Fig 2 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Source publication
The application of hyperspectral imaging (HSI) has gained significant importance in the past decade, particulary in the context of food analysis, including potatoes. However, the current literature lacks a comprehensive systematic review of the application of this technique in potato cultivation. Therefore, the aim of this work was to conduct a sys...
Context in source publication
Context 1
... abstracts, 63 articles were selected for full-text and eligibility. Finally, 52 articles met the inclusion criteria and were incorporated into the present review. During full-text assessment for eligibility, the percentage of agreement between researchers (CMPA, ALM, SA) was 97.5%. After a meeting, a 100% agreement was reached. The flowchart in Fig. 2 summarises the inclusion selection ...
Citations
... For instance, Wang et al. [33] evaluated the advancements, limitations, and challenges of HSI to assess wheat quality. Furthermore, Peraza-Alemán et al. [34] reviewed its application in potatoes, Li et al. [35] in different fruit and vegetable products, Ismail et al. [36] in seafood products and Matenda et al. [37] in meat. ...
Hyperspectral imaging is now establishing itself as a transformative analytical technique in the food safety and quality domains, offering unique capabilities for non-destructive, real-time, and high-resolution analysis of food at different levels of its production. Hyperspectral imaging combines the strengths of computer vision and classical spectroscopy. It provides both spatial and spectral information, making it a powerful and green alternative to the conventional techniques employed in this field.
This critical review explores the advances in hyperspectral imaging applications, highlighting its potential to revolutionize food quality and safety assessment, including adulteration, contamination and non-conformity detection. Recent breakthroughs in sensor technology, data processing algorithms, and machine learning integration are discussed, emphasizing the most popular data analysis strategies and their role in addressing the challenges of complex food matrices and dynamic production environments. This review underlines the data analysis approaches applied in each of the collected works, highlighting two trends: studying food samples as a whole or analyzing them as a set of pixel-spectra. Machine learning methods such as principal component analysis, partial least squares regression, partial least squares discriminant analysis, soft independent modelling of class analogy, and support vector machines have been widely applied for the analysis of food samples. These techniques are used for both qualitative and quantitative purposes, regardless of the sample’s origin (plant- or animal-based) or its complexity. Additionally, this review outlines the limitations of hyperspectral imaging, such as high costs, computational demands, and the need for standardized protocols, while identifying opportunities for future research and industrial implementation.
... Hyperspectral imaging has been used to analyze the most relevant compounds, diseases, and stress factors in potatoes [18]. This study used hyperspectral imaging technology combined with machine learning models to detect five different external defects of potatoes: scab, black skin, broken skin, green skin, and mechanical damage. ...
For potato external defect detection, ordinary spectral technology has limitations in detail detection and processing accuracy, while the machine vision method has the limitation of a long feedback time. To realize accurate and rapid external defect detection for red-skin potatoes, a non-destructive detection method using hyperspectral imaging and a machine learning model was explored in this study. Firstly, Savitzky–Golay (SG), standard normal variate transformation (SNV), multiplicative scatter correction (MSC), the normalization algorithm, and different preprocessing algorithms combined with SG were used to preprocess the hyperspectral data. Then, principal component regression (PCR), support vector machine (SVM), partial least squares regression (PLSR), and least squares support vector machine (LSSVM) algorithms were used to establish quantitative models to find the most suitable preprocessing algorithm. The successive projections algorithm (SPA) was used to obtain various characteristic wavelengths. Finally, the qualitative models were established to detect the external defects of potatoes using the machine learning algorithms of backpropagation neural network (BPNN), k-nearest neighbors (KNN), classification and regression tree (CART), and linear discriminant analysis (LDA). The experimental results showed that the SG–SNV fusion hyperspectral data preprocessing algorithm and the KNN machine learning model were the most suitable for the detection of external defects in red-skin potatoes. Moreover, multiple external defects can be detected without multiple models. For healthy potatoes, black/green-skin potatoes, and scab/mechanical-damage/broken-skin potatoes, the detection accuracy was 93%,93%, and 83%, which basically meets the production requirements. However, enhancing the prediction accuracy of the scab/mechanical-damage/broken-skin potatoes is still a challenge. The results also demonstrated the feasibility of using hyperspectral imaging technology and machine learning technology to detect potato external defects and provided new insights for potato external defect detection.
... The raw hyperspectral images require a series of operations to extract meaningful information, enhance interpretability, and prepare them for further analysis, such as feature extraction, development of classification, prediction, and regression model development (Peraza-Alemán et al., 2024). The basic steps of hyperspectral data analysis and visualization are shown in Fig. 3. ...
