Pub Date : 2025-09-01Epub Date: 2025-04-12DOI: 10.1016/j.aiia.2025.03.009
Wang Dai , Kebiao Mao , Zhonghua Guo , Zhihao Qin , Jiancheng Shi , Sayed M. Bateni , Liurui Xiao
The rapid advancement of artificial intelligence in domains such as natural language processing has catalyzed AI research across various fields. This study introduces a novel strategy, the AutoKeras-Knowledge Distillation (AK-KD), which integrates knowledge distillation technology for joint optimization of large and small models in the retrieval of surface temperature and emissivity using thermal infrared remote sensing. The approach addresses the challenges of limited accuracy in surface temperature retrieval by employing a high-performance large model developed through AutoKeras as the teacher model, which subsequently enhances a less accurate small model through knowledge distillation. The resultant student model is interactively integrated with the large model to further improve specificity and generalization capabilities. Theoretical derivations and practical applications validate that the AK-KD strategy significantly enhances the accuracy of temperature and emissivity retrieval. For instance, a large model trained with simulated ASTER data achieved a Pearson Correlation Coefficient (PCC) of 0.999 and a Mean Absolute Error (MAE) of 0.348 K in surface temperature retrieval. In practical applications, this model demonstrated a PCC of 0.967 and an MAE of 0.685 K. Although the large model exhibits high average accuracy, its precision in complex terrains is comparatively lower. To ameliorate this, the large model, serving as a teacher, enhances the small model's local accuracy. Specifically, in surface temperature retrieval, the small model's PCC improved from an average of 0.978 to 0.979, and the MAE decreased from 1.065 K to 0.724 K. In emissivity retrieval, the PCC rose from an average of 0.827 to 0.898, and the MAE reduced from 0.0076 to 0.0054. This research not only provides robust technological support for further development of thermal infrared remote sensing in temperature and emissivity retrieval but also offers important references and key technological insights for the universal model construction of other geophysical parameter retrievals.
{"title":"Joint optimization of AI large and small models for surface temperature and emissivity retrieval using knowledge distillation","authors":"Wang Dai , Kebiao Mao , Zhonghua Guo , Zhihao Qin , Jiancheng Shi , Sayed M. Bateni , Liurui Xiao","doi":"10.1016/j.aiia.2025.03.009","DOIUrl":"10.1016/j.aiia.2025.03.009","url":null,"abstract":"<div><div>The rapid advancement of artificial intelligence in domains such as natural language processing has catalyzed AI research across various fields. This study introduces a novel strategy, the AutoKeras-Knowledge Distillation (AK-KD), which integrates knowledge distillation technology for joint optimization of large and small models in the retrieval of surface temperature and emissivity using thermal infrared remote sensing. The approach addresses the challenges of limited accuracy in surface temperature retrieval by employing a high-performance large model developed through AutoKeras as the teacher model, which subsequently enhances a less accurate small model through knowledge distillation. The resultant student model is interactively integrated with the large model to further improve specificity and generalization capabilities. Theoretical derivations and practical applications validate that the AK-KD strategy significantly enhances the accuracy of temperature and emissivity retrieval. For instance, a large model trained with simulated ASTER data achieved a Pearson Correlation Coefficient (PCC) of 0.999 and a Mean Absolute Error (MAE) of 0.348 K in surface temperature retrieval. In practical applications, this model demonstrated a PCC of 0.967 and an MAE of 0.685 K. Although the large model exhibits high average accuracy, its precision in complex terrains is comparatively lower. To ameliorate this, the large model, serving as a teacher, enhances the small model's local accuracy. Specifically, in surface temperature retrieval, the small model's PCC improved from an average of 0.978 to 0.979, and the MAE decreased from 1.065 K to 0.724 K. In emissivity retrieval, the PCC rose from an average of 0.827 to 0.898, and the MAE reduced from 0.0076 to 0.0054. This research not only provides robust technological support for further development of thermal infrared remote sensing in temperature and emissivity retrieval but also offers important references and key technological insights for the universal model construction of other geophysical parameter retrievals.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 3","pages":"Pages 407-425"},"PeriodicalIF":8.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143837982","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01Epub Date: 2025-01-27DOI: 10.1016/j.aiia.2025.01.011
Leng Han , Zhichong Wang , Miao He , Yajia Liu , Xiongkui He
Precision application in orchards enhancing deposition uniformity and environmental sustainability by accurately matching nozzle output with canopy parameters. This study provides a pipeline for creating 3D prescription maps using a UAV and performing offline variable application. It also evaluates the accuracy of ground altitude measurements at various flight heights. At a flight height of 30 m, with a three-dimensional reconstruction method without phase-control points, the root mean square error (RMSE) for ground altitude measurement was 0.214 m and the mean absolute error (MAE) was 0.211 m; for the canopy area, these values were 0.591 m and 0.541 m, respectively. As flight height increased, the accuracy of altitude measurements declined and tended to be underestimated. Moreover, during offline variable spraying, the shape of the spray area influenced deposition accuracy, with collision detection area of a line segment achieving greater precision than conical ones. Field tests showed that the offline variable application method reduced pesticide usage by 32.43 % and enhanced spray uniformity. This newly developed process does not require costly sensors on each sprayer and has potential for field applications.
