No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Cattle weight estimation using active contour models and regression trees Bagging
Abstract
Monitoring the weight of beef cattle is important for productive strategies. The main goal of this work was to automatically extract measurements from 2D images of the dorsal area of Nellore cattle to estimate the weight of these cattle using regression algorithms. For this purpose, Euclidean distances from points generated by the Active Contour Model, together with features obtained from the dorsal Convex Hull, were selected. These were submitted to Bagging, Regression by Discretization and Random Forest algorithms for analysis of the predicted error metrics. The Bagging algorithm showed the best results, with Mean Absolute Error (MAE) of 13.44 kg (±2.76), Square Root of the Mean Error (RMSE) of 15.88 kg (±2.86), Mean Absolute Percentage Error (MAPE) of 2.27% and correlation coefficient at 0.75.
... The cattle farming industry has benefits as a food source, livelihood, economic contribution, environmental land restoration, and energy source [2]. Cattle are an essential resource that contributes to agricultural practices, supports biodiversity conservation, facilitates research, and has cultural significance [3]. ...
... The highest beef production based on market value is at prime age, namely between 18 months and 24 months, and has reached optimal weight, and beef muscle mass has developed. In the context of the livestock buying and selling business, it is to help farmers make better decisions regarding selling, purchasing livestock, managing feed, health services, and efficient livestock maintenance [3]. The need for sacrificial animals for Eid al-Adha 2023 is estimated to reach 1.7 million. ...
... The strength of Random Forest lies in its capacity to reduce overfitting, increase model stability, and offer practical solutions in various classification and prediction scenarios [9]. Many industries involving random forest techniques in advanced data processing, including bioinformatics, finance, health, and others, have effectively used this approach [3]. Random forests can offer new perspectives in the investigation of predictive models, support the reliability of research, and offer reliable answers to problems posed by the complexity of modern data [8]. ...
The global cattle farming industry has benefits as a food source, livelihood, economic contribution, land environmental restoration, and energy source. The importance of predicting cow weight for farmers is to monitor animal development. Meanwhile, for traders, knowing the animal's weight makes it easier to calculate the price of the animal meat they buy. The authors propose estimating cattle weighting linear regression and random forest regression. Linear regression can interpret the linear relationship between dependent and independent variables, and random forest regression can generalize the data well. The data set used in this study consisted of ten variables: live body weight, withers height, sacrum height, chest depth, chest width, maclocks width, hip joint width, oblique body length, oblique back length and chest circumference. Find the model that produces the smallest MAE value. The results show that the linear regression algorithm can produce estimated weight values for cattle with the best performance. This model produces a mean absolute error (MAE) of 0.35 kg, a mean absolute percentage error (MAPE) of 0.07%, a root mean square error (RMSE) of 0.5 kg, and an R² of 0.99. Each variable has excellent correlation performance results and contributes to computer vision and machine learning.
... Our MAE demonstrates superiority over the findings in [10]. Although [4] shows a better MAE compared to ours, it is important to note that their experiments were conducted in a feeding fence system, whereas our experiments operated in real farm environments. Table 10 presents a performance comparison between our approach and previous methodologies. ...
... In terms of MAPE, our approach outperforms all other methods. Regarding MAE, this metric is used only in the studies by Ruchay et al. [10] and Weber et al. [4]. Our MAE demonstrates superiority over the findings in [10]. ...
... Our MAE demonstrates superiority over the findings in [10]. Although [4] shows a better MAE compared to ours, it is important to note that their experiments were conducted in a feeding fence system, whereas our experiments operated in real farm environments. ...
Accurate weight measurement is critical for monitoring the growth and well-being of cattle. However, the traditional weighing process, which involves physically placing cattle on scales, is labor-intensive and stressful for the animals. Therefore, the development of automated cattle weight prediction techniques assumes critical significance. This study proposes a weight prediction approach for Korean cattle using 3D segmentation-based feature extraction and regression machine learning techniques from incomplete 3D shapes acquired from real farm environments. Firstly, we generated mesh data of 3D Korean cattle shapes using a multiple-camera system. Subsequently, deep learning-based 3D segmentation with the PointNet network model was employed to segment 3D mesh data into two dominant parts: torso and center body. From these segmented parts, the body length, chest girth, and chest width of Korean cattle were extracted. Finally, we implemented five regression machine learning models (CatBoost regression, LightGBM, polynomial regression, random forest regression, and XGBoost regression) for weight prediction. To validate our approach, we captured 270 Korean cattle in various poses, totaling 1190 poses of 270 cattle. The best result was achieved with mean absolute error (MAE) of 25.2 kg and mean absolute percent error (MAPE) of 5.85% using the random forest regression model.
... Definition of the body of animals can be collected in 2D and 3D using still images or video analysis to determine LW, animal performance, and body composition in live animals (Gomes et al., 2016;Song et al., 2018;Weber et al., 2020). This recent increase in the use of image analysis was particularly triggered by computer-aided analyses such as machine learning (Miller et al., 2019). ...
... Studies reported modest to high correlations between body measures obtained from 2D and 3D imagery and LW in beef and dairy cattle Ozkaya et al., 2016;Song et al., 2018;Stajnko et al., 2008;Tasdemir et al., 2011;Weber et al., 2020). Ozkaya et al. (2016) predicted LW from body area (R 2 = 0.61) and the accuracy of such predictions were improved (R 2 = 0.88) when combined with body measures such as wither height, body length, and chest girth in Limousine cattle. ...
... Ozkaya et al. (2016) predicted LW from body area (R 2 = 0.61) and the accuracy of such predictions were improved (R 2 = 0.88) when combined with body measures such as wither height, body length, and chest girth in Limousine cattle. Similarly, Weber et al. (2020) estimated LW (R 2 = 0.75) from 2D images of the dorsal area of confined Nellore (Bos indicus cattle). The latter authors suggested that automatic segmentation techniques and previous LW data from animals (e.g., static weighing operations) could improve the accuracy of LW prediction. ...
Pig farming systems face an increasingly diversified challenge to consider simultaneously the economic, environmental, and social pillars of sustain ability. For animal nutrition, this requires the development of smart feeding strategies able to integrate these different dimensions in a dynamic way and to be adapted as much as possible to each individual animal. These developments can be supported by digital technologies including data collection and processing, decision making and automation of applications. Classical traits such as feed intake and growth benefit from new technologies that can be measured more frequently. New sensors can be indicative for other traits related to body composition, physiological status, activity, feed efficiency, or rearing environment. A challenge for data collection is to obtain information on a large number of animals and with sufficient frequency, quality, and precision and use it cost-effectively. Another challenge is to analyse the ever-increasing volume of data and use it in decision-making. Nutritional models for pigs and sows, classically mechanistic, have to evolve to integrate real-time data. With the development of data-driven modelling methods (e.g., machine-learning or deep-learning), a synergy between mechanistic models and data-driven approaches is required in smart pig nutrition. Moreover, the practical application of smart pig nutrition must consider the evolution in pig farming systems towards increased diversity in terms of size, space allowance, and outdoor access, and return on investment. Finally, the transition of pig nutrition in the digital era must consider the social acceptance of an increasing role of digital technologies in animal production systems.KeywordsActivityArtificial intelligenceAutomatonConcept-driven modellingData collectionData-driven modellingData processingDecision support systemFattening pigsFeed efficiencyFeed intakeGestating sowHealth statusLactating sowMineralNutritionNutritional requirementsPerformancePhysiological statusPig farming systemPrecision feedingRearing environmentSensors
... For the state-of-the-art models, a comparison with our proposal was performed (shown in Table 6). Weber et al. (2020) extracted measurements from 2D images of the dorsal area of Nellore cattle to estimate the BW using machine learning algorithms, which reported the BW of both steers and heifers, with the R 2 of 75%, MAPE of 2.27% and RMSE of 15.88 kg. Our study achieved an RMSE of 10.2 kg compared to 15.88 kg reached by Weber et al. (2020), which is an improvement of 35%. ...
... Weber et al. (2020) extracted measurements from 2D images of the dorsal area of Nellore cattle to estimate the BW using machine learning algorithms, which reported the BW of both steers and heifers, with the R 2 of 75%, MAPE of 2.27% and RMSE of 15.88 kg. Our study achieved an RMSE of 10.2 kg compared to 15.88 kg reached by Weber et al. (2020), which is an improvement of 35%. Nonetheless, we achieved a MAPE of 3.2% compared to 2.27% reached by their study, which is an improvement of 29%. ...
... Image Type Method Environment [7] cattle 2D segmentation + convex hull, random forest regression fence system [8] cow 2D ANN Regression - [9] pig 2D ANN Regression - [10] cattle 2D gabor filter, fuzzy logic - [11] cow 3D segmentation fence system [12] pig 3D segmentation, linear regression - [13] cattle 3D segmentation, linear and non-linear regression - [14] cow 3D Lasso regression fence system [15] heifer 3D ellipse fitting, linear regression narrow passage [16] cow 3D linear regression - [17] pig 3D linear and non-linear regression - [18] cow 3D full-body scan linear regression special scanning station [19] cow thermal linear regression - [20] pig stereo vision least squares regression fence system [21] calf stereo vision linear regression - [22] calf stereo vision linear regression - [23] heifer 2D deep learning-based image processing and regression - [24] pig 3D deep learning-based image processing and regression - [25] pig 2D deep learning-based image processing and regression - ...
... There are several studies in the literature that are based on image processing techniques on 2D images. In a study by Weber et al., the live body weight of cattle was estimated using dorsal area images taken from above using a kind of fence system [7]. Their system first performs segmentation and then generates a convex hull around the segmented area to obtain features to feed a Random Forest-based regression model. ...
In cattle breeding, regularly taking the animals to the scale and recording their weight is important for both the performance of the enterprise and the health of the animals. This process, which must be carried out in businesses, is a difficult task. For this reason, it is often not performed regularly or not performed at all. In this study, we attempted to estimate the weights of cattle by using stereo vision and semantic segmentation methods used in the field of computer vision together. Images of 85 animals were taken from different angles with a stereo setup consisting of two identical cameras. The distances of the animals to the camera plane were calculated by stereo distance calculation, and the areas covered by the animals in the images were determined by semantic segmentation methods. Then, using all these data, different artificial neural network models were trained. As a result of the study, it was revealed that when stereo vision and semantic segmentation methods are used together, live animal weights can be predicted successfully.
... Alternatively, body weight can be predicted both manually and digitally (Weber et al., 2020a;Wang et al., 2021). In the manual method, data can be collected on independent variables such as body measurement, using measuring tape and sticks, facilitating the prediction of the dependent variable namely body weight. ...