Conventional egg quality analysis and compliance monitoring methods have inherent limitations, necessitating non-destructive techniques in the modern egg industry. Hyperspectral imaging (HSI) has emerged as a fast, accurate, and non-destructive tool for quality assessment, effectively determining egg's internal and external properties. Following the fundamentals of HSI and image analysis, this review consolidates and discusses recent applications of HSI technology for table and hatching egg analysis, addressing its limitations and potential challenges. Current research demonstrates HSI's efficacy in rapidly and accurately determining parameters such as freshness, shell integrity, defects, chemical composition, and detection of fake eggs. Despite its promising performance, the widespread industrial application of HSI would encounter multiple challenges due to the inherent properties of the samples (e.g., complex shape, light diffusivity, light and heat sensitivity) and current technological limitations. The existing research is insufficient for early predicting certain critical parameters such as fertility, egg sex, and embryonic mortality in high-throughput screening. However, current research underscores HSI's enormous potential, highlighting that advanced machine learning combined with HSI technology can revolutionize conventional egg and hatchery operations, enhancing automation, economic sustainability, and global animal welfare. This review can guide researchers and policymakers in understanding the contemporary challenges of HSI technology, developing innovative solutions, improving regulatory frameworks, and fostering advancements to maximize the benefits of this cutting-edge green technology.
... The raw hyperspectral images require a series of operations to extract meaningful information, enhance interpretability, and prepare them for further analysis, such as feature extraction, development of classification, prediction, and regression model development (Peraza-Alemán et al., 2024). The basic steps of hyperspectral data analysis and visualization are shown in Fig. 3. ...
Conventional egg quality analysis and compliance monitoring methods have inherent limitations, necessitating non-destructive techniques in the modern egg industry. Hyperspectral imaging (HSI) has emerged as a fast, accurate, and non-destructive tool for quality assessment, effectively determining egg's internal and external properties. Following the fundamentals of HSI and image analysis, this review consolidates and discusses recent applications of HSI technology for table and hatching egg analysis, addressing its limitations and potential challenges. Current research demonstrates HSI's efficacy in rapidly and accurately determining parameters such as freshness, shell integrity, defects, chemical composition, and detection of fake eggs. Despite its promising performance, the widespread industrial application of HSI would encounter multiple challenges due to the inherent properties of the samples (e.g., complex shape, light diffusivity, light and heat sensitivity) and current technological limitations. The existing research is insufficient for early predicting certain critical parameters such as fertility, egg sex, and embryonic mortality in high-throughput screening. However, current research underscores HSI's enormous potential, highlighting that advanced machine learning combined with HSI technology can revolutionize conventional egg and hatchery operations, enhancing automation, economic sustainability, and global animal welfare. This review can guide researchers and policymakers in understanding the contemporary challenges of HSI technology, developing innovative solutions, improving regulatory frameworks, and fostering advancements to maximize the benefits of this cutting-edge green technology.
... These findings indicate that hyperspectral images are effective in illustrating the reflectance change of potatoes before and after spouting. Hyperspectral imaging [40,41] and 3D imaging [42,43] have each been employed independently to assess potato quality. In this work, we present a case to support that the integration of hyperspectral imaging with 3D imaging provides a more effective approach for detecting potato sprouts. ...
This paper proposes a four-dimensional (4D) line-scan hyperspectral imaging system to acquire 3D spatial data and hyperspectral images covering from visible to short-wave infrared (Vis-SWIR) wavelength range. The system captures visible and near-infrared (VNIR) and SWIR hyperspectral images using two line-scan sensors, while 3D spatial data is acquired via a fringe projection profilometry subsystem. To align the VNIR and SWIR hyperspectral images, we utilize a line-scan homography method and propose a transformation method to register 3D spatial data with hyperspectral images. The mean reprojection error for hyperspectral image registration is 0.5396 pixels, and the registration of 3D spatial data with hyperspectral images achieves subpixel accuracy. Spatial accuracy is demonstrated with a root mean squared error (RMSE) of 0.1321 mm and a mean absolute error (MAE) of 0.1066 mm by measuring a standard sphere with a 20.0512 mm radius. The spectral resolutions are 11.2 nm in the VNIR range and 5 nm in the SWIR range. Two case studies were conducted: one involving a colorful object with rich features and colors, and another involving a potato before and after sprouting. Results from the measurement of a colorful object demonstrate the proposed system’s registration accuracy and image intensity variation across wavelengths, while the potato study highlights the system’s potential applications in the food industry.
... Moreover, the effective preprocessing and deliberate choice of wavelengths in hyperspectral imaging are recommended to reduce extraneous factors that can affect the quality of measured spectra, such as the morphological variations and light-scattering characteristics of different food items [33,35,36]. So, the integration of hyperspectral imaging with AI, including optimized preprocessing and wavelength selection in view of food quality monitoring, remains so far poorly addressed, and relevant studies have been identified only in [1,10,15,18,37,38]. ...