{"title":"PWM offline variable application based on UAV remote sensing 3D prescription map","authors":"Leng Han , Zhichong Wang , Miao He , Yajia Liu , Xiongkui He","doi":"10.1016/j.aiia.2025.01.011","DOIUrl":"10.1016/j.aiia.2025.01.011","url":null,"abstract":"<div><div>Precision application in orchards enhancing deposition uniformity and environmental sustainability by accurately matching nozzle output with canopy parameters. This study provides a pipeline for creating 3D prescription maps using a UAV and performing offline variable application. It also evaluates the accuracy of ground altitude measurements at various flight heights. At a flight height of 30 m, with a three-dimensional reconstruction method without phase-control points, the root mean square error (RMSE) for ground altitude measurement was 0.214 m and the mean absolute error (MAE) was 0.211 m; for the canopy area, these values were 0.591 m and 0.541 m, respectively. As flight height increased, the accuracy of altitude measurements declined and tended to be underestimated. Moreover, during offline variable spraying, the shape of the spray area influenced deposition accuracy, with collision detection area of a line segment achieving greater precision than conical ones. Field tests showed that the offline variable application method reduced pesticide usage by 32.43 % and enhanced spray uniformity. This newly developed process does not require costly sensors on each sprayer and has potential for field applications.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 3","pages":"Pages 496-507"},"PeriodicalIF":8.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144106612","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01Epub Date: 2025-02-13DOI: 10.1016/j.aiia.2025.02.002
Zhenni He , Fahui Yuan , Yansuo Zhou , Bingbo Cui , Yong He , Yufei Liu
At present, automatic broccoli harvest in field still faces some issues. It is difficult to segment broccoli in real time under complex field background, and hard to pick tilt-growing broccoli for the end-effector of robot. In this research, an improved YOLOv8n-seg model, named YOLO-Broccoli-Seg was proposed for broccoli recognition. Through adding a triplet attention module to YOLOv8-Seg model, the feature fusion capability of the algorithm is improved significantly. The mean average precision mAP50 (Mask), mAP95 (Mask), mAP50 (Bounding Box, Bbox) and mAP95 (Bbox) of YOLO-Broccoli-Seg are 0.973, 0.683, 0.973 and 0.748 respectively. Precision P-value was improved the most, with an increment of 8.7 %. In addition, an attitude estimation method based on three-dimensional point cloud is proposed. When the tilt angle of broccoli is between −30°and 30°, the R2 between the estimated value and the true value is 0.934. It indicated that this method can well represent the growth attitude of broccoli. This research can provide the rich broccoli information and technical basis for the automated broccoli picking.
{"title":"Stereo vision based broccoli recognition and attitude estimation method for field harvesting","authors":"Zhenni He , Fahui Yuan , Yansuo Zhou , Bingbo Cui , Yong He , Yufei Liu","doi":"10.1016/j.aiia.2025.02.002","DOIUrl":"10.1016/j.aiia.2025.02.002","url":null,"abstract":"<div><div>At present, automatic broccoli harvest in field still faces some issues. It is difficult to segment broccoli in real time under complex field background, and hard to pick tilt-growing broccoli for the end-effector of robot. In this research, an improved YOLOv8n-seg model, named YOLO-Broccoli-Seg was proposed for broccoli recognition. Through adding a triplet attention module to YOLOv8-Seg model, the feature fusion capability of the algorithm is improved significantly. The mean average precision mAP50 (Mask), mAP95 (Mask), mAP50 (Bounding Box, Bbox) and mAP95 (Bbox) of YOLO-Broccoli-Seg are 0.973, 0.683, 0.973 and 0.748 respectively. Precision <em>P</em>-value was improved the most, with an increment of 8.7 %. In addition, an attitude estimation method based on three-dimensional point cloud is proposed. When the tilt angle of broccoli is between −30°and 30°, the R<sup>2</sup> between the estimated value and the true value is 0.934. It indicated that this method can well represent the growth attitude of broccoli. This research can provide the rich broccoli information and technical basis for the automated broccoli picking.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 3","pages":"Pages 526-536"},"PeriodicalIF":8.