This study aims to develop and validate a model for predicting the body weight (BW) of Ongole Crossbred (OC) cattle using body measurements. To achieve this, a combination of meta-analysis and field experiments was employed. The meta-analysis involved identifying relevant keywords and databases, reviewing titles and abstracts, extracting data, and subsequently tabulating and analyzing the data. A total of 1,141 animal records were included in the quantitative synthesis process. Following the meta-analysis, a BW prediction model for OC cattle was developed. The model incorporated recommendations obtained from the meta-analysis, considering body measurement, age, and sex. Data from 507 animals were utilised to construct the model. Finally, a field experiment was conducted on 35 animals to assess the accuracy of the model. The meta-analysis revealed that body volume (BV) (r=0.96) and heart girth (HG) (r=0.89) exhibited stronger correlations with BW compared to body length (BL) (r=0.68). Linear regression modeling of OC cattle BW, demonstrated that HG yielded high correlation coefficients for both male (r=0.98) and female (r=0.94) cattle. Similarly, BV showed strong correlations for male (r=0.99) and female (r=0.95) cattle. Furthermore, the analysis revealed that both HG and BV were effective predictors across different age groups, with high correlation coefficients observed for cattle aged 1-12 months and over 24 months. The field experiment confirmed the high reliability of the model, achieving an accuracy of 90.8% for HG and 91% for BV. In conclusion, HG and BV are strong predictors of OC cattle BW, with categorization by breed further improving prediction accuracy.
... As estimativas podem ser melhoradas pelo uso de algoritmos de aprendizado de máquina (Dohmen et al., 2022), porém algoritmos adequados para extração de informações relevantes das imagens ainda são raros (Barbedo et al., 2020). Weber et al. (2020b) no Brasil mostraram que pode ser estimado o peso corporal de bovinos Girolando e Nelore, com coeficientes de correlação de 0,71 a 0,75, por medidas corporais extraídas de imagens desses bovinos. ...
Automation, digital agriculture (DA), and precision livestock (PL) technologies can contribute to farms’ efficient and sustainable management. Sensors, equipment, images, actuators, systems for monitoring edaphoclimatic variability, pastures and animals, and inputs can collect, store, and provide data from systems
(intensive or integrated) of meat and milk production. The conversion and integration of climate, soil, pastures, and animals’ raw data and their interactions into useful information create opportunities for more efficient use of natural resources and production factors, reducing the risk associated with the activity and increasing the sustainability of animal production in pasture systems. Due to a large amount of data and information generated from different sources, there is a need for technologies and methods for standardization and processing through
ICT (information and communication technologies). The development of technological solutions based on the
application of DA and PL to pastures and animals can contribute robust systems to support decision-making in production systems based on pastures.
As tecnologias de automação, agricultura digital (AD) e pecuária de precisão (PP) podem contribuir para o manejo eficiente e sustentável de propriedades rurais. Os sensores, equipamentos, imagens, atuadores, sistemas para monitoramento da variabilidade edafoclimática, das pastagens e dos animais e do uso dos insumos podem ser utilizados para a coleta, armazenamento e disponibilização de dados de sistemas (intensivos ou integrados) de produção de carne e leite. A conversão e integração de dados brutos sobre clima, solo, pastagens, animais e
suas interações em informações úteis cria oportunidades para o uso mais eficiente dos recursos naturais e dos
fatores de produção, reduzindo o risco associado à atividade e aumentando a sustentabilidade dos sistemas de produção animal em pastagens. Devido à grande quantidade de dados e informações geradas de diferentes fontes, há a necessidade de tecnologias e métodos para padronização e processamento por meio de TIC (tecnologias de informação e comunicação). O desenvolvimento de soluções tecnológicas baseadas na aplicação de AD e PP às pastagens e animais podem contribuir para o desenvolvimento de sistemas robustos de apoio à tomada de decisão nos sistemas de produção baseado em pastagens.
... With pigs, a body weight estimation and measurement of local body sizes can be made based on a point cloud of the pig's back using a convolutional neural network with multiple heads of attention [9]. Many other researchers have used different systems of extracting feature information from single-view images for the automated measurement of livestock bodies [10][11][12][13][14]. However, there are limitations to the single-view angle measurement method. ...
Cattle farming is an important part of the global livestock industry, and cattle body size is the key indicator of livestock growth. However, traditional manual methods for measuring body sizes are not only time-consuming and labor-intensive but also incur significant costs. Meanwhile, automatic measurement techniques are prone to being affected by environmental conditions and the standing postures of livestock. To overcome these challenges, this study proposes a multi-view fusion-driven automatic measurement system for full-attitude cattle body measurements. Outdoors in natural light, three Zed2 cameras were installed covering different views of the channel. Multiple images, including RGB images, depth images, and point clouds, were automatically acquired from multiple views using the YOLOv8n algorithm. The point clouds from different views undergo multiple denoising to become local point clouds of the cattle body. The local point clouds are coarsely and finely aligned to become a complete point cloud of the cattle body. After detecting the 2D key points on the RGB image created by the YOLOv8x-pose algorithm, the 2D key points are mapped onto the 3D cattle body by combining the internal parameters of the camera and the depth values of the corresponding pixels of the depth map. Based on the mapped 3D key points, the body sizes of cows in different poses are automatically measured, including height, length, abdominal circumference, and chest circumference. In addition, support vector machines and Bézier curves are employed to rectify the missing and deformed circumference body sizes caused by environmental effects. The automatic body measurement system measured the height, length, abdominal circumference, and chest circumference of 47 Huaxi Beef Cattle, a breed native to China, and compared the results with manual measurements. The average relative errors were 2.32%, 2.27%, 3.67%, and 5.22%, respectively, when compared with manual measurements, demonstrating the feasibility and accuracy of the system.
... ACMs effectively segment images by grayscale information, structural information, and other prior information. ACMs have exerted a tremendous fascination in industrial testing [7,8], imaging medicine [9,10], monitoring and safety [11], and other fields due to its robustness to uneven brightness and poorly-defined edge. ...
... This information includes the animal's identification, such as RFID or NLISID, type, and description. Identification of the animal is essential for tracking the movements The weight observation (Weber et al., 2020) in the units specified (usually kilogrammes). 2 Score Body condition scoring is a management score designed to assess an animal's body reserves or fat accumulation (Qiao et al., 2021). ...
... The link between the weights and body size measurements (features) of cows is investigated through the use of machine-learning-based data analysis. Weber et al. [16] used regression methods to automatically extract measures from images of the dorsal area of Nellore cattle in order to determine the cattle's weight. Euclidean distances from locations produced by the active contour model were chosen for this purpose by the authors together with characteristics gleaned from the dorsal Convex Hull. ...
We investigate the impact of different data modalities for cattle weight estimation. For this purpose, we collect and present our own cattle dataset representing the data modalities: RGB, depth, combined RGB and depth, segmentation, and combined segmentation and depth information. We explore a recent vision-transformer-based zero-shot model proposed by Meta AI Research for producing the segmentation data modality and for extracting the cattle-only region from the images. For experimental analysis, we consider three baseline deep learning models. The objective is to assess how the integration of diverse data sources influences the accuracy and robustness of the deep learning models considering four different performance metrics: mean absolute error (MAE), root mean squared error (RMSE), mean absolute percentage error (MAPE), and R-squared (R2). We explore the synergies and challenges associated with each modality and their combined use in enhancing the precision of cattle weight prediction. Through comprehensive experimentation and evaluation, we aim to provide insights into the effectiveness of different data modalities in improving the performance of established deep learning models, facilitating informed decision-making for precision livestock management systems.
... In accordance with [1], the authors implemented a method for automatically extracting measurements to estimate the weight of Nellore cattle based on regression algorithms using 2D images of the dorsal area. Additionally, the use of depth images along with an algorithm for automatically estimating heifer height and body mass for cattle, as presented in [2], has demonstrated that in single-view measurement methods utilizing a single RGB camera or depth camera for body condition and body size characteristics evaluation, challenges persist in obtaining multi-scale information, such as chest girth, abdominal circumference, rump angle, and so on. ...
This paper introduces an approach to the automated measurement and analysis of dairy cows using 3D point cloud technology. The integration of advanced sensing techniques enables the collection of non-intrusive, precise data, facilitating comprehensive monitoring of key parameters related to the health, well-being, and productivity of dairy cows. The proposed system employs 3D imaging sensors to capture detailed information about various parts of dairy cows, generating accurate, high-resolution point clouds. A robust automated algorithm has been developed to process these point clouds and extract relevant metrics such as dairy cow stature height, rump width, rump angle, and front teat length. Based on the measured data combined with expert assessments of dairy cows, the quality indices of dairy cows are automatically evaluated and extracted. By leveraging this technology, dairy farmers can gain real-time insights into the health status of individual cows and the overall herd. Additionally, the automated analysis facilitates efficient management practices and optimizes feeding strategies and resource allocation. The results of field trials and validation studies demonstrate the effectiveness and reliability of the automated 3D point cloud approach in dairy farm environments. The errors between manually measured values of dairy cow height, rump angle, and front teat length, and those calculated by the auto-measurement algorithm were within 0.7 cm, with no observed exceedance of errors in comparison to manual measurements. This research contributes to the burgeoning field of precision livestock farming, offering a technological solution that not only enhances productivity but also aligns with contemporary standards for sustainable and ethical animal husbandry practices.
... Instead, the method integrating area and height takes advantage of threedimensional size information of cows and becomes an accurate and reliable measurement, which is also consistent with the favorite indicators of experienced farmers for weight estimation artificially. For this purpose, besides the reference cards and image processing software (Ozkaya and Bozkurt, 2008;Weber et al., 2020a), different computer vision methods have been attempted to calculate the body areas and height including the Euclidean distances (Weber et al., 2020b), EfficientNet, ResNet, Recurrent Attention Model (Gjergji et al., 2020). Moreover, considering the strong correlation between the body parameters from the images and cattle weight, the regression-based machine learning methods, for instance, multiple linear regression (MLR) (Freund et al., 2006), support vector machine (SVM) (Boser et al., 1992), backpropagation (BP) neural network (Hakem et al., 2022) were used to predict the body weight. ...
Accurate prediction of cattle weight is essential for enhancing the efficiency and sustainability of livestock management practices. However, conventional methods often involve labor-intensive procedures and lack instant and non-invasive solutions. This study proposed an intelligent weight prediction approach for cows based on semantic segmentation and Back Propagation (BP) neural network. The proposed semantic segmentation method leveraged a hybrid model which combined ResNet-101-D with the Squeeze-and-Excitation (SE) attention mechanism to obtain precise morphological features from cow images. The body size parameters and physical measurements were then used for training the regression-based machine learning models to estimate the weight of individual cattle. The comparative analysis methods revealed that the BP neural network achieved the best results with an MAE of 13.11 pounds and an RMSE of 22.73 pounds. By eliminating the need for physical contact, this approach not only improves animal welfare but also mitigates potential risks. The work addresses the specific needs of welfare farming and aims to promote animal welfare and advance the field of precision agriculture.
... The scope is to achieve an integrated vision between production characteristics, animal welfare, and security issues. In particular, the adoption of new types of approaches for weight assessment aims to increase the accuracy of measurement and, accordingly, to improve the monitoring of animal performance, thus potentially providing major benefits to both the herdsmen and the animals in their care [6], [22], [23]. ...
A brief overview is presented of the main findings related to the role of Industry 4.0 (in terms of smart computing and sensing technologies) for the Precision Grazing, as strictly connected to the principles of both Precision Agriculture and Precision Livestock Farming.
... The scope is to achieve an integrated vision between production characteristics, animal welfare, and security issues. In particular, the adoption of new types of approaches for weight assessment aims to increase the accuracy of measurement and, accordingly, to improve the monitoring of animal performance, thus potentially providing major benefits to both the herdsmen and the animals in their care [6]- [8]. ...
