Figure - available from: Animals
This content is subject to copyright.
Decomposition movement of sheep’s typical walking behavior in one period. (i) The starting state of walking behavior; (ii) the sheep lifts the left leg; (iii) the body moves forward with the left leg; and (iv) the sheep lifts the right leg and moves forward with the right leg to the starting state.
Source publication
Simple Summary
The monitoring and analysis of sheep behavior can reflect their welfare and health, which is beneficial for grazing management. For automatic classification and the continuous monitoring of grazing sheep behavior, wearable devices based on inertial measurement unit (IMU) sensors are important. The accuracy of different machine learni...
Similar publications
The aging population has led to a growing number of falls in our society, affecting global public health worldwide. This paper presents CareFall, an automatic Fall Detection System (FDS) based on wearable devices and Artificial Intelligence (AI) methods. CareFall considers the accelerometer and gyroscope time signals extracted from a smartwatch. Tw...
Citations
... Untuk mendukung pengambilan keputusan dalam manajemen hewan, sangat penting untuk fokus pada pengumpulan data secara real-time dari peternakan [33]. Identifikasi domba individu di peternakan dan perilaku mereka yang berbeda sangat penting untuk memantau kesehatan dan praktik peternakan lainnya [34]. Peneliti sebelumnya telah mengembangkan sistem pelacakan otomatis berbasis radar gelombang milimeter yang menawarkan pemantauan hewan secara akurat dan real-time [35]. ...
Sektor peternakan merupakan sektor yang berkembang pesat di Indonesia hal ini sejalan dengan program pemerintah terkait pemenuhan gizi protein asal hewan untuk masyarakat Indonesia. Pemanfaatan Teknologi Kecerdasan Buatan atau Artificial Intelligence (AI) dalam sektor peternakan di Indonesia menawarkan solusi inovatif untuk mengatasi berbagai masalah terkait program di Indonesia dalam meningkatkan pemenuhan kebutuhan protein asal hewan. Metode penelitian ini meliputi kajian literatur tentang aplikasi AI terkini dalam peternakan, serta studi kasus dari implementasi di Indonesia. Artificial Inteligence dapat memainkan peran kunci dalam mengoptimalkan produksi dan distribusi produk peternakan. Penelitian ini mengeksplorasi penerapan AI dalam meningkatkan efisiensi operasional peternakan melalui manajemen pakan cerdas, pemantauan kesehatan hewan secara real-time, dan otomatisasi proses produksi. Artificial Inteligence memungkinkan penggunaan sensor dan analisis data untuk memberikan rekomendasi terkait pakan, mendeteksi dini penyakit, dan memantau kesejahteraan hewan, yang dapat mengurangi kerugian dan meningkatkan hasil produksi. Penggunaan AI dalam sektor peternakan menghadapi berbagai tantangan seperti biaya implementasi AI yang tinggi, kebutuhan infrastruktur teknologi yang memadai, dan keterbatasan tenaga kerja terampil. Artificial Intelligence dapat secara signifikan meningkatkan efisiensi dan produktivitas peternakan, serta mendukung pemenuhan kebutuhan protein hewani masyarakat. Kesimpulannya, meskipun terdapat tantangan yang harus diatasi, integrasi AI dalam sektor peternakan diharapkan dapat memperkuat ketahanan pangan dan meningkatkan keberlanjutan industri peternakan di Indonesia.
... Accurate identification of feeding behavior in sheep is essential for health monitoring and enables the optimization of feed formulation and feeding strategies. The feeding behavior of sheep is divided into two categories: foraging and ruminating [2]. Foraging is the behavior of sheep to actively ingest food, which directly affects their growth, development, and health status. ...
In precision feeding, non-contact and pressure-free monitoring of sheep feeding behavior is crucial for health monitoring and optimizing production management. The experimental conditions and real-world environments differ when using acoustic sensors to identify sheep feeding behaviors, leading to discrepancies and consequently posing challenges for achieving high-accuracy classification in complex production environments. This study enhances the classification performance by integrating the deep spectrogram features and acoustic characteristics associated with feeding behavior. We conducted the task of collecting sound data in actual production environments, considering noise and complex surroundings. The method included evaluating and filtering the optimal acoustic features, utilizing a customized convolutional neural network (SheepVGG-Lite) to extract Short-Time Fourier Transform (STFT) spectrograms and Constant Q Transform (CQT) spectrograms’ deep features, employing cross-spectrogram feature fusion and assessing classification performance through a support vector machine (SVM). Results indicate that the fusion of cross-spectral features significantly improved classification performance, achieving a classification accuracy of 96.47%. These findings highlight the value of integrating acoustic features with spectrogram deep features for accurately recognizing sheep feeding behavior.