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144106614","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Sudden reductions in crop yield (i.e., yield shocks) severely disrupt the food supply, intensify food insecurity, depress farmers' welfare, and worsen a country's economic conditions. Here, we study the spatiotemporal patterns of wheat yield shocks, quantified by the lower quantiles of yield fluctuations, in 86 countries over 30 years. Furthermore, we assess the relationships between shocks and their key ecological and socioeconomic drivers using quantile regression based on statistical (linear quantile mixed model) and machine learning (quantile random forest) models. Using a panel dataset that captures spatiotemporal patterns of yield shocks and possible drivers in 86 countries, we find that the severity of yield shocks has been increasing globally since 1997. Moreover, our cross-validation exercise shows that quantile random forest outperforms the linear quantile regression model. Despite this performance difference, both models consistently reveal that the severity of shocks is associated with higher weather stress, nitrogen fertilizer application rate, and gross domestic product (GDP) per capita (a typical indicator for economic and technological advancement in a country). While the unexpected negative association between more severe wheat yield shocks and higher fertilizer application rate and GDP per capita does not imply a direct causal effect, they indicate that the advancement in wheat production has been primarily on achieving higher yields and less on lowering the possibility and magnitude of sharp yield reductions. Hence, in the context of growing extreme weather stress, there is a critical need to enhance the technology and management practices that mitigate yield shocks to improve the resilience of the world food systems.
{"title":"Unveiling the drivers contributing to global wheat yield shocks through quantile regression","authors":"Srishti Vishwakarma , Xin Zhang , Vyacheslav Lyubchich","doi":"10.1016/j.aiia.2025.03.004","DOIUrl":"10.1016/j.aiia.2025.03.004","url":null,"abstract":"<div><div>Sudden reductions in crop yield (i.e., yield shocks) severely disrupt the food supply, intensify food insecurity, depress farmers' welfare, and worsen a country's economic conditions. Here, we study the spatiotemporal patterns of wheat yield shocks, quantified by the lower quantiles of yield fluctuations, in 86 countries over 30 years. Furthermore, we assess the relationships between shocks and their key ecological and socioeconomic drivers using quantile regression based on statistical (linear quantile mixed model) and machine learning (quantile random forest) models. Using a panel dataset that captures spatiotemporal patterns of yield shocks and possible drivers in 86 countries, we find that the severity of yield shocks has been increasing globally since 1997. Moreover, our cross-validation exercise shows that quantile random forest outperforms the linear quantile regression model. Despite this performance difference, both models consistently reveal that the severity of shocks is associated with higher weather stress, nitrogen fertilizer application rate, and gross domestic product (GDP) per capita (a typical indicator for economic and technological advancement in a country). While the unexpected negative association between more severe wheat yield shocks and higher fertilizer application rate and GDP per capita does not imply a direct causal effect, they indicate that the advancement in wheat production has been primarily on achieving higher yields and less on lowering the possibility and magnitude of sharp yield reductions. Hence, in the context of growing extreme weather stress, there is a critical need to enhance the technology and management practices that mitigate yield shocks to improve the resilience of the world food systems.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 3","pages":"Pages 564-572"},"PeriodicalIF":8.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144106557","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01Epub Date: 2025-04-14DOI: 10.1016/j.aiia.2025.04.005
Yi Zhang , Yu Zhang , Meng Gao , Xinjie Wang , Baisheng Dai , Weizheng Shen
The recognition of dairy cow behavior is essential for enhancing health management, reproductive efficiency, production performance, and animal welfare. This paper addresses the challenge of modality loss in multimodal dairy cow behavior recognition algorithms, which can be caused by sensor or video signal disturbances arising from interference, harsh environmental conditions, extreme weather, network fluctuations, and other complexities inherent in farm environments. This study introduces a modality mapping completion network that maps incomplete sensor and video data to improve multimodal dairy cow behavior recognition under conditions of modality loss. By mapping incomplete sensor or video data, the method applies a multimodal behavior recognition algorithm to identify five specific behaviors: drinking, feeding, lying, standing, and walking. The results indicate that, under various comprehensive missing coefficients (λ), the method achieves an average accuracy of 97.87 % ± 0.15 %, an average precision of 95.19 % ± 0.4 %, and an average F1 score of 94.685 % ± 0.375 %, with an overall accuracy of 94.67 % ± 0.37 %. This approach enhances the robustness and applicability of cow behavior recognition based on multimodal data in situations of modality loss, resolving practical issues in the development of digital twins for cow behavior and providing comprehensive support for the intelligent and precise management of farms.