Precision Livestock Farming, as a specific sub-sector of Public Health Informatics, focuses on the application of process engineering
principles and techniques to achieve an automatic monitoring, modelling, and management of animal productions. In the present work
a timely "protocol" is proposed for unobtrusive direct/indirect monitoring of biometric parameters for the estimation of body conditions
on Mediterranean Buffalo populations, using low-cost automated systems already present on the market i.e., smart cameras endowed
with depth perception capabilities
... They captured sheep images from top view, created masks of the top view, and measured six distances in the mask as features to feed a random forest regression model. Weber et al. [4] proposed a cattle weight estimation approach using active contour and regression trees bagging. They first segmented the image, then created a hull from the segmented image, then extracted features, and predicted weight using a random forest model. ...
Accurate weight measurement is pivotal for monitoring the growth and well-being of cattle. However, the conventional weighing process, which involves physically placing cattle on scales, is labor-intensive and distressing for the animals. Hence, the development of automated cattle weight prediction techniques assumes critical significance. This study proposes a weight prediction approach for Korean cattle using 3D segmentation-based feature extraction and regression machine learning techniques from incomplete 3D shapes acquired from real farm environments. In the initial phase, we generated mesh data of 3D Korean cattle shapes using a multiple-camera system. Subsequently, deep learning-based 3D segmentation with the PointNet network model was employed to segment two dominant parts of the cattle. From these segmented parts, three crucial dimensions of Korean cattle were extracted. Finally, we implemented five regression machine learning models (CatBoost regression, LightGBM, Polynomial regression, Random Forest regression, and XGBoost regression) for weight prediction. To validate our approach, we captured 270 Korean cattle in various poses, totaling 1190 poses of 270 cattle. The best result was achieved with mean absolute error (MAE) of 25.2 kg and mean absolute percent error (MAPE) of 5.81% using the random forest regression model.
... Table 3 summarises 34 events captured by the LEI schema, along with their definitions. The weight observation (Weber et al., 2020) in the units specified (usually kilograms). ...
Data-driven advances have resulted in significant improvements in dairy production. However, the meat industry has lagged behind in adopting data-driven approaches, underscoring the crucial need for data standardisation to facilitate seamless data transmission to maximise productivity, save costs, and increase market access. To address this gap, we propose a novel data schema, Livestock Event Information (LEI) schema, designed to accurately and uniformly record livestock events. LEI complies with the International Committee for Animal Recording (ICAR) and Integrity System Company (ISC) schemas to deliver this data standardisation and enable data sharing between producers and consumers. To validate the superiority of LEI, we conducted a structural metrics analysis and a comprehensive case study. The analysis demonstrated that LEI outperforms the ICAR and ISC schemas in terms of design, while the case study confirmed its superior ability to capture livestock event information. Our findings lay the foundation for the implementation of the LEI schema, unlocking the potential for data-driven advances in livestock management. Moreover, LEI's versatility opens avenues for future expansion into other agricultural domains, encompassing poultry, fisheries, and crops. The adoption of LEI promises substantial benefits, including improved data accuracy, reduced costs, and increased productivity, heralding a new era of sustainability in the meat industry.
... Extra Trees Regression has several advantages in predicting the body weight of cattle. It can handle both numerical and categorical features, can capture non-linear relationships, and requires less computational resources compared to some other complex regression algorithms (Biase et al., 2022b;Wang et al., 2021a;Weber et al., 2020) RF regression uses a large number of decision trees constructed using random subsets of the training data. Each decision tree is built by selecting a random subset of features at each split. ...
In order to forecast and model the weight of cattle a number of techniques have been used. Nonetheless, no machine algorithm has been utilized to estimate the weight of Bali cattle. This article examines the use of machine learning regression to create models for Bali cattle's body weight prediction. The response variables consist of body weight as the dependent variable and body length, girth circumference, height at wither of 228 male and 211 female cattle of similar ages (285 days). The descriptive statistics of female Bali cattle in our investigation revealed that the morphological measurements were similar to those documented by other researchers. To predict body weight on the basis of different characteristics, machine learning models such as Random Forest, Support Vector, K-Neighbors, and Extra Tree regressions have been used. Additionally, linear regression was utilized to estimate the body weight for comparison with the traditional approach. The assessment standards used included the determination coefficient, the root mean square error, the average absolute error, and the average absolute percentage error as measures of evaluation efficiency. We found that Linear Regression performs the best among all the regressors for female cattle. Similarly for male, it is about the same as extra tree regression. The machine learning algorithm (MLA) was discovered to furnish more precise estimate of the weight of the body cattle, surpassing the conventional algorithm.
... As a result, a mean square error (MSE) of 1046.0 was obtained. Weber et al. [8] obtained a mean absolute percentage error (MAPE) of 2.27% using a manually created segmentation mask to find the distances between points on the outline of the segmentation mask, and estimated the cattle weight using those values. Seo et al. [9] used cattle top-view and side-view data and achieved an error rate of 5% to 10.7% using multiple regression equations. ...
Weight information is important in cattle breeding because it can measure animal growth and be used to calculate the appropriate amount of daily feed. To estimate the weight, we developed an image-based method that does not stress cattle and requires no manual labor. From a 2D image, a mask was obtained by segmenting the animal and background, and weights were estimated using a deep neural network with residual connections by extracting weight-related features from the segmentation mask. Two image segmentation methods, fully and weakly supervised segmentation, were compared. The fully supervised segmentation method uses a Mask R-CNN model that learns the ground truth mask generated by labeling as the correct answer. The weakly supervised segmentation method uses an activation visualization map that is proposed in this study. The first method creates a more precise mask, but the second method does not require ground truth segmentation labeling. The body weight was estimated using statistical features of the segmented region. In experiments, the following performance results were obtained: a mean average error of 17.31 kg and mean absolute percentage error of 5.52% for fully supervised segmentation, and a mean average error of 35.91 kg and mean absolute percentage error of 10.1% for the weakly supervised segmentation.
... Many studies evaluate body weight using measurements in several animal species, such as buffalo, sheep, dogs, cattle, goats, rabbits and camels [14,[17][18][19][20][21][22][23][24][25][26][27]. ...
Simple Summary: This study aimed to estimate body weight from various biometric measurements and features such as genotype (share of Suffolk and Polish Merino genotypes), birth weight (BiW), sex, birth type and body weight at 12 months of age (LBW) and some body measurements such as withers height (WH), sacrum height (SH), chest depth (CD), chest width (CW), chest circumference (CC), shoulder width (SW) and rump width (RW). Three hundred and forty-four animals were used in the study. Data mining and machine learning algorithms such as Random Forest Regression, Support Vector Regression and classification and regression tree were used to estimate the body weight from various features. Results show that the random forest procedure may help breeders improve characteristics of great importance. In this way, the breeders can get an elite population and determine which features are essential for estimating the body weight of the herd in Poland. Abstract: The study's main goal was to compare several data mining and machine learning algorithms to estimate body weight based on body measurements at a different share of Polish Merino in the genotype of crossbreds (share of Suffolk and Polish Merino genotypes). The study estimated the capabilities of CART, support vector regression and random forest regression algorithms. To compare the estimation performances of the evaluated algorithms and determine the best model for estimating body weight, various body measurements and sex and birth type characteristics were assessed. Data from 344 sheep were used to estimate the body weights. The root means square error, standard deviation ratio, Pearson's correlation coefficient, mean absolute percentage error, coefficient of determination and Akaike's information criterion were used to assess the algorithms. A random forest regression algorithm may help breeders obtain a unique Polish Merino Suffolk cross population that would increase meat production.
... There has been a surge in the use of these methods in the agriculture domain [252][253][254][255][256][257][258][259]. Recent research has been directed towards using bagging [260][261][262][263] and boosting [198]. Ensembles of NNs have been investigated in [264][265][266][267][268][269][270]. ...
Rapid advancements in technology, particularly in soil tools and agricultural machinery, have led to the proliferation of mechanized agriculture. The interaction between such tools/machines and soil is a complex, dynamic process. The modeling of this interactive process is essential for reducing energy requirements, excessive soil pulverization, and soil compaction, thereby leading to sustainable crop production. Traditional methods that rely on simplistic physics-based models are not often the best approach. Computational intelligence-based approaches are an attractive alternative to traditional methods. These methods are highly versatile, can handle various forms of data, and are adaptive in nature. Recent years have witnessed a surge in adapting such methods in all domains of engineering, including agriculture. These applications leverage not only classical computational intelligence methods, but also emergent ones, such as deep learning. Although classical methods have routinely been applied to the soil-machine interaction studies, the field is yet to harness the more recent developments in computational intelligence. The purpose of this review article is twofold. Firstly, it provides an in-depth description of classical computational intelligence methods, including their underlying theoretical basis, along with a survey of their use in soil-machine interaction research. Hence, it serves as a concise and systematic reference for practicing engineers as well as researchers in this field. Next, this article provides an outline of various emergent methods in computational intelligence, with the aim of introducing state-of-the-art methods to the interested reader and motivating their application in soil-machine interaction research.
... For example, a selected set of pretrained convolutional neural networks (CNN) are retrained to detect various kind of insects [8]. Weber et al. [9] used data from 2D images to estimate cow's weight using active contour models and regression trees Bagging. Lee et al. [10] employed object detection and segmentation to estimate tomato's weight. Foglia et al. [3] used simple image processing techniques to identify and localize radicchios in robotic harvesting application. ...
—In agricultural robotics, to perform automatic harvesting task, it is important that the size of thetarget can be reasonably approximated.Stereo cameras can be used as a sensor to detect the targeted object and approximate its size. In this paper we present two methods for estimating the size of mangoes: pinhole model and ellipsoid model methods, using data obtained from a stereo camera.Object detection scheme using Mask-RCNN deep learning neural network on an RGB image and point cloud data from stereo camera are employed. The pinhole model method assumes simple triangular projection of the object onto the image plane.It makes use of the bounding boxes predicted by the neural network and the 3-D point cloud data. The ellipsoid model method, on the other hand, uses three-dimensional point cloud data obtained by processing the images from the stereo camera.It assumes that mango shape can be approximated as an ellipsoid and uses the predicted masked region and its corresponding point cloud data to fit an ellipsoid whose center and axes are extracted to estimate the position and the size of the mango, respectively. In the scope of this study where the full shape of mango can be seen in the image view, results from experiments suggest that the simpler pinhole model can effectively estimate the size of the mango with greater accuracy than the ellipsoid model. Keywords—Object Detection; Agricultural Robotics; Artificial Intelligent
... Non-contact based body measurement methods for the livestock can be mainly divided into two types: single-view measurement methods and multi-view measurement methods. Single-view measurement methods use a single depth camera or RGB camera to obtain images at a specific angle of view, and body size information was obtained through image processing (Pallottino et al., 2015;Rodríguez Alvarez et al., 2018;Nir et al., 2018;Zhang et al., 2019;Shi et al., 2019;Weber et al., 2020). The single-view measurement methods evaluate body condition and body size characteristics through local information of the livestock (Rodríguez Alvarez et al., 2019). ...