Hyperspectral imaging (HSI) is one of the non-destructive quality assessment methods providing both spatial and spectral information. HSI in food quality and safety can detect the presence of contaminants, adulterants, and quality attributes, such as moisture, ripeness, and microbial spoilage, in a non-destructive manner by analyzing spectral signatures of food components in a wide range of wavelengths with speed and accuracy. However, analyzing HSI data can be quite complicated and time consuming, in addition to needing some special expertise. Artificial intelligence (AI) has shown immense promise in HSI for the assessment of food quality because it is so powerful at coping with irrelevant information, extracting key features, and building calibration models. This review has shown various machine learning (ML) approaches applied to HSI for quality and safety control of foods. It covers the basic concepts of HSI, advanced preprocessing methods, and strategies for wavelength selection and machine learning methods. The application of HSI to AI increases the speed with which food safety and quality can be inspected. This happens through automation in contaminant detection, classification, and prediction of food quality attributes. So, it can enable decisions in real-time by reducing human error at food inspection. This paper outlines their benefits, challenges, and potential improvements while again assessing the validity and practical usability of HSI technologies in developing reliable calibration models for food quality and safety monitoring. The review concludes that HSI integrated with state-of-the-art AI techniques has good potential to significantly improve the assessment of food quality and safety, and that various ML algorithms have their strengths, and contexts in which they are best applied.
... CAFAD depends on the weather conditions and crop phenological stage (Strydom et al. 2024). The information on the CAFAD is essential to decide the mode of control of the disease in addition to the type and amount of pesticide to be applied; it will help the farmers to take appropriate measures to protect the crop and sustain the potatoes to taste better (Peraza-Alemán et al. 2024). Simulating the disease occurrence is an environment-friendly and low-cost method of crop characterisation. ...
Weather-based simulation models were developed in the present study to characterise late blight of potato in the Lower Gangetic Plains using AI-based machine learning and multiple linear regression approaches. Weather variables considered in this study were, daily maximum temperature, minimum temperature, rainfall, morning relative humidity, evening relative humidity, wind speed, and solar radiation, observed during the potato crop growing period, along with the disease parameter — crop age at first appearance of disease (CAFAD) — from 2006 to 2020 (15 years). A total of 56 simple and weighted weather-based indices were developed which served as inputs for the model development and validation. Simulation models were developed using machine learning approaches like LASSO, SVM, and multiple linear regression technique SMLR, along with two hybrid models like LASSO-SVM and SMLR-SVM under early, normal, late, and all (pooled data) planting conditions in lower Gangetic plains. Out of all the AI-based machine learning and multiple linear regression models developed, based on the overall standardised Ranking Performance Index (sRPI), the SMLR model was considered to be the best to simulate crop age at first appearance of the disease under early and normal planting conditions, whereas SVM and LASSO were the best under late and all planting conditions, respectively. The developed models can be used in decision support systems to predict potato late blight and can also be extended to develop similar models for other locations.
Despite the high spectral resolution and abundant information of hyperspectral images (HSI), their spatial resolution is relatively low due to limitations in sensor technology. Sensors often need to sacrifice some spatial resolution to ensure accurate light energy measurement when pursuing high spectral resolution. This trade-off results in HSI’s inability to capture fine spatial details, thereby limiting its application in scenarios requiring high-precision spatial information. HSI and multispectral images (MSI) fusion is a commonly used technique for generating high-resolution HSI (HR-HSI). However, many deep learning-based HSI-MSI fusion algorithms ignore correlation and multi-scale information between input images. To address this issue, we propose an asymptotic multi-scale symmetric fusion network (AMSF-Net) for hyperspectral and multispectral image fusion. AMSF-Net consists of two parts: the multi-level feature fusion (MFF) module and the progressive cross-scale spatial perception (PCP) module. The MFF module uses multi-stream feature extraction branches to perform information interaction between HSI and MSI at the same scale layer by layer, compensating for the spatial details lacking in HSI and the spectral details absent in MSI. The PCP module combines the input and output features of MFF, utilizes multi-scale bidirectional strip convolution and deep convolution to further refine edge features, and reconstructs HR-HSI by learning the features of different expansion roll branches by connecting across scales. Comparative experiments with several state-of-the-art HSI-MSI fusion algorithms on four publicly available datasets, CAVE, Chikusei, Houston and WorldView-3 are conducted to validate the effectiveness and superiority of AMSF-Net. On the Chikusei dataset, improvements were 9.1%, 12.5%, and 5.1%, respectively, on the indicators RMSE, ERGAS, and SAM, compared to the suboptimal method.