{"title":"Multimodal behavior recognition for dairy cow digital twin construction under incomplete modalities: A modality mapping completion network approach","authors":"Yi Zhang , Yu Zhang , Meng Gao , Xinjie Wang , Baisheng Dai , Weizheng Shen","doi":"10.1016/j.aiia.2025.04.005","DOIUrl":"10.1016/j.aiia.2025.04.005","url":null,"abstract":"<div><div>The recognition of dairy cow behavior is essential for enhancing health management, reproductive efficiency, production performance, and animal welfare. This paper addresses the challenge of modality loss in multimodal dairy cow behavior recognition algorithms, which can be caused by sensor or video signal disturbances arising from interference, harsh environmental conditions, extreme weather, network fluctuations, and other complexities inherent in farm environments. This study introduces a modality mapping completion network that maps incomplete sensor and video data to improve multimodal dairy cow behavior recognition under conditions of modality loss. By mapping incomplete sensor or video data, the method applies a multimodal behavior recognition algorithm to identify five specific behaviors: drinking, feeding, lying, standing, and walking. The results indicate that, under various comprehensive missing coefficients (λ), the method achieves an average accuracy of 97.87 % ± 0.15 %, an average precision of 95.19 % ± 0.4 %, and an average F1 score of 94.685 % ± 0.375 %, with an overall accuracy of 94.67 % ± 0.37 %. This approach enhances the robustness and applicability of cow behavior recognition based on multimodal data in situations of modality loss, resolving practical issues in the development of digital twins for cow behavior and providing comprehensive support for the intelligent and precise management of farms.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 3","pages":"Pages 459-469"},"PeriodicalIF":8.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143873553","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01Epub Date: 2025-04-04DOI: 10.1016/j.aiia.2025.03.008
Rabiu Aminu , Samantha M. Cook , David Ljungberg , Oliver Hensel , Abozar Nasirahmadi
<div><div>To reduce damage caused by insect pests, farmers use insecticides to protect produce from crop pests. This practice leads to high synthetic chemical usage because a large portion of the applied insecticide does not reach its intended target; instead, it may affect non-target organisms and pollute the environment. One approach to mitigating this is through the selective application of insecticides to only those crop plants (or patches of plants) where the insect pests are located, avoiding non-targets and beneficials. The first step to achieve this is the identification of insects on plants and discrimination between pests and beneficial non-targets. However, detecting small-sized individual insects is challenging using image-based machine learning techniques, especially in natural field settings. This paper proposes a method based on explainable artificial intelligence feature selection and machine learning to detect pests and beneficial insects in field crops. An insect-plant dataset reflecting real field conditions was created. It comprises two pest insects—the Colorado potato beetle (CPB, <em>Leptinotarsa decemlineata</em>) and green peach aphid (<em>Myzus persicae</em>)—and the beneficial seven-spot ladybird (<em>Coccinella septempunctata</em>). The specialist herbivore CPB was imaged only on potato plants (<em>Solanum tuberosum</em>) while green peach aphids and seven-spot ladybirds were imaged on three crops: potato, faba bean (<em>Vicia faba)</em>, and sugar beet (<em>Beta vulgaris</em> subsp. <em>vulgaris</em>). This increased dataset diversity, broadening the potential application of the developed method for discriminating between pests and beneficial insects in several crops. The insects were imaged in both laboratory and outdoor settings. Using the GrabCut algorithm, regions of interest in the image were identified before shape, texture, and colour features were extracted from the segmented regions. The concept of explainable artificial intelligence was adopted by incorporating permutation feature importance ranking and Shapley Additive explanations values to identify the feature set that optimized a model's performance while reducing computational complexity. The proposed explainable artificial intelligence feature selection method was compared to conventional feature selection techniques, including mutual information, chi-square coefficient, maximal information coefficient, Fisher separation criterion and variance thresholding. Results showed improved accuracy (92.62 % Random forest, 90.16 % Support vector machine, 83.61 % K-nearest neighbours, and 81.97 % Naïve Bayes) and a reduction in the number of model parameters and memory usage (7.22 <em>×</em> 10<sup>7</sup> Random forest, 6.23 <em>×</em> 10<sup>3</sup> Support vector machine, 3.64 <em>×</em> 10<sup>4</sup> K-nearest neighbours and 1.88 <em>×</em> 10<sup>2</sup> Naïve Bayes) compared to using all features. Prediction and training times were also reduced by approxima