Body measurement plays an important role in the breeding and production of dairy cows. Body measurement is mainly done manually, which is laborious, imprecise, and prone to stressful behavior in animals. A portable non-contact 3D measurement system of dairy cow body using smartphones was proposed in this study. Reconstruction of 3D point clouds scene containing the cow through a Structure-from-Motion (SfM) photogrammetry, the RANdom SAmple Consensus (RANSAC), and the Euclidean clustering were used to segment the scene and extract cow’s point clouds. Aiming to solve the holes in the cow point cloud caused by the barrier and legs occlusion, a completion method based on the spline curve with a smoothing factor was proposed. The automatic measurement of body size was realized through morphological features. The accuracy of the proposed system was evaluated in a commercial dairy farm. Compared with mesh-based method, the proposed point cloud hole completion method reduced the maximum relative errors of chest girth and chest width by 2.22% and 2.18%. Under different cow postures, the average relative error between automatic body size measurement algorithm and manual calibration was less than 4.67%. The average relative error between the measured value of the non-contact measurement system and the actual value of withers height, body length, chest girth, and chest width were 2.41%, 3.18%, 4.37%, and 6.12%, respectively. The results showed that the proposed method can be used as an automatic and non-contact approach for measuring animal body size.
... Many studies have been reported about the prediction of body weight from body measurements in different animal species such as sheep [7][8][9] , cattle [10,11] , rabbit [12] , dog [13] and camel [14] . In the literature, there are many practical approaches to estimate body weight for sheep breeds by body measurements within the scope of multiple regression [15] , Classification and Regression Tree (CART) and Chi-square automatic interaction detection (CHAID) and MARS algorithm [6] and artificial neural networks [16] . ...
In this study, it is aimed to compare several data mining and artificial neural network algorithms to predict body weight from biometric
measurements for the Th alli sheep breed. For this purpose, the prediction capabilities of Bayesian Regularized Neural Network (BRNN), Support
Vector Regression (SVR), Random Forest Regression (RFR) and Multivariate Adaptive Regression Splines (MARS) algorithms were comparatively
investigated. To measure the predictive performances of the evaluated algorithms, body measurements such as body length, heart girth, ear length,
ear width, head width, head length, withers height, rump length, rump width neck length, neck width of Th alli sheep were used for predicting
the body weight. In this context, 270 female Th alli sheep were used to predict body weight. Model comparison criteria such as root-mean square
error (RMSE), standard deviation ratio (SDR), performance index (PI), global relative approximation error (RAE), mean absolute percentage error
(MAPE), Pearson"s correlation coefficient (r), determination of coefficient (R2) and Akaike"s information criteria (AIC) were used to compare all
algorithms. In conclusion, the MARS algorithm can be recommended to enable breeders to obtain an elite population of Th alli sheep breed.
... Using a balanced distribution of weak and hard data, which makes the data set, difficult instances are identified by out-of-bag handlers, so that when a sample is considered "hard" it is incorrectly classified by the ensemble. This hard data is always added to the next data set while easy data has little chance of getting into the dataset [20,[81][82][83]. Performance of the Bagging based hybrid models developed in this study is slightly better than other ML models such as LASSO (R2 = 0.911), Random Forest (R2 = 0.936) and SVM (R2 = 0.935) carried out by Nguyen et al. [13] on Mekong River. ...
Water level predictions in the river, lake and delta play an important role in flood management. Every year Mekong River delta of Vietnam is experiencing flood due to heavy monsoon rains and high tides. Land subsidence may also aggravate flooding problems in this area. Therefore, accurate predictions of water levels in this region are very important to forewarn the people and authorities for taking timely adequate remedial measures to prevent losses of life and property. There are so many methods available to predict the water levels based on historical data but nowadays Machine Learning (ML) methods are considered the best tool for accurate prediction. In this study, we have used surface water level data of 18 water level measurement stations of the Mekong River delta from 2000 to 2018 to build novel time-series Bagging based hybrid ML models namely: Bagging (RF), Bagging (SOM) and Bagging (M5P) to predict historical water levels in the study area. Performances of the Bagging-based hybrid models were compared with Reduced Error Pruning Trees (REPT), which is a benchmark ML model. The data of 19 years period was divided into 70:30 ratio for the modeling. The data of the period 1/2000 to 5/2013 (which is about 70% of total data) was used for the training and for the period 5/2013 to 12/2018 (which is about 30% of total data) was used for testing (validating) the models. Performance of the models was evaluated using standard statistical measures: Coefficient of Determination (R2), Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). Results show that the performance of all the developed models is good (R2 > 0.9) for the prediction of water levels in the study area. However, the Bagging-based hybrid models are slightly better than another model such as REPT. Thus, these Bagging-based hybrid time series models can be used for predicting water levels at Mekong data.
... ith the development of precision livestock farming (PLF) in modern agriculture, videobased intelligent management, such as automatic detection, live weight estimation, noncontact welfare measurement, and behavior monitoring and analysis, has been widely used (Gjergji et al., 2020;Weber et al., 2020;Li et al., 2021;Qiao et al., 2021;Yang et al., 2020;Guo et al., 2021). In particular, the realization of intelligent detection and classification of animal behavior is a frequent topic (Zhao et al., 2018). ...
Highlights
BiGRU-attention based cow behavior classification was proposed.
Key spatial-temporal features were captured for behavior representation.
BiGRU-attention achieved >82% classification accuracy on calf and adult cow datasets.
The proposed method could be used for similar animal behavior classification.
Abstract . Animal behavior consists of time series activities, which can reflect animals’ health and welfare status. Monitoring and classifying animal behavior facilitates management decisions to optimize animal performance, welfare, and environmental outcomes. In recent years, deep learning methods have been applied to monitor animal behavior worldwide. To achieve high behavior classification accuracy, a BiGRU-attention based method is proposed in this article to classify some common behaviors, such as exploring, feeding, grooming, standing, and walking. In our work, (1) Inception-V3 was first applied to extract convolutional neural network (CNN) features for each image frame in videos, (2) bidirectional gated recurrent unit (BiGRU) was used to further extract spatial-temporal features, (3) an attention mechanism was deployed to allocate weights to each of the extracted spatial-temporal features according to feature similarity, and (4) the weighted spatial-temporal features were fed to a Softmax layer for behavior classification. Experiments were conducted on two datasets (i.e., calf and adult cow), and the proposed method achieved 82.35% and 82.26% classification accuracy on the calf and adult cow datasets, respectively. In addition, in comparison with other methods, the proposed BiGRU-attention method outperformed long short-term memory (LSTM), bidirectional LSTM (BiLSTM), and BiGRU. Overall, the proposed BiGRU-attention method can capture key spatial-temporal features to significantly improve animal behavior classification, which is favorable for automatic behavior classification in precision livestock farming. Keywords: BiGRU, Cow behavior, Deep learning, LSTM, Precision livestock farming.
... Bootstrap aggregating (Ba) is employed to enlarge inconsistency/instability extents as well as classification plots. Bagging has demonstrated by evidence or argument to be true or existing to be very sensitive to highlight the variations in training data that is contributory to boost the categorization precision of incipient intention tree classifier by decreasing variance of categorization wrong (Weber et al. 2020). The points to be made about the Ba algorithm note that the punctuality of the single ML algorithm is not high, so the principal ML algorithm is repeated several times to enhancement the prediction precision, as well as the final precision of the model using the results. ...
In this study, two novel hybrid models namely Bagging-based Rough Set (BRS) and AdaBoost-based Rough Set (ABRS) were used to generate landslide susceptibility maps of Son La hydropower reservoir basin, Vietnam. In total, 186 past landslide events and twelve landslides affecting factors (slope degree, slope aspect, elevation, curvature, focal flow, river density, rainfall, aquifer, weathering crust, lithology, fault density and road density) were considered in the modeling study. The landslide data was split into training (70%) and testing (30%) for the model’s development and validation. One R feature selection method was used to select and prioritize the landslide affecting factors based on their importance in model prediction. Performance of the hybrid developed models was evaluated and also compared with single rough set (RS) and support vector machine (SVM) models using various standard statistical measures including area under the curve (AUC)-receiver operating characteristics (ROC) curve. The results show that the developed hybrid model BRS (AUC = 0.845) is the most accurate model in comparison to other models (ABRS, SVM and RS) in predicting landslide susceptibility. Therefore, the BRS model can be used as an effective tool in the development of an accurate landslide susceptibility map of the hilly area.
... The R 2 values from testing the developed models by multiple regression, partial least square regression, and an artificial neural network with seven significant variables were 0.91, 0.91, and 0.92, respectively (relative errors were within 4%). Weber et al. (2020b) measured cattle dorsal area from 2D Image and then predicted cattle body weight using active contour models and regression trees Bagging, which achieved Mean Absolute Error of 13.44 kg. ...
There has been an increasing demand for animal protein due to several factors such as global population growth, rising incomes, etc. However, farming productivity is stagnating due to a mix of traditional practice, climate change, socioeconomic , and environmental phenomena. Precision livestock farming, with intelligent perception tools at its core, and vast amounts of data being acquired from different sensors or platforms, has the ability to analyse individual animal for improved management, and the potential to dramatically enhance farm productivity. In order to facilitate research and promote the development of related areas, this review summarises and analyses the main existing techniques used in precision cattle farming, focusing on those related to identification, body condition score evaluation, and live weight estimation. More than 100 relevant papers have been discussed in a cohesive manner. From this review and extensive discussions of recent trends, we anticipate that intelligent perception for precision cattle farming will develop through non-contact, high precision, automated technologies, combined with emerging 3D model reconstruction and deep learning technologies. Existing challenges and future research opportunities will also be highlighted and discussed.
The aim of this study was to define the growth by using nonlinear functions in intensive feedlot system with XGBoost algorithm. To achieve this aim, five nonlinear functions were implemented. To implementation of the study, Brown Swiss (n=41) and Simental (n=95) breed were used. Each nonlinear functions were examined for each breed. According to the results of the nonlinear functions, logistic model was the best prediction model for defining the growth of each breed. In this study, the parameters in the best prediction model were calculated individually and the relationship of these parameters with body weight was evaluated with the XGBoost algorithm. Model comparison criteria such as standard deviation ratio (SDratio), Pearson’s correlation coefficient (PC), determination of coefficient (R2) and Akaike’s information criteria (AIC) were used to evaluate the XGBoost algorithm. In conclusion, the XGBoost algorithm can be an effective and optional approach that allows breeders to estimate live weight from growth parameters. This algorithm can operate on large data sets with high accuracy and speed, leading to significant improvements in agricultural productivity and animal health management. XGBoost enables more accurate predictions by analyzing the effects of various characteristics (e.g., nutritional level, breed, age). Therefore, this method can be used to determine critical parameters such as body weight in animal breeding practices, serving as a powerful support tool for operational decisions.
This paper presents a method for the measurement of morphological features of horses that is suitable for mobile applications and requires only a smartphone camera to perform measurements with a short preparation time. This study compares, various body lengths of 46 warmblood horses. The 3D-coordinates were determined photogrammetrically i.e.), using overlapping photographs and video recordings. To measure the relevant points, 12 markers were affixed at significant points to the left body side of the horses. A local coordinate system was defined by using a mobile reference frame to ensure that the measurements can be collected independently of local conditions and still are directly comparable. The measurements were repeated three times and analyzed for their accuracy by various statistical methods. With an average standard deviation of 1.75 cm for the three repeated body length measurements, the method is very reliable. The repetitions of the body length measurements showed that body parts with a high mobility, especially the measurements from the neck to the withers, have a significant impact on the attainable accuracy. To determine the absolute accuracy of the photogrammetric measurement method, a single horse was also measured with a terrestrial laser scanner (TLS). The standard deviation of the two measurements was between 0.24 cm and 1.79 cm, which demonstrates the high absolute accuracy of the measurement method.