... AdaBoost (Adaptive Boosting) is an ensemble learning algorithm introduced by Yoav Freund and Robert Schapire in 1996 [29]. As a boosting method [30], AdaBoost aims to enhance overall classification performance by combining multiple weak classifiers (weak learners) into a strong classifier [31]. ...
To analyze the impact of the marine environment on the relative abundance of Illex argentinus (high and low categories) in the southwest Atlantic, this study collected logbook data from Chinese pelagic trawlers from December 2014 to June 2024, including vessel position data and oceanographic variables such as sea surface temperature, 50 m and 100 m water temperature, sea surface salinity, sea surface height, chlorophyll-a concentration, and mixed layer depth. Vessel positions were used to enhance the logbook data quality, allowing an analysis of the annual trends in the resource center of this squid at a spatial resolution of 0.1° × 0.1° and a temporal resolution of ten days. The findings showed that the resource center is primarily located around 42° S in the north and between 45° S and 47° S in the south, with a trend of northward movement during the study period. Additionally, we constructed two ensemble learning models based on decision trees—AdaBoost and PSO-RF—aiming to identify the most critical environmental factors that affect its resource abundance; we found that the optimal model was the PSO-RF model with max_depth of 5 and n_estimators of 46. The importance analysis revealed that sea surface temperature, mixed layer depth, sea surface height, sea surface salinity, and 50 m water temperature are critical environmental factors affecting this species’ resources. Given that deep learning models generally have shorter running times and higher accuracy than other models, we developed a CNN-Attention model based on the five most important input factors. This model achieved an accuracy of 73.6% in forecasting this squid for 2024, predicting that the population would first appear near the Argentine exclusive economic zone around mid-December 2023 and gradually move east and south thereafter. The predictions of the model, validated through log data, maintained over 70% accuracy during most periods at a time scale of ten days. The successful construction of the resource abundance forecasting model and its accuracy improvements can help enterprises save fuel and time costs associated with blind searches for target species. Moreover, this research contributes to improving resource utilization efficiency and reducing fishing duration, thereby aiding in lowering carbon emissions from pelagic trawling activities, offering valuable insights for the sustainable development of this species’ resources.
... McLennan et al. [6] validated an automated recording system for evaluating behavioral activity levels in sheep. Current research on gait recognition algorithms primarily focuses on analyzing behavioral patterns of sheep [7][8][9] and cattle [10][11][12]. This study has drawn on their research method, utilizing gait analysis to measure activity levels. ...
Featured Application
This research introduces a novel wearable device that uses an acceleration threshold behavior recognition method to classify horse activities into three levels: low (standing), medium (walking), and high (trotting, cantering, and galloping). The recognition algorithm is directly implemented in the hardware, which horses wear during their training sessions. This device allows for the real-time analysis of horse activity levels and the accurate calculation of the time spent in each activity state. This method provides scientific data support for horse training, facilitating the optimization of training programs.
Abstract
This study demonstrated that wearable devices can distinguish between different levels of horse activity, categorized into three types based on the horse’s gaits: low activity (standing), medium activity (walking), and high activity (trotting, cantering, and galloping). Current research in activity level classification predominantly relies on deep learning techniques, known for their effectiveness but also their demand for substantial data and computational resources. This study introduces a combined acceleration threshold behavior recognition method tailored for wearable hardware devices, enabling these devices to classify the activity levels of horses directly. The approach comprises three sequential phases: first, a combined acceleration interval counting method utilizing a non-linear segmentation strategy for preliminary classification; second, a statistical analysis of the variance among these segments, coupled with multi-level threshold processing; third, a method using variance-based proximity classification for recognition. The experimental results show that the initial stage achieved an accuracy of 87.55% using interval counting, the second stage reached 90.87% with variance analysis, and the third stage achieved 91.27% through variance-based proximity classification. When all three stages are combined, the classification accuracy improves to 92.74%. Extensive testing with the Xinjiang Wild Horse Group validated the feasibility of the proposed solution and demonstrated its practical applicability in real-world scenarios.
... The contact sensors are mainly wearable accelerometers. Jin et al. evaluated the effect of multiple combinations of algorithms, time windows, and accelerometer data on the performance of sheep behavior classification [20]. Hu et al. input accelerometer data into various machine learning models to identify sheep grazing behavior and improve the accuracy of recognition by using features derived from mixed-size time windows [21]. ...
... , (20) where is the mean of each column, and is the standard deviation of each column of temporal features. Second, the covariance matrix C can be computed as follows: , ...