{"title":"Improving the performance of machine learning algorithms for detection of individual pests and beneficial insects using feature selection techniques","authors":"Rabiu Aminu , Samantha M. Cook , David Ljungberg , Oliver Hensel , Abozar Nasirahmadi","doi":"10.1016/j.aiia.2025.03.008","DOIUrl":"10.1016/j.aiia.2025.03.008","url":null,"abstract":"<div><div>To reduce damage caused by insect pests, farmers use insecticides to protect produce from crop pests. This practice leads to high synthetic chemical usage because a large portion of the applied insecticide does not reach its intended target; instead, it may affect non-target organisms and pollute the environment. One approach to mitigating this is through the selective application of insecticides to only those crop plants (or patches of plants) where the insect pests are located, avoiding non-targets and beneficials. The first step to achieve this is the identification of insects on plants and discrimination between pests and beneficial non-targets. However, detecting small-sized individual insects is challenging using image-based machine learning techniques, especially in natural field settings. This paper proposes a method based on explainable artificial intelligence feature selection and machine learning to detect pests and beneficial insects in field crops. An insect-plant dataset reflecting real field conditions was created. It comprises two pest insects—the Colorado potato beetle (CPB, <em>Leptinotarsa decemlineata</em>) and green peach aphid (<em>Myzus persicae</em>)—and the beneficial seven-spot ladybird (<em>Coccinella septempunctata</em>). The specialist herbivore CPB was imaged only on potato plants (<em>Solanum tuberosum</em>) while green peach aphids and seven-spot ladybirds were imaged on three crops: potato, faba bean (<em>Vicia faba)</em>, and sugar beet (<em>Beta vulgaris</em> subsp. <em>vulgaris</em>). This increased dataset diversity, broadening the potential application of the developed method for discriminating between pests and beneficial insects in several crops. The insects were imaged in both laboratory and outdoor settings. Using the GrabCut algorithm, regions of interest in the image were identified before shape, texture, and colour features were extracted from the segmented regions. The concept of explainable artificial intelligence was adopted by incorporating permutation feature importance ranking and Shapley Additive explanations values to identify the feature set that optimized a model's performance while reducing computational complexity. The proposed explainable artificial intelligence feature selection method was compared to conventional feature selection techniques, including mutual information, chi-square coefficient, maximal information coefficient, Fisher separation criterion and variance thresholding. Results showed improved accuracy (92.62 % Random forest, 90.16 % Support vector machine, 83.61 % K-nearest neighbours, and 81.97 % Naïve Bayes) and a reduction in the number of model parameters and memory usage (7.22 <em>×</em> 10<sup>7</sup> Random forest, 6.23 <em>×</em> 10<sup>3</sup> Support vector machine, 3.64 <em>×</em> 10<sup>4</sup> K-nearest neighbours and 1.88 <em>×</em> 10<sup>2</sup> Naïve Bayes) compared to using all features. Prediction and training times were also reduced by approxima","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 3","pages":"Pages 377-394"},"PeriodicalIF":8.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143835178","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01Epub Date: 2025-04-11DOI: 10.1016/j.aiia.2025.04.002
Hao Fu , Xueguan Zhao , Haoran Tan , Shengyu Zheng , Changyuan Zhai , Liping Chen
To address the low recognition accuracy of open-field vegetables under light occlusion, this study focused on cabbage and developed an online target recognition model based on deep learning. Using Yolov8n as the base network, a method was proposed to mitigate the impact of light occlusion on the accuracy of online cabbage recognition. A combination of cabbage image filters was designed to eliminate the effects of light occlusion. A filter parameter adaptive learning module for cabbage image filter parameters was constructed. The image filter combination and adaptive learning module were embedded into the Yolov8n object detection network. This integration enabled precise real-time recognition of cabbage under light occlusion conditions. Experimental results showed recognition accuracies of 97.5 % on the normal lighting dataset, 93.1 % on the light occlusion dataset, and 95.0 % on the mixed dataset. For images with a light occlusion degree greater than 0.4, the recognition accuracy improved by 9.9 % and 13.7 % compared to Yolov5n and Yolov8n models. The model achieved recognition accuracies of 99.3 % on the Chinese cabbage dataset and 98.3 % on the broccoli dataset. The model was deployed on an Nvidia Jetson Orin NX edge computing device, achieving an image processing speed of 26.32 frames per second. Field trials showed recognition accuracies of 96.0 % under normal lighting conditions and 91.2 % under light occlusion. The proposed online cabbage recognition model enables real-time recognition and localization of cabbage in complex open-field environments, offering technical support for target-oriented spraying.