Objective: This study aimed to develop a method for predicting the body weight of beef cattle using meta-analysis based on digital image processing.
Materials and Methods: The meta-analysis process commenced by collecting studies with the keywords “beef cattle,” “correlation,” “digital image,” and “body weight” from Google Scholar and Science Direct. The obtained studies were reviewed papers based on their titles, abstracts, and content, and then categorized by authors, year, country, sample size, and correlation coefficient. A digital image of body measurements used included wither and hip height, chest depth, heart girth, body length, and top view. The statistical analysis was conducted by calculating effect sizes using the correlation coefficient and sample sizes.
Results: The results of the meta-analysis, based on 3,017 cattle from 13 selected studies, showed the highest and lowest correlation coefficients for the top view variable and hip height. Based on cattle breed, significant differences (p < 0.05) were observed in the wither height variable with correlation coefficients of 0.94, 0.79, and 0.66 for Hanwoo, Holstein, and Simmental, respectively. Based on sex, significant differences (p < 0.05) were seen in the wither height variable, with cor-
relation coefficients of 0.73 for males and 0.90 for females, while for hip height, the values were 0.70 and 0.87, respectively.
Conclusion: In conclusion, to achieve the best accuracy in predicting the body weight of beef cattle based on a digital image, the top view variable can be used. However, for ease of field experimentation, body length or chest depth can also be used while taking breed and sex categories into the model.
In numerous systems of animal production, there is increasing interest in the use of 3D imaging technology on farms for its ability to easily and safely measure traits of interest on living animals. With this information, it is possible to evaluate multiple morphological indicators of interest, either directly or indirectly, and follow them through time. Several tools for this purpose were developed, but one of their main weaknesses was their sensitivity to light and animal movement, which limited their potential for large-scale application on farms. To address this, a new device, called Deffilait3D and based on depth camera technology, was developed. In tests on 31 Holstein dairy cows and 13 Holstein heifers, the values generated for most measured indicators were highly repeatable and reproducible, with coefficients of variation lower than 4%. A comparison of measurements obtained from both Deffilait3D and previous validated system, called Morpho3D, revealed a high degree of similarity for most selected traits, e.g., less than 0.2% variation for animal volume and 1.2% for chest depth, with the highest degree of difference (8%) noted for animal surface area. Previously published equations used to estimate body weight with the Morpho3D device were equally valid using Deffilait3D. This new device was able to record 3D images regardless of the movement of animals and it is affected only by direct daylight. The on-going step is now to develop methods for automated analysis and extraction from images, which should enable the rapid development of new tools and potentially lead to the large-scale adoption of this type of device on commercial farms.
1
Summary
Changes in body mass are a key indicator of health and disease in humans and model organisms. Animal body mass is routinely monitored in husbandry and preclinical studies. In rodent studies, the current best method requires manually weighing the animal on a balance which has at least two consequences. First, direct handling of the animal induces stress and can have confounding effects on studies. Second, the acquired mass is static and not amenable to continuous assessment, and rapid mass changes can be missed. A noninvasive and continuous method of monitoring animal mass would have utility in multiple areas of biomedical research. Here, we test the feasibility of determining mouse body mass using video data. We combine computer vision methods with statistical modeling to demonstrate the feasibility of our approach. Our methods determine mouse mass with 4.8% error across highly genetically diverse mouse strains, with varied coat colors and mass. This error is low enough to replace manual weighing with image-based assessment in most mouse studies. We conclude that visual determination of rodent mass using video enables noninvasive and continuous monitoring and can improve animal welfare and preclinical studies.
Recently, there has been an increase in the popularity of breeding insect larvae (Tenebrio Molitor and Hermetia Illucens). Dimensioning larvae and observing their growth over time is a key component of monitoring insect larvae breeding. Due to the high number of larvae in the analysed images (dense scenes) and their overlap, determining the size distribution of larvae in real-time is a research challenge. In this work, we proposed an efficient method for determining the size distribution of larvae based on a regression convolutional neural network (RegCNN) and knowledge transfer. Larval width was chosen as the main measured larval parameter due to its ease of registration in dense scenes. The larval length L and its volume V were determined indirectly using determined regression models L(width) and V(width). RegCNN training was performed using knowledge transfer to omit the time-consuming labelling of multiple images containing larvae at different growth stages. Training used quartiles (lower quartile, median, upper quartile) of larval widths determined using improved multistage larvae phenotyping based on classical computer vision methods and larvae segmentation model. Finally, our approach required labelling only a few images for calibration purposes. The study evaluated different RegCNN architectures: pre-trained on ImageNet (ResNet, EfficientNet) and custom with a reduced number of model parameters. The proposed method was validated for the distribution of larvae characterised by width quartiles taking values from 1.7 mm to 3.1 mm, corresponding to an average larval length of 16 mm to 28 mm. For the best evaluated model (ResNet18) in larval width estimation, we obtained RMSE=0.131 mm (average RMSE=1.12 mm for larval length estimation) and R^2=0.870 (coefficient of determination) with an average inference time of 0.30 s/box. The best proposed custom architecture (TenebrioRegCNN_v3) achieved slightly lower accuracy (RMSE=0.134 mm, R^2=0.864) with about five times lower inference time per image than ResNet18. The quantitative results confirmed the proposed method's potential to be applied in real breeding conditions.
The nutritional management of grazing livestock in extensive conditions is challenging because of the difficulty to measure diet selection, feed and nutrient intake, and excretion, energy expenditure, and ultimately animal performance. The large variability in space and time of weather, pasture characteristics, and animal requirements and performance pose an additional challenge. However, various sensor technologies exist today to measure key attributes related to feed availability and quality, energy and nutrient requirements, animal performance, and environmental footprint in near real-time. Requirements for maintenance are a function of body weight (LW), physical activities, and environmental conditions, which can be measured using automatic systems for LW determination, animal behaviour, and weather. Requirements for production (body growth, gestation, and lactation) can be measured directly or indirectly via automatic LW determination, and technologies for the detection of oestrus and birthing events. Feed efficiency could be measured using face masks, heart rate monitors, and open-circuit gas-quantification systems of gas exchanges (CO2, CH4, and O2). Finally, mathematical nutrition models play a very important roll to integrate these technologies and predict hard to measure variables. Examples of such automatic model-data fusion approach are presented to demonstrate its potential as part of smart nutrition systems of extensively kept livestock. The combination of data collected automatically using digital technologies, data analytics, and mathematical prediction models have the potential to revolutionize animal nutrition of extensively reared livestock. This will improve productivity, animal welfare, and the sustainability of these systems.KeywordsAnimal ProductionBig dataCattleDigitalFeed supplementationExtensive animal productionGrazingPasturesRangelandRuminantsSensorsSheepSimulation modelsSmart NutritionSustainabilityTechnologiesWelfare
Airlines overestimate the weight of their passengers by simply assigning a constant weight for everyone, causing each plane to burn more fuel than needed to carry the extra weight. Accurately estimating the passenger weights is a difficult problem for airlines as naively weighing all passengers with scales is impractical in already busy airports. Hence, we propose CamScale, a novel vision-based weight inference system that is augmented by an off-the-shelf viscoelastic mat (e.g., memory foam mat). CamScale takes the video feed of the mat placed on the floor as the passengers walk over it. It utilizes the inherent strain, or deformation of the mat due to the passengers' footsteps to infer their weights. CamScale is advantageous because it does not incur additional weighing time, while being cost-effective and accurate. We evaluate CamScale through real-world experiments by deploying RGB and infrared cameras and inviting 36 participants to walk a total of more than 17,000 steps over viscoelastic mats, equivalent to walking approximately 13.1 km. We demonstrate that CamScale is able to accurately estimate an individual's weight with an average error of 1.12 kg.
Introduction: The use of automation and sensor-based systems in livestock production allows monitoring of individual cows in real-time and provides the possibility of early warning systems to take necessary management actions against possible anomalies. Among the different RT monitoring parameters, body weight (BW) plays an important role in tracking the productivity and health status.
Methods: In this study, various supervised learning techniques representing different families of methods in the machine learning space were implemented and compared for performance in the prediction of body weight from 3D image data in dairy cows. A total of 83,011 records of contour data from 3D images and body weight measurements taken from a total of 914 Danish Holstein and Jersey cows from 3 different herds were used for the predictions. Various metrics including Pearson’s correlation coefficient (r), the root mean squared error (RMSE), and the mean absolute percentage error (MAPE) were used for robust evaluation of the various supervised techniques and to facilitate comparison with other studies. Prediction was undertaken separately within each breed and subsequently in a combined multi-breed dataset.
Results and discussion: Despite differences in predictive performance across the different supervised learning techniques and datasets (breeds), our results indicate reasonable prediction accuracies with mean correlation coefficient (r) as high as 0.94 and MAPE and RMSE as low as 4.0 % and 33.0 (kg), respectively. In comparison to the within-breed analyses (Jersey, Holstein), prediction using the combined multi-breed data set resulted in higher predictive performance in terms of high correlation coefficient and low MAPE. Additional tests showed that the improvement in predictive performance is mainly due to increase in data size from combining data rather than the multi-breed nature of the combined data. Of the different supervised learning techniques implemented, the tree-based group of supervised learning techniques (Catboost, AdaBoost, random forest) resulted in the highest prediction performance in all the metrics used to evaluate technique performance. Reported prediction errors in our study (RMSE and MAPE) are one of the lowest in the literature for prediction of BW using image data in dairy cattle, highlighting the promising predictive value of contour data from 3D images for BW in dairy cows under commercial farm conditions.
Weight prediction in live animals remains challenging. Several studies have been carried out trying to predict the body weight in livestock through morphometric measurements, the Schaeffer's model is one of them. However, the fit of those studies in small ruminants is not well covered. Therefore, a novel model to predict the weight of Pelibuey sheep through morphometric measurements and the Gray Wolf Optimizer algorithm is presented. The model involves calculating the volume of the specimen through a truncated cone and leaving density as an estimation parameter of the algorithm. Also, two alternative models were made where the original Schaeffer's model was optimized. The modified models from the original Schaeffer's formula showed improvements up to 22.61% in R-squared and decreases up to 33.48% in RMSE. However, the truncated cone model had the best estimates, with an RMSE of 2.57, R-squared of 89.02%, and the lowest AIC. This represented a 25.13% improvement in R-squared and a 38.31% reduction in the RMSE. The model is expected to improve its efficiency if the cattle sample is larger, and it is also intended to be implemented in animals of other proportions.