The rapid development of agricultural cyber-physical systems sheds new light on facilitating agricultural production. The grazing behavior recognition of herded sheep is a paramount issue in animal husbandry. Triaxial accelerometers of agricultural cyber-physical systems provide fine-grained observations of herded sheep but also generate temporal-spatial correlated acceleration data with inherently large-scale dimensions and massive volumes. These inherent characteristics of the data constrain the direct application of existing recognition algorithms. Motivated by the unique features of triaxial accelerometers of agricultural cyber-physical systems, we design a hybrid temporal-spatial fuzzy deep neural network (TSFDNN) approach for predicting the grazing behaviors of herded sheep. We first extract temporal-spatial features and reduce data dimensionality using bidirectional long short-term memory network (Bi-LSTM) and convolutional neural network (CNN) in parallel, then control feature dimensions through principal component analysis (PCA), and finally use fuzzy neural network (FNN) to achieve feature enhancement and category mapping. The superiority of the designed TSFDNN is demonstrated through its empirical comparison with other state-of-the-art machine learning algorithms by using two datasets from sheep pastures. Furthermore, we analyze the rationale of each component in the designed TSFDNN by performing several ablation studies. We also conduct robustness experiments with heterogeneous dimension reduction and optimization algorithms to explore the generalization capabilities of TSFDNN. The managerial implications of precisely identifying herded sheep behaviors for production decision-making, agricultural management, animal welfare, and ecological protection are discussed.
... To bolster decision-making in animal management, it is essential to focus on real-time data collection from the farm (Lewis Baida et al., 2021). The identification of individual sheep on the farm and their various behaviours is crucial for monitoring their health and other farming practices (Jin et al., 2022b(Jin et al., , 2022a. Dore et al. (2021) have developed an automated tracking system based on millimeter-wave radars which offers precise and real-time animal monitoring. ...
... Referred to simply as "drones", UAVs offer large-coverage and can acquire data in real-time. Jin et al. (2022bJin et al. ( , 2022a) developed an AI model utilizing videos captured by drones. The model was trained to detect potential anomalies. ...
... With limited resources, it is not suitable for running deep learning or machine learning model algorithms.). Furthermore, current research on behavior recognition algorithms primarily focuses on analyzing behavioral patterns of sheep [9][10][11] and cattle [12][13][14]. Therefore, in developing a new horse behavior classification algorithm, this study referenced research methodologies from other domains. ...
Analyzing horse behavior is crucial for assessing training quality, particularly in accurately identifying actions like standing, walking, and running. Research has been conducted both domestically and internationally in this regard; however, challenges persist, including reliance on single sensors, poor real-time performance, and low identification accuracy. In view of these challenges, this study investigated real-time identification of horse behavior based on wearable devices. The system, centered around a microcontroller and utilizing a 4G network as a carrier, employs multi-axis IMU sensors as input sources to perceive horse posture. The study proposed a behavior classification method based on analysis of acceleration thresholds. The method consists primarily of two sequential stages: first, the resultant acceleration interval counting method, which employs a nonlinear segmentation approach for initial behavior classification; and second, the statistical analysis of variance parameters between segments, which when coupled with multi-level threshold processing, achieves a refined classification. Experimental results indicated that the interval counting method alone achieved an accuracy of 87.55%, while for the variance analysis method alone the accuracy was 90.87%. The proposed method, comprising the two stages, reached a classification accuracy of 91.57%, underpinning its usefulness in supporting future equine research.
... These accelerometer data are usually collected at high frequencies (typically tens of hertz). Jin et al. (2022) used a sampling frequency of 20 Hz to classify walking, standing, grazing, lying down, and running behaviors in sheep. Bloch et al. (2023), for the feeding behavior classification of dairy cows, used a sampling frequency of 25 Hz, while Alvarenga et al. (2020) used the same frequency to discriminate biting and chewing behaviors in sheep. ...