针对光照遮挡下露地蔬菜识别准确率低的问题，本研究以白菜为研究对象，开发了一种基于深度学习的在线目标识别模型。以Yolov8n为基础网络，提出了一种减轻光遮挡对白菜在线识别精度影响的方法。白菜图像过滤器的组合设计，以消除光遮挡的影响。构建了白菜图像滤波参数的滤波参数自适应学习模块。将图像滤波组合和自适应学习模块嵌入到Yolov8n目标检测网络中。这种整合使得在光遮挡条件下对卷心菜进行精确的实时识别。实验结果表明，正常光照数据集的识别准确率为97.5%，光遮挡数据集的识别准确率为93.1%，混合数据集的识别准确率为95.0%。对于光遮挡度大于0.4的图像，与Yolov5n和Yolov8n模型相比，识别准确率分别提高了9.9%和13.7%。该模型对大白菜和西兰花的识别准确率分别达到99.3%和98.3%。该模型部署在Nvidia Jetson Orin NX边缘计算设备上，实现了每秒26.32帧的图像处理速度。野外试验表明，在正常光照条件下识别准确率为96.0%，在光遮挡条件下识别准确率为91.2%。所提出的在线大白菜识别模型能够实现复杂开阔环境下大白菜的实时识别和定位，为定向喷洒提供技术支持。
{"title":"Effective methods for mitigate the impact of light occlusion on the accuracy of online cabbage recognition in open fields","authors":"Hao Fu , Xueguan Zhao , Haoran Tan , Shengyu Zheng , Changyuan Zhai , Liping Chen","doi":"10.1016/j.aiia.2025.04.002","DOIUrl":"10.1016/j.aiia.2025.04.002","url":null,"abstract":"<div><div>To address the low recognition accuracy of open-field vegetables under light occlusion, this study focused on cabbage and developed an online target recognition model based on deep learning. Using Yolov8n as the base network, a method was proposed to mitigate the impact of light occlusion on the accuracy of online cabbage recognition. A combination of cabbage image filters was designed to eliminate the effects of light occlusion. A filter parameter adaptive learning module for cabbage image filter parameters was constructed. The image filter combination and adaptive learning module were embedded into the Yolov8n object detection network. This integration enabled precise real-time recognition of cabbage under light occlusion conditions. Experimental results showed recognition accuracies of 97.5 % on the normal lighting dataset, 93.1 % on the light occlusion dataset, and 95.0 % on the mixed dataset. For images with a light occlusion degree greater than 0.4, the recognition accuracy improved by 9.9 % and 13.7 % compared to Yolov5n and Yolov8n models. The model achieved recognition accuracies of 99.3 % on the Chinese cabbage dataset and 98.3 % on the broccoli dataset. The model was deployed on an Nvidia Jetson Orin NX edge computing device, achieving an image processing speed of 26.32 frames per second. Field trials showed recognition accuracies of 96.0 % under normal lighting conditions and 91.2 % under light occlusion. The proposed online cabbage recognition model enables real-time recognition and localization of cabbage in complex open-field environments, offering technical support for target-oriented spraying.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 3","pages":"Pages 449-458"},"PeriodicalIF":8.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143855790","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01Epub Date: 2025-03-17DOI: 10.1016/j.aiia.2025.03.003
Mehdi Fasihi , Mirko Sodini , Alex Falcon , Francesco Degano , Paolo Sivilotti , Giuseppe Serra
Predicting grapevine phenological stages (GPHS) is critical for precisely managing vineyard operations, including plant disease treatments, pruning, and harvest. Solutions commonly used to address viticulture challenges rely on image processing techniques, which have achieved significant results. However, they require the installation of dedicated hardware in the vineyard, making it invasive and difficult to maintain. Moreover, accurate prediction is influenced by the interplay of climatic factors, especially temperature, and the impact of global warming, which are difficult to model using images. Another problem frequently found in GPHS prediction is the persistent issue of missing values in viticultural datasets, particularly in phenological stages. This paper proposes a semi-supervised approach that begins with a small set of labeled phenological stage examples and automatically generates new annotations for large volumes of unlabeled climatic data. This approach aims to address key challenges in phenological analysis. This novel climatic data-based approach offers advantages over common image processing methods, as it is non-intrusive, cost-effective, and adaptable for vineyards of various sizes and technological levels. To ensure the robustness of the proposed Pseudo-labelling strategy, we integrated it into eight machine-learning algorithms. We evaluated its performance across seven diverse datasets, each exhibiting varying percentages of missing values. Performance metrics, including the coefficient of determination (R2) and root-mean-square error (RMSE), are employed to assess the effectiveness of the models. The study demonstrates that integrating the proposed Pseudo-labeling strategy with supervised learning approaches significantly improves predictive accuracy. Moreover, the study shows that the proposed methodology can also be integrated with explainable artificial intelligence techniques to determine the importance of the input features. In particular, the investigation highlights that growing degree days are crucial for improved GPHS prediction.