In animal husbandry, it is of great interest to determine and control the key factors that affect the production characteristics of animals, such as milk yield. In this study, simplified selective tree-based ensembles were used for modeling and forecasting the 305-day average milk yield of Holstein-Friesian cows, depending on 12 external traits and the farm as an environmental factor. The preprocessing of the initial independent variables included their transformation into rotated principal components. The resulting dataset was divided into learning (75%) and holdout test (25%) subsamples. Initially, three diverse base models were generated using Classifiction and Regression Trees (CART) ensembles and bagging and arcing algorithms. These models were processed using the developed simplified selective algorithm based on the index of agreement. An average reduction of 30% in the number of trees of selective ensembles was obtained. Finally, by separately stacking the predictions from the non-selective and selective base models, two linear hybrid models were built. The hybrid model of the selective ensembles showed a 13.6% reduction in the test set prediction error compared to the hybrid model of the non-selective ensembles. The identified key factors determining milk yield include the farm, udder width, chest width, and stature of the animals. The proposed approach can be applied to improve the management of dairy farms.
Some scientific researches have been conducted recently based on Artificial Intelligence (AI) to solve animal welfare and health related problems. However, no review study was conducted to assess the potential of AI in solving the diverse application problems in several animal species including ruminants, pig and poultry. To fill in this gap, this paper provides a systematic review on scientific research advancement in AI related animal farming based on 131 research publications over the past four decades. It was shown that the number of related studies has increased significantly since 2016, and the most intensive studies were focused on animal behavior detection and recognition, and concentrated mostly on farm animal species of pig (37.95%), cattle (37.44%), and poultry (16.92%); Also, most scientific research in animal farming driven by sensors and AI models were focused on data collection, processing, assessment, and analysis in the areas of animal behavior detection, disease monitoring, growth estimation, and environment monitoring at the experimental stage; Moreover, some technical challenges on AI like the accuracy and cost need to be improved before it could come into use in the commercial animal farming.
Share Link:
https://authors.elsevier.com/a/1eCGB3QCo9bRvZ
Morphological characteristics are still the most used tools for the identification of plant species. In this context, leaves are the most available plant organ used, given their perenniality and diversity. Computer-based image analysis help extract morphological features for botanical identification and maybe a solution to taxonomic problems requiring extensively trained specialists that use visual identification as the primary method for this approach. In this study, were collected 40 leaves from 30 trees and shrub species from 19 different families. Here, we compared two popular image capture devices: a scanner and a mobile phone. Features analyzed comprised color, shape, and texture. The performance of both devices was compared through three machine learning algorithms (adaptive boosting—AdaBoost, random forest, support vector machine—SVM) and an artificial neural network model (deep learning). Computer vision showed to be efficient in the identification of species (higher than 93%), with similar results obtained for both mobile phones and scanners. The algorithms SVM, random forest and deep learning performed more efficiently than AdaBoost. Based on the results, we present the Inovtaxon Plant Species Identification Software, available at https://github.com/DeborahBambil/Inovtaxon.
The objective with this study was to analyze the body measurements of Girolando cattle, as well as measurements extracted from their images, to generate a model to understand which measures further explain the cattle body weight. Therefore, the experiment physically measured 34 Girolando cattle (two males and 32 females), for the following traits: heart girth (HG P), circumference of the abdomen, body length, occipito-ischial length, wither height, and hip height. In addition, images of the dorsum and the body lateral area of these animals allowed measurements of hip width (HW I), body length, tail distance to the neck, dorsum area (DA I), dorsum perimeter, wither height, hip height, body lateral area, perimeter of the lateral area, and rib height. The measurements extracted from the images were subjected to the stepwise regression method and regression-based machine learning algorithms. The HG p was the physical measure with stronger positive correlation with respect to body weight. In the stepwise method, the final model generated R² of 0.70 and RMSE of 42.52 kg and the equation: WEIGHT (kg) = 6.15421 * HW I (cm) + 0.01929 * DA I (cm 2) + 70.8388. The linear regression and SVM algorithms obtained the best results, followed by discretization regression with random forests. The set of rules presented in this study can be recommended for estimating body weight in Girolando cattle, at a correlation coefficient of 0.71, by measurements of hip width and dorsum area, both extracted from cattle images.
The accurate and reliable counting of animals in quadcopter acquired imagery is one of the most promising but challenging tasks in intelligent livestock management in the future. In this paper we demonstrate the application of the cutting-edge instance segmentation framework, Mask R-CNN, in the context of cattle counting in different situations such as extensive production pastures and also in intensive housing such as feedlots. The optimal IoU threshold (0.5) and the full-appearance detection for the algorithm in this study are verified through performance evaluation. Experimental results in this research show the framework’s potential to perform reliably in offline quadcopter vision systems with an accuracy of 94% in counting cattle on pastures and 92% in feedlots. Compared with the existing typical competing algorithms, Mask R-CNN outperforms both in the counting accuracy and average precision especially on the datasets with occlusion and overlapping. Our research shows promising steps towards the incorporation of artificial intelligence using quadcopters for enhanced management of animals.
Large-scale phenotyping of animal behaviour traits is time consuming and has led to increased demand for technologies that can automate these procedures. Automated tracking of animals has been successful in controlled laboratory settings, but recording from animals in large groups in highly variable farm settings presents challenges. The aim of this review is to provide a systematic overview of the advances that have occurred in automated, high throughput image detection of farm animal behavioural traits with welfare and production implications. Peer-reviewed publications written in English were reviewed systematically following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. After identification, screening, and assessment for eligibility, 108 publications met these specifications and were included for qualitative synthesis. Data collected from the papers included camera specifications, housing conditions, group size, algorithm details, procedures, and results. Most studies utilized standard digital colour video cameras for data collection, with increasing use of 3D cameras in papers published after 2013. Papers including pigs (across production stages) were the most common (n = 63). The most common behaviours recorded included activity level, area occupancy, aggression, gait scores, resource use, and posture. Our review revealed many overlaps in methods applied to analysing behaviour, and most studies started from scratch instead of building upon previous work. Training and validation sample sizes were generally small (mean±s.d. groups = 3.8±5.8) and in data collection and testing took place in relatively controlled environments. To advance our ability to automatically phenotype behaviour, future research should build upon existing knowledge and validate technology under commercial settings and publications should explicitly describe recording conditions in detail to allow studies to be reproduced.
Automated high-throughput phenotyping with sensors, imaging, and other on-farm technologies has resulted in a flood of data that are largely under-utilized. Drastic cost reductions in sequencing and other omics technology have also facilitated the ability for deep phenotyping of livestock at the molecular level. These advances have brought the animal sciences to a cross-roads in data science where increased training is needed to manage, record, and analyze data to generate knowledge and advances in Agriscience related disciplines. This paper describes the opportunities and challenges in using high-throughput phenotyping, “big data,” analytics, and related technologies in the livestock industry based on discussions at the Livestock High-Throughput Phenotyping and Big Data Analytics meeting, held in November 2017 (see: https://www.animalgenome.org/bioinfo/community/workshops/2017/). Critical needs for investments in infrastructure for people (e.g., “big data” training), data (e.g., data transfer, management, and analytics), and technology (e.g., development of low cost sensors) were defined by this group. Though some subgroups of animal science have extensive experience in predictive modeling, cross-training in computer science, statistics, and related disciplines are needed to use big data for diverse applications in the field. Extensive opportunities exist for public and private entities to harness big data to develop valuable research knowledge and products to the benefit of society under the increased demands for food in a rapidly growing population.
For the time-consuming and stressful body measuring task of Qinchuan cattle and farmers, the demand for the automatic measurement of body dimensions has become more and more urgent. It is necessary to explore automatic measurements with deep learning to improve breeding efficiency and promote the development of industry. In this paper, a novel approach to measuring the body dimensions of live Qinchuan cattle with on transfer learning is proposed. Deep learning of the Kd-network was trained with classical three-dimensional (3D) point cloud datasets (PCD) of the ShapeNet datasets. After a series of processes of PCD sensed by the light detection and ranging (LiDAR) sensor, the cattle silhouettes could be extracted, which after augmentation could be applied as an input layer to the Kd-network. With the output of a convolutional layer of the trained deep model, the output layer of the deep model could be applied to pre-train the full connection network. The TrAdaBoost algorithm was employed to transfer the pre-trained convolutional layer and full connection of the deep model. To classify and recognize the PCD of the cattle silhouette, the average accuracy rate after training with transfer learning could reach up to 93.6%. On the basis of silhouette extraction, the candidate region of the feature surface shape could be extracted with mean curvature and Gaussian curvature. After the computation of the FPFH (fast point feature histogram) of the surface shape, the center of the feature surface could be recognized and the body dimensions of the cattle could finally be calculated. The experimental results showed that the comprehensive error of body dimensions was close to 2%, which could provide a feasible approach to the non-contact observations of the bodies of large physique livestock without any human intervention.
In order to accurately count the number of animals grazing on grassland, we present a livestock detection algorithm using modified versions of U-net and Google Inception-v4 net. This method works well to detect dense and touching instances. We also introduce a dataset for livestock detection in aerial images, consisting of 89 aerial images collected by quadcopter. Each image has resolution of about 3000×4000 pixels, and contains livestock with varying shapes, scales, and orientations. We evaluate our method by comparison against Faster RCNN and Yolo-v3 algorithms using our aerial livestock dataset. The average precision of our method is better than Yolo-v3 and is comparable to Faster RCNN.
Thermal stress is the result of a misbalance between heat produced or gained from the environment and the amount of heat lost to the environment. The level of thermal stress can range from minor or no effect to death of vulnerable animals. Under summertime conditions, thermal stress results in hyperthermia or heat stress. Heat stress in feedlot cattle is a common summertime occurrence in cattle-producing parts of the world (USA, Australia, Brazil, etc.). Effects on animals experiencing heat stress include decreases in feed intake, animal growth, and production efficiency. During these extreme events, animal losses can exceed 5% of all cattle on feed in a single feedlot. Luckily, these extreme events are generally very localized and last only a day or two. However, these losses can be devastating to individual producers within the affected area. The level of heat stress an individual animal will experience is a result of a combination of three distinct components: environmental conditions, individual animal susceptibility, and management of the herd. Environmental components include temperature, humidity, wind speed, and solar radiation. Several indices have been developed to summarize the different components into a single value. Individual animal susceptibility is influenced by many different factors including coat color, sex, temperament, previous health history, acclimation, and condition score. Finally, management greatly influences the effects of thermal stress. Management factors can be broken into four distinct categories: feed, water, environmental influences, and handling. Understanding these risk factors and how each one influences animal stress will aid in the development of management strategies and how to implement them. Management strategies that can be employed at the right time and to the correct groups of animals will increase benefits to the animals and limit costs for the producers.
Information on the body shape of pigs is a key indicator to monitor their performance and health and to control or predict their market weight. Manual measurements are among the most common ways to obtain an indication of animal growth. However, this approach is laborious and difficult, and it may be stressful for both the pigs and the stockman. The present paper proposes the implementation of a Structure from Motion (SfM) photogrammetry approach as a new tool for on-barn animal reconstruction applications. This is possible also to new software tools allowing automatic estimation of camera parameters during the reconstruction process even without a preliminary calibration phase. An analysis on pig body 3D SfM characterization is here proposed, carried out under different conditions in terms of number of camera poses and animal movements. The work takes advantage of the total reconstructed surface as reference index to quantify the quality of the achieved 3D reconstruction, showing how as much as 80% of the total animal area can be characterized.