Avaluating ingestive behavior is crucial for making decisions about animal management; however, direct observation can be laborious, time-consuming, and exhaustive for the observer. In this context, the development of measurement devices coupled with computational methods to reliably discriminate animal behavior and identify an appropriate sampling frequency for battery autonomy without compromising precision is important. This study aimed to develop an electronic device for discriminating ingestive behavior in sheep using a triaxial accelerometer and machine learning, considering different sampling frequencies and classification models. The analyzed sampling frequencies were 1, 0.2, 0.1, 0.067, and 0.05 Hz, validated in 18 sheep. The electronic device successfully distinguished feeding, rumination, and other behaviors. The highest values in the results for ingestive behavior classification were achieved by the Random Forest model with and without variable selection, with accuracies ranging from 85.2 to 88.6 %, Kappa indices from 0.78 to 0.83, and false-positive rates from 5.4 to 7.2. Furthermore, the best-performing model differentiated feeding, rumination, and other behaviors with sensitivities ranging from 85.70 to 88.33 % and specificities from 92.71 to 94.11 %. Sampling frequencies of 1, 0.2, 0.1, and 0.05 Hz showed statistically similar means (p < 0.05) for the average error between observed and device-recorded data, ranging from 1.50 to 2.82 min/hour for feeding behavior, 4.61 to 6.10 min/hour for rumination behavior, and 2.43 to 3.60 min/hour for other behaviors. Therefore, the developed electronic device successfully discriminated ingestive behavior in sheep using a triaxial accelerometer and machine learning. Additionally, the 0.05 Hz sampling frequency yielded good results, indicating its potential to reduce device energy consumption without compromising precision in assessing feeding, rumination, and other behaviors.
... In addition, wearable devices based on the Inertial Measurement Unit (IMU) sensor device are becoming more prevalent. It can be used, for example, to support the automatic behavior classification of grazing sheep and the evaluation of production performance [14,15]. There are also promising applications for the interaction between different animals, search and rescue missions, and the protection of animals from poaching and theft [7]. ...
... This review includes articles with Qualsyst score percentages ≥60%. As a result, 23 articles were included for in-depth analysis [7,[13][14][15][16][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42][43][44]; the remaining 5 articles were excluded since their score was below the acceptance percentage [45][46][47][48][49]. The flow diagram of the review phases' is shown in Figure 2. The first study identified dates back to 2016 [39], and 95.7% (22/23) of the studies were conducted in the last five years (2018-2022). ...
... However, sensor fusion can be applied to other animal populations. In the high level: pigs [33], felines [13,32], koalas [31], goats [29], and birds [32], regarding feature level: pigs [35], fish [43], dogs [27], camouflaged animals [34], and sheep [14]. Meanwhile, for the low-level: horses [39], dogs [7], primates [16], and sheep [15]. ...
The development of technology, such as the Internet of Things and artificial intelligence, has significantly advanced many fields of study. Animal research is no exception, as these technologies have enabled data collection through various sensing devices. Advanced computer systems equipped with artificial intelligence capabilities can process these data, allowing researchers to identify significant behaviors related to the detection of illnesses, discerning the emotional state of the animals, and even recognizing individual animal identities. This review includes articles in the English language published between 2011 and 2022. A total of 263 articles were retrieved, and after applying inclusion criteria, only 23 were deemed eligible for analysis. Sensor fusion algorithms were categorized into three levels: Raw or low (26%), Feature or medium (39%), and Decision or high (34%). Most articles focused on posture and activity detection, and the target species were primarily cows (32%) and horses (12%) in the three levels of fusion. The accelerometer was present at all levels. The findings indicate that the study of sensor fusion applied to animals is still in its early stages and has yet to be fully explored.
... The MPU6050 sensor can collect three-axis acceleration and three-axis angular velocity. The acceleration directions are illustrated in Figure 2, where the x, y, and z axes represent lateral, vertical, and horizontal motion, respectively [13]. The three-axis acceleration values ranged from ±8 g (m/s 2 ), while the triaxial angular velocity values ranged from ±500 dps ( • /s). ...
... False negative (FN) indicated that the model predicted the category as negative, but the actual category was positive. The precision, as well as the recall, were calculated by Equations (6) and (7) [13]. ...
Step counting is an effective method to assess the activity level of grazing sheep. However, existing step-counting algorithms have limited adaptability to sheep walking patterns and fail to eliminate false step counts caused by abnormal behaviors. Therefore, this study proposed a step-counting algorithm based on behavior classification designed explicitly for grazing sheep. The algorithm utilized regional peak detection and peak-to-valley difference detection to identify running and leg-shaking behaviors in sheep. It distinguished leg shaking from brisk walking behaviors through variance feature analysis. Based on the recognition results, different step-counting strategies were employed. When running behavior was detected, the algorithm divided the sampling window by the baseline step frequency and multiplied it by a scaling factor to accurately calculate the number of steps for running. No step counting was performed for leg-shaking behavior. For other behaviors, such as slow and brisk walking, a window peak detection algorithm was used for step counting. Experimental results demonstrate a significant improvement in the accuracy of the proposed algorithm compared to the peak detection-based method. In addition, the experimental results demonstrated that the average calculation error of the proposed algorithm in this study was 6.244%, while the average error of the peak detection-based step-counting algorithm was 17.556%. This indicates a significant improvement in the accuracy of the proposed algorithm compared to the peak detection method.