{"title":"Boosting grapevine phenological stages prediction based on climatic data by pseudo-labeling approach","authors":"Mehdi Fasihi , Mirko Sodini , Alex Falcon , Francesco Degano , Paolo Sivilotti , Giuseppe Serra","doi":"10.1016/j.aiia.2025.03.003","DOIUrl":"10.1016/j.aiia.2025.03.003","url":null,"abstract":"<div><div>Predicting grapevine phenological stages (GPHS) is critical for precisely managing vineyard operations, including plant disease treatments, pruning, and harvest. Solutions commonly used to address viticulture challenges rely on image processing techniques, which have achieved significant results. However, they require the installation of dedicated hardware in the vineyard, making it invasive and difficult to maintain. Moreover, accurate prediction is influenced by the interplay of climatic factors, especially temperature, and the impact of global warming, which are difficult to model using images. Another problem frequently found in GPHS prediction is the persistent issue of missing values in viticultural datasets, particularly in phenological stages. This paper proposes a semi-supervised approach that begins with a small set of labeled phenological stage examples and automatically generates new annotations for large volumes of unlabeled climatic data. This approach aims to address key challenges in phenological analysis. This novel climatic data-based approach offers advantages over common image processing methods, as it is non-intrusive, cost-effective, and adaptable for vineyards of various sizes and technological levels. To ensure the robustness of the proposed Pseudo-labelling strategy, we integrated it into eight machine-learning algorithms. We evaluated its performance across seven diverse datasets, each exhibiting varying percentages of missing values. Performance metrics, including the coefficient of determination (R<sup>2</sup>) and root-mean-square error (RMSE), are employed to assess the effectiveness of the models. The study demonstrates that integrating the proposed Pseudo-labeling strategy with supervised learning approaches significantly improves predictive accuracy. Moreover, the study shows that the proposed methodology can also be integrated with explainable artificial intelligence techniques to determine the importance of the input features. In particular, the investigation highlights that growing degree days are crucial for improved GPHS prediction.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 3","pages":"Pages 550-563"},"PeriodicalIF":8.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144106556","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01Epub Date: 2025-04-05DOI: 10.1016/j.aiia.2025.04.001
Josué Kpodo , A. Pouyan Nejadhashemi
Agricultural Extension (AE) research faces significant challenges in producing relevant and practical knowledge due to rapid advancements in artificial intelligence (AI). AE struggles to keep pace with these advancements, complicating the development of actionable information. One major challenge is the absence of intelligent platforms that enable efficient information retrieval and quick decision-making. Investigations have shown a shortage of AI-assisted solutions that effectively use AE materials across various media formats while preserving scientific accuracy and contextual relevance. Although mainstream AI systems can potentially reduce decision-making risks, their usage remains limited. This limitation arises primarily from the lack of standardized datasets and concerns regarding user data privacy. For AE datasets to be standardized, they must satisfy four key criteria: inclusion of critical domain-specific knowledge, expert curation, consistent structure, and acceptance by peers. Addressing data privacy issues involves adhering to open-access principles and enforcing strict data encryption and anonymization standards. To address these gaps, a conceptual framework is introduced. This framework extends beyond typical user-oriented platforms and comprises five core modules. It features a neurosymbolic pipeline integrating large language models with physically based agricultural modeling software, further enhanced by Reinforcement Learning from Human Feedback. Notable aspects of the framework include a dedicated human-in-the-loop process and a governance structure consisting of three primary bodies focused on data standardization, ethics and security, and accountability and transparency. Overall, this work represents a significant advancement in agricultural knowledge systems, potentially transforming how AE services deliver critical information to farmers and other stakeholders.