Keywords: pig barn; automatic measurement; animal weight; three-dimensional reconstruction; structure from motion; precision livestock farming.
The use of unmanned aerial systems (UASs) in agriculture has been growing steadily in the last decade, but their use to monitor and count cattle has been very limited. This article analyses the reasons for this apparent lack of progress, considering both the technical challenges and the difficulties in defining target users who would benefit from a UAS-based system for monitoring cattle. Such an analysis is combined with the findings reported in several investigations dedicated to counting and monitoring wildlife to draw a comprehensive picture on the current situation, to suggest possible solutions to technical issues and to delineate applications that could be useful to both cattle farmers and governments. The text concludes by showing that there are unexplored viable uses for UAS in livestock monitoring, especially in countries like Brazil, where extensive stockbreeding prevails.
Identification of individual livestock such as pigs and cows has become a pressing issue in recent years as intensification practices continue to be adopted and precise objective measurements are required (e.g. weight). Current best practice involves the use of RFID tags which are time-consuming for the farmer and distressing for the animal to fit. To overcome this, non-invasive biometrics are proposed by using the face of the animal. We test this in a farm environment, on 10 individual pigs using three techniques adopted from the human face recognition literature: Fisherfaces, the VGG-Face pre-trained face convolutional neural network (CNN) model and our own CNN model that we train using an artificially augmented data set. Our results show that accurate individual pig recognition is possible with accuracy rates of 96.7% on 1553 images. Class Activated Mapping using Grad-CAM is used to show the regions that our network uses to discriminate between pigs.
Lameness in dairy cattle is a persistent problem, indicating pain caused by underlying disease states and is associated with reduced milk yields. Digital dermatitis is a common cause of lameness. Thermal imaging is a technique that may facilitate early detection of this disease and has the potential for use in automated detection systems. Previous studies with thermal imaging have imaged either the heels or the coronary band of the foot and typically only used the maximum temperature (Max) value as the outcome measure. This study investigated the utility of other statistical descriptors: 90th percentile (90PCT), 95th percentile (95PCT), standard deviation (SD) and coefficient of variation (CoV) and compared the utility of imaging the heel or coronary band. Images were collected from lame and healthy cows using a high-resolution thermal camera. Analyses were done at the cow and foot level. There were significant differences between lame and healthy feet detectable at the heels (95th percentile: P < 0.05; SD: P < 0.05) and coronary band (SD: P < 0.05). Within lame cows, 95PCT values were higher at the heel (P < 0.05) and Max values were higher at the coronary band (P < 0.05) in the lame foot compared to the healthy foot. ROC analysis showed an AUC value of 0.72 for Max temperature and 0.68 for 95PCT at the heels. It was concluded that maximum temperature is the most accurate measure, but other statistical descriptors of temperature can be used to detect lameness. These may be useful in certain contexts, such as where there is contamination. Differentiation of lame from healthy feet was most apparent when imaging the heels.
Information on the daily growth rate of pigs enables the stockman to monitor their performance and health and to predict and control their market weight and date. Manual measurements are among the most common ways to get an indication of animal growth. However, this approach is laborious and difficult, and it may be stressful for both the pigs and the stockman. As a consequence, manual measurements can be very time consuming, induce costs and sometimes cause injuries to the animals and the stockman. The present work proposes the implementation of a Microsoft Kinect v1 depth camera for the fast, non-contact measurement of pig body dimensions such as heart girth, length and height. In the present work, these dimension values were related to animal weight, and two models (linear and non-linear) were developed and applied to the Kinect and manual measurement data. Both models were highly correlated with the direct weight measurements considered as references, as demonstrated by high coefficients of determination (R2 > 0.95). Specifically, in the case of the non-linear model based on non-contact depth camera measurements, the mean absolute error exhibited a reduction of over 40% compared to the same non-linear model based on manual measurements (from 0.82 to 0.48 kg).
Frequent checks on livestock's body growth can help reducing problems related to cow infertility or other welfare implications, and recognizing health's anomalies. In the last ten years, optical methods have been proposed to extract information on various parameters while avoiding direct contact with animals' body, generally causes stress. This research aims to evaluate a new monitoring system, which is suitable to frequently check calves and cow's growth through a three-dimensional analysis of their bodies' portions. The innovative system is based on multiple acquisitions from a low cost Structured Light Depth-Camera (Microsoft Kinect™ v1). The metrological performance of the instrument is proved through an uncertainty analysis and a proper calibration procedure. The paper reports application of the depth camera for extraction of different body parameters. Expanded uncertainty ranging between 3 and 15 mm is reported in the case of ten repeated measurements. Coefficients of determination R 2 > 0.84 and deviations lower than 6% from manual measurements where in general detected in the case of head size, hips distance, withers to tail length, chest girth, hips, and withers height. Conversely, lower performances where recognized in the case of animal depth (R 2 = 0.74) and back slope (R 2 = 0.12).
The objectives of this study were to quantify the error of body weight prediction using automatically measured morphological traits in a 3-dimensional (3-D) vision system and to assess the influence of various sources of uncertainty on body weight prediction. In this case study, an image acquisition setup was created in a cow selection box equipped with a top-view 3-D camera. Morphological traits of hip height, hip width, and rump length were automatically extracted from the raw 3-D images taken of the rump area of dairy cows (n = 30). These traits combined with days in milk, age, and parity were used in multiple linear regression models to predict body weight. To find the best prediction model, an exhaustive feature selection algorithm was used to build intermediate models (n = 63). Each model was validated by leave-one-out cross-validation, giving the root mean square error and mean absolute percentage error. The model consisting of hip width (measurement variability of 0.006 m), days in milk, and parity was the best model, with the lowest errors of 41.2 kg of root mean square error and 5.2% mean absolute percentage error. Our integrated system, including the image acquisition setup, image analysis, and the best prediction model, predicted the body weights with a performance similar to that achieved using semi-automated or manual methods. Moreover, the variability of our simplified morphological trait measurement showed a negligible contribution to the uncertainty of body weight prediction. We suggest that dairy cow body weight prediction can be improved by incorporating more predictive morphological traits and by improving the prediction model structure.
Since animals express their internal state through behaviour, changes in said behaviour may be used to detect early signs of problems, such as in animal health. Continuous observation of livestock by farm staff is impractical in a commercial setting to the degree required to detect behavioural changes relevant for early intervention. An automated monitoring system is developed; it automatically tracks pig movement with depth video cameras, and automatically measures standing, feeding, drinking, and locomotor activities from 3D trajectories. Predictions of standing, feeding, and drinking were validated, but not locomotor activities. An artificial, disruptive challenge; i.e., introduction of a novel object, is used to cause reproducible behavioural changes to enable development of a system to detect the changes automatically. Validation of the automated monitoring system with the controlled challenge study provides a reproducible framework for further development of robust early warning systems for pigs. The automated system is practical in commercial settings because it provides continuous monitoring of multiple behaviours, with metrics of behaviours that may be considered more intuitive and have diagnostic validity. The method has the potential to transform how livestock are monitored, directly impact their health and welfare, and address issues in livestock farming, such as antimicrobial use.
The objective of this study was to validate an electronic system for monitoring individual feeding and drinking behaviour and intake developed for young cattle housed in group. A total of 35 Holstein-Gyr crossbred heifers (BW: 180±52 kg; age: 121.5±32.5 days), fitted with an ear tag containing a unique passive transponder, were distributed in three groups of 12, 12 and 11 animals per period and had free access to 12 electronic feed bins and two electronic water bins (Intergado® Ltd). The dimensions of feed and water bins, as well as the sensors position were appropriate for young cattle. The system documented the visit frequency and duration, as well as the feed and water intakes, by recording the animal's identification tag, bin number, initial and final times of visits and the difference of feed/water weight at the start and end of each bin visit. Feed bins were monitored using time-lapse video recording over 4 days and the water bins were monitored over 6 days. For each feed bin, two feeding events were monitored using manual weighings with an external scale immediately before and after the animal's visit and the difference between them was assumed as feed intake (n=24 observations). For the water bins, 60 manual weighings were made. Video and manual weighing data were regressed on the electronic feeding and drinking behaviour and intake data to evaluate the system's precision and accuracy. The system showed high specificity (98.98% and 98.56% for the feed and water bins, respectively) and sensitivity (99.25% and 98.74%, respectively) for identifying an animal's presence or absence. Duration of feed and water bin visits as well as feed and water consumption per visit estimated by the system were highly correlated and precise compared with the observed video and manual weighing data (r 2=0.917, 0.963, 0.973 and 0.986, respectively). It was concluded that Intergado® system is a useful tool for monitoring feeding and drinking behaviour as well as water and feed intakes in young cattle housed in groups.
This study was realized to evaluate prediction models for body weight (BW) using body measurements as heart girth (HG), wither height (WH), hip width (HW), body length (BL) and hip height (HH), and develop new Equations to predict the BW of Holstein and crossbred heifers. Prediction Equations (1, 2, 3, 4, 5 and 6) proposed in previous studies were tested, while body measurements, which were more significant in terms of predicting BW, were used to obtain new Equations. Heart girth was the body measurement with the greater correlation with BW. Equation 4 presented the best prediction results, showing a low average deviation, high accuracy and precision. Equations 1 and 2 predicted with a high presence of bias, while 2 had the lowest value of the concordance correlation coefficient and a great contribution made by systematic error. The 3 model presented a high variability in predicting BW. For both Holstein and crossbred heifers, the Equations evaluated are inadequate for predicting BW. Among the Equations evaluated, the model using HG presented the best adjustment and the greater adjusted coefficient of determination. Body measurements relating to HG and BL were the most significant variables in predicting the BW, resulting in the following Equation: BW = - 372.89 + 2.8072 × HG + 1.6087 × BL. © 2017, Eduem - Editora da Universidade Estadual de Maringa. All rights reserved.
This paper presents a study on the use of low resolution infrared images to detect ticks in cattle. Emphasis is given to the main factors that influence the quality of the captured images, as well as to the actions that can increase the amount of information conveyed by these images. In addition, a new automatic method for analyzing the images and counting the ticks is introduced. The proposed algorithm relies only on color transformations and simple mathematical morphology operations, thus being easy to implement and computationally light. Tests were carried out using a large database containing images of the neck and hind end of the animals. It was observed that the proposed algorithm is very effective in detecting ticks visible in the images, even if the contrast with the background is not high. On the other hand, due to both intrinsic and extrinsic factors, the thermographic images used in this study did not always succeed in creating enough contrast between ticks and cattle's hair coat. Although these problems can be mitigated by following some directives, currently only rough estimates for tick counts can be achieved using infrared images with low spatial resolution.
Missed, swapped, false insurance claimed and reallocation of cattle at slaughter houses is global problems throughout the world. Only few researches have been done so far to solve such major problems. Traditional identification approaches and non-biometrics techniques have such severe problems due to their own boundaries and limitations for the registration and traceability of cattle. These techniques are not able to provide a competent level of security to livestock/cattle. Therefore, cattle identification needs an innovative research for the development of efficient, scalable, affordable, non-invasive, automatic recognition systems for better registration, traceability and security of livestock/cattle. In this paper, an attempt has been made to solve the above problems by using computer vision approaches for cattle recognition using their facial images. The major contributions of this research are in three folds: firstly, the preparations of a facial images database of cattle, secondly, the extraction of set of discriminatory features from the database and implementation of computer vision based face recognition algorithms for recognizing cattle and finally, the experimental results and discussion of face recognition algorithms. Thus, this paper presents a comprehensive review of the performances of various computer vision and pattern recognition approaches for the application of cattle face recognition.