{"title":"Navigating challenges/opportunities in developing smart agricultural extension platforms: Multi-media data mining techniques","authors":"Josué Kpodo , A. Pouyan Nejadhashemi","doi":"10.1016/j.aiia.2025.04.001","DOIUrl":"10.1016/j.aiia.2025.04.001","url":null,"abstract":"<div><div>Agricultural Extension (AE) research faces significant challenges in producing relevant and practical knowledge due to rapid advancements in artificial intelligence (AI). AE struggles to keep pace with these advancements, complicating the development of actionable information. One major challenge is the absence of intelligent platforms that enable efficient information retrieval and quick decision-making. Investigations have shown a shortage of AI-assisted solutions that effectively use AE materials across various media formats while preserving scientific accuracy and contextual relevance. Although mainstream AI systems can potentially reduce decision-making risks, their usage remains limited. This limitation arises primarily from the lack of standardized datasets and concerns regarding user data privacy. For AE datasets to be standardized, they must satisfy four key criteria: inclusion of critical domain-specific knowledge, expert curation, consistent structure, and acceptance by peers. Addressing data privacy issues involves adhering to open-access principles and enforcing strict data encryption and anonymization standards. To address these gaps, a conceptual framework is introduced. This framework extends beyond typical user-oriented platforms and comprises five core modules. It features a neurosymbolic pipeline integrating large language models with physically based agricultural modeling software, further enhanced by Reinforcement Learning from Human Feedback. Notable aspects of the framework include a dedicated human-in-the-loop process and a governance structure consisting of three primary bodies focused on data standardization, ethics and security, and accountability and transparency. Overall, this work represents a significant advancement in agricultural knowledge systems, potentially transforming how AE services deliver critical information to farmers and other stakeholders.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 3","pages":"Pages 426-448"},"PeriodicalIF":8.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143842756","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01Epub Date: 2025-04-10DOI: 10.1016/j.aiia.2025.04.003
Shi Yinyan, Zhu Yangxu, Wang Xiaochan, Zhang Xiaolei, Zheng Enlai, Zhang Yongnian
Environmental impacts and economic demands are driving the development of variable rate fertilization (VRF) technology for precision agriculture. Despite the advantages of a simple structure, low cost and high efficiency, uneven fertilizer-spreading uniformity is becoming a key factor restricting the application of centrifugal fertilizer spreaders. Accordingly, the particle application characteristics and variation laws for centrifugal VRF spreaders with multi-pass overlapped spreading needs to be urgently explored, in order to improve their distribution uniformity and working accuracy. In this study, the working performance of a self-developed centrifugal VRF spreader, based on real-time growth information of rice and wheat, was investigated and tested through the test methods of using the collection trays prescribed in ISO 5690 and ASAE S341.2. The coefficient of variation (CV) was calculated by weighing the fertilizer mass in standard pans, in order to evaluate the distribution uniformity of spreading patterns. The results showed that the effective application widths were 21.05, 22.58 and 23.67 m for application rates of 225, 300 and 375 kg/ha, respectively. The actual fertilizer application rates of multi-pass overlapped spreading were generally higher than the target rates, as well as the particle distribution CVs within the effective spreading widths were 11.51, 9.25 and 11.28 % for the respective target rates. Field test results for multi-pass overlapped spreading showed that the average difference between the actual and target application was 4.54 %, as well as the average particle distribution CV within the operating width was 11.94 %, which met the operation requirements of particle transverse distribution for centrifugal fertilizer spreaders. The results and findings of this study provide a theoretical reference for technical innovation and development of centrifugal VRF spreaders and are of great practical and social significance for accelerating their application in implementing precision agriculture.
{"title":"Assessing particle application in multi-pass overlapping scenarios with variable rate centrifugal fertilizer spreaders for precision agriculture","authors":"Shi Yinyan, Zhu Yangxu, Wang Xiaochan, Zhang Xiaolei, Zheng Enlai, Zhang Yongnian","doi":"10.1016/j.aiia.2025.04.003","DOIUrl":"10.1016/j.aiia.2025.04.003","url":null,"abstract":"<div><div>Environmental impacts and economic demands are driving the development of variable rate fertilization (VRF) technology for precision agriculture. Despite the advantages of a simple structure, low cost and high efficiency, uneven fertilizer-spreading uniformity is becoming a key factor restricting the application of centrifugal fertilizer spreaders. Accordingly, the particle application characteristics and variation laws for centrifugal VRF spreaders with multi-pass overlapped spreading needs to be urgently explored, in order to improve their distribution uniformity and working accuracy. In this study, the working performance of a self-developed centrifugal VRF spreader, based on real-time growth information of rice and wheat, was investigated and tested through the test methods of using the collection trays prescribed in ISO 5690 and ASAE S341.2. The coefficient of variation (CV) was calculated by weighing the fertilizer mass in standard pans, in order to evaluate the distribution uniformity of spreading patterns. The results showed that the effective application widths were 21.05, 22.58 and 23.67 m for application rates of 225, 300 and 375 kg/ha, respectively. The actual fertilizer application rates of multi-pass overlapped spreading were generally higher than the target rates, as well as the particle distribution CVs within the effective spreading widths were 11.51, 9.25 and 11.28 % for the respective target rates. Field test results for multi-pass overlapped spreading showed that the average difference between the actual and target application was 4.54 %, as well as the average particle distribution CV within the operating width was 11.94 %, which met the operation requirements of particle transverse distribution for centrifugal fertilizer spreaders. The results and findings of this study provide a theoretical reference for technical innovation and development of centrifugal VRF spreaders and are of great practical and social significance for accelerating their application in implementing precision agriculture.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 3","pages":"Pages 395-406"},"PeriodicalIF":8.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143835179","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号