According to FAO organization, by 2050 annual meat production should be risen by over 200 M tones to reach 470 M tones as the world population will reach 9.1 billion (34% higher than today). So, there is a need for controlling safety policies of animals and efficient management of food production. One way to help achieve this need is the automatic animal identification/identification and traceability systems. In this paper, two classifiers fusion using Gabor feature extraction technique and the notion of features and classifiers fusion. Gabor features are extracted from three different scales of muzzle print images. Two different levels of fusion are then used, i.e., feature fusion and classifier fusion, to accurately identify animal individuals using three different classifiers (support vector machine – SVM, k-nearest neighbor, and minimum distance classifier). The experimental results show that, the proposed two approaches are robust and accurate in comparing them with the existed works as the proposed approaches achieve 99.5% identification accuracy. In addition, the results prove that the features fusion-based model achieved accuracy better than the classifier fusion-based model.
The objective of this study was to validate an electronic system for monitoring individual feeding behavior and feed intake (Intergado Ltd., Contagem, Minas Gerais, Brazil) in freestall-housed dairy cattle. No data have been published that validate either the behavioral measurement or the feed intake of this system. Feeding behavior data were recorded for 12 Holstein cows over 5 d using an Intergado system and time-lapse video. The cows were fitted with an ear tag containing a unique passive transponder and provided free access to 12 feed bins. The system documented the visit duration and feed intake by recording the animal identification number, bin number, initial and final times, and the difference between feed weight at start and end of each feed bin visit. These data were exported to Intergado web software and reports were generated. Electronic data on animal behavior were compared with video data collected during the same evaluation period. An external scale was used to manually measure and validate the electronic system's ability to monitor dairy cow feed intake for each feed bin visit. The feed intake was manually measured for 4-h time periods and compared with the sum of the feed intake recorded by the monitoring system for each cow visit during the same time period. Video and manual weight data were regressed on the electronic feeding behavior and feeding intake data to evaluate the precision of the monitoring system. The Intergado system presented high values for specificity (99.9%) and sensitivity (99.6%) for cow detection. The visit duration and feed intake per visit collected using the electronic monitoring system were similar to the video and manual weighing data, respectively. The difference between the feed intake measured manually and the sum of the electronically recorded feed intake was less than 250 g (25,635 ± 2,428 and 25,391 ± 2,428 g estimated using manual weighing and the electronic system, respectively). In conclusion, the Intergado system is a reasonable tool to monitor feeding behavior and feed intake for freestall-housed dairy cows.
Copyright © 2015 American Dairy Science Association. Published by Elsevier Inc. All rights reserved.
Scale is the most accurate means of measuring the actual body weight (BW) in developing heifers. However, when it is not available, indirect tools to estimate BW are needed. The objective of this study was to evaluate the distance between the left and right coxal tuberosity (DLRCT) as potential estimator of shrunk BW (SBW) of developing Holstein heifers as compared with the actual weight measured by an electronic scale. The study included pre-breeding Argentinean Holstein females from 3 to 21 months of age. A dataset comprising 496 observations was used to quantify the relationship between the DLRCT (cm) and the SBW (kg) by the least squared method: The power function Y = 0.131X(2.0758), where X = DLRCT provided the best tit (P < 0.0001) for predicting SBW. A dataset comprising 194 observations was used to assess the strength of agreement of the power function. The Lin correlation coefficient value was 0.97 and the computed 95% CI was 0.965-0.979. The mean difference between observed and estimated SBW was -0.99 kg. There was no significant difference (t = -0.83; P = 0.41) in the mean SBW between observed and estimated data. As a predictor of SBW, particularly in heifers >= 5 months of age and 5 350 kg BW (i.e. <= 45 cm), the DLRCT demonstrated to be a useful alternative that can be easily applied during any practice usually performed on replacement heifers, without requiring a squeeze chute.
The objective of the current study was to determine the accuracy of the prediction of live weight (LW) from body measurements (BMs) by using digital image analysis on female Holstein calves. The calves were measured with a measurement stick and digital image analysis. The following linear parameters were taken: body length (BL), wither height (WH), chest depth (CD), hip height (HH) and hip width (HW). LW and BMs were recorded at birth, at weaning and at 24 weeks of age. Regression coefficients, which included all BMs at birth, gave a low R
2 value (66·7%), but the R
2 value was found to be 87·6 and 86·0% at weaning and 24 weeks of age, respectively. A high correlation coefficient was found among LW and CD, HH and HW at weaning (0·90, 0·91 and 090, respectively) and at 24 weeks of age (0·89, 0·90 and 0·91, respectively). The results confirm that for female Holstein calves, digital image analysis is an effective measuring system for the prediction of LW from BMs.
The objective of this paper is to provide recognition for Pantaneira cattle breed using Convolutional Neural Networks (CNN). Fifty-one animals from the Aquidauana Pantaneira cattle Center (NUBOPAN) were studied. The center is located in the Midwest region of Brazil. Four monitoring cameras were distributed in the fences and took 27,849 images of Pantaneira cattle breed using different angles and positions. The following three CNN architectures were used for the experiment: DenseNet-201, Resnet50 and Inception-Resnet-V. All networks were submitted to 10-fold stratified cross-validation over 50 epochs. The results showed an accuracy of 99% in all networks, which is encouraging for future research.
This study aimed to determine body weight (BW), body condition score (BCS), and dairy type traits (DTT) of Holstein heifers and lactating cows using three-dimensional (3D) cameras, in addition to evaluating the sensitivity of this system over time and the best sensor position. Twenty-eight cows and twenty-seven heifers were used. Measurements of BW and BCS were taken over five months, and one single measurement was taken for each of the 23 lactating cows that had the official Holstein register to evaluate DTT. Images were taken using a Microsoft Kinect 3D camera from lateral and dorsal perspectives to predict BW and BCS, synced with MATLAB software. Fourteen and thirteen measurements were taken from dorsal and lateral perspectives, respectively. Then, the SAS GLMSELECT LASSO procedure was used to test the linear and quadratic effects, and the ratios of the obtained variables. Subsequently, selected characteristics were tested using PROC MIXED of SAS to fit the models and predict BW and BCS. In addition, DTT were evaluated using 3D camera images to estimate the Holstein Association official grade. The udder, chest, and close back side were used to complement the lateral and dorsal images. Biometric measurements and 3D camera data were also compared to each other using a paired t-test. The obtained models to predict BW had an R² of 0.89 and 0.96 and RMSE of 49.20 and 26.89 for lateral and dorsal perspectives, respectively. The lateral model was composed of body weight, height, body depth, and body lateral volume. The dorsal model was composed of rump width, thorax width, and dorsal area. The model obtained to predict BCS had an R² of 0.63 and 0.61 and RMSE of 0.16 and 0.17 for lateral and dorsal images respectively. The lateral model was composed of body depth, lateral area, and body weight divided by height. The dorsal model was composed of dorsal length, dorsal area, dorsal volume, and body weight to dorsal area ratio. Among all official 15 evaluated DTT, 4 were adequately predicted (P < 0.05) by 3D cameras. Ten DTT were adequately predicted (P < 0.05) by biometric measurements. In conclusion, 3D cameras have a good prospective future commercial use and either lateral or dorsal images could be used for BW prediction however the BCS models still need improvements. The udder traits were those DTT with the best prospective use, due to the highest accuracy.
This study entailed the design and implementation of a computer vision system for cow individual feed intake measurement, based on deep Convolutional Neural Networks (CNNs) models, and a low-cost RGB-D (Red, Green, Blue, Depth) camera. Individual feed intake of dairy cows is an important variable currently unavailable in commercial dairies. An RGB-D camera was positioned above the feeding area in an open cowshed. Feed intake was estimated by combining information from the RGB and depth images. Cow identification was conducted using the RGB image. Deep learning algorithms for identification and intake estimation were developed using CNN models. Data for CNN training were acquired by a specially developed automatic data acquisition system. A range of feed weights under varied configurations were collected over a period of seven days with the setup, which included an automatic scale, cameras, and a micro-controller. Test data for feed intake was acquired in an open cowshed research dairy farm, wherein the cows were fed Total Mix Ration (TMR). Images of cows eating over a period of 36 h provided the test data for cow identification. The system was able to accurately identify 93.65% of the cows. The amount of feed consumed, which ranged from 0 to 8 kg per meal, was measured with mean absolute and square errors (MAE and MSE) of 0.127 kg, and 0.034 respectively. The analysis showed that the amount and diversity of data are important for model training. Better results were achieved for the model that was trained with high-diversity data than the model trained with homogeneous data (MAE of 1.025 kg, and MSE of 2.845 kg^2 for a model trained on shadow conditions only). Additionally, the training analysis shows that the model based on RGB-D data shows better results than the model based on depth channel data without RGB (MAE of 0.241 kg, and MSE of 0.106 kg^2). These results suggest the potential of low-cost cameras for individual feed intake measurements in advanced dairy farms.
Frequent measurements of body weight (BW) in livestock systems are very important because they allow assessing growth. However, real-time monitoring of animal growth through traditional weighing scales is stressful for animals, costly and labor-intensive. Thus, the objectives of this study were to: 1) assess the predictive quality of an automated computer vision system used to predict BW and average daily gain (ADG) in beef cattle; and 2) compare different predictive approaches, including Multiple Linear Regression (MLR), Least Absolute Shrinkage and Selection Operator (LASSO), Partial Least Squares (PLS), and Artificial Neutral Networks (ANN). A total of 234 images of Nellore beef cattle were collected during the weaning, stocker and feedlot phases. First, biometric body measurements of each animal, such as body volume, area, length, and others, were performed using three-dimensional images captured with the Kinect® sensor, and their respective BW were acquired using an electronic scale. Next, the biometric measurements were used as explanatory variables in the four predictive approaches (MLR, LASSO, PLS, and ANN). To evaluate prediction quality, a leave-one-out cross-validation was adopted. The ANN was the best prediction approach in terms of Root Mean Square Error of Prediction (RMSEP) and squared predictive correlation (r2). The results for Weaning were RMSEP = 8.6 kg and r2 = 0.91; for Stocker phase, RMSEP = 11.4 kg and r2 = 0.79; and for Beginning of feedlot, RMSEP = 7.7 kg and r2 = 0.92. The ANN was also the best method for prediction of ADG, with RMSEP = 0.02 kg/d and r2 = 0.67 for the period between Weaning and Stocker, RMSEP = 0.02 kg/d and r2 = 0.85 for the Weaning and Beginning of Feedlot phase, RMSEP = 0.03 kg/d and r2 = 0.80 for Weaning and Final of Feedlot phase, RMSEP = 0.10 kg/d and r2 = 0.51 for Stocker and Beginning of feedlot phase, and RMSEP = 0.09 kg/d and r2 = 0.82 for the Beginning and Final of feedlot phase. Overall, the results indicate that the proposed automated computer vision system can be successfully used to predict BW and ADG in real-time in beef cattle.