Computers and Electronics in Agriculture最新文献_第9页

Modeling the impact of multi-rotor UAV downwash on granular fertilizer distribution in precision agriculture 多旋翼无人机下冲对精准农业中颗粒肥料分布影响的建模

IF 8.9 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2026-01-16 DOI: 10.1016/j.compag.2025.111389

Wang Xunwei , Zhou Zhiyan , Chen Boqian , Deng Konghong , Lin Jianqin

Research on unmanned aerial vehicle (UAV) downwash has largely focused on spray applications, with millimeter-scale granular fertilizer spreading receiving considerably less attention. This study developed a coupled computational fluid dynamics and discrete element method (CFD-DEM) model to simulate wind field-particle interactions during fertilizer spreading, and its correctness was verified under specific parameters. Using this model, UAV downwash characteristics and their relationship with resultant deposition patterns were investigated, involving the three key parameters of the UAV rotor layout, flight speed, and fertilizer particle size. The results indicated that the axisymmetric rotor structure generated symmetric helical vortex tubes, which helped prevent skewness in the deposition pattern. The downwash wind field of four-rotor X-shaped under low-speed flight conditions manifested as four helical downward air columns. As flight speed increased, these columns tilted backward and subsequently detached from the ground, forming wake vortices. The symmetrical downwash airflow caused central deposition accumulation, reducing spreading range while inducing irregular patterns. Both reduced flight speeds and smaller particle sizes amplified these distortions: deposition patterns evolved from elliptical forms to irregular polygon at lower speeds/small size particles. Compared to no-wind conditions, the downwash reduced maximum effective width. Quantitative analysis via the relative deviation (RD) of lateral distribution curves evaluated the impact of wind field. The primary negative impact of downwash on spreading performance was a reduction in effective swath width. Based on the simulation results, it was reasonable to select the parameters of four-rotor X-shaped layout (RD = 19.45 %), flight speeds ≥ 5 m/s (RD ≤ 19.85 %) and fertilizer particle size ≥ 2 mm (RD ≤ 11.60 %) for operation. This study provides a valuable theoretical framework for predicting UAV fertilizer deposition patterns, while its broader applicability requires further validation across a wider parameter space, thereby contributing to the advancement of precision aerial application in agriculture.

无人机下洗的研究主要集中在喷雾应用上，毫米级颗粒肥料的施用受到的关注相对较少。本文建立了计算流体力学与离散元法（CFD-DEM）耦合模型，模拟了施肥过程中风场-粒子相互作用，并在特定参数下验证了模型的正确性。利用该模型，研究了无人机旋翼布局、飞行速度和肥料粒径三个关键参数下洗特性及其与沉积模式的关系。结果表明，轴对称转子结构产生了对称的螺旋涡管，有助于防止沉积模式的偏斜。低速飞行条件下，四旋翼x型下洗风场表现为四个螺旋向下的气柱。随着飞行速度的增加，这些柱子向后倾斜，随后与地面分离，形成尾流漩涡。对称的下冲气流导致中心沉积堆积，减小了扩散范围，同时形成不规则的形态。降低的飞行速度和较小的颗粒尺寸都放大了这些扭曲：在较低的速度/较小的颗粒下，沉积模式从椭圆形演变为不规则的多边形。与无风条件相比，下冲减少了最大有效宽度。通过横向分布曲线的相对偏差（RD）进行定量分析，评价风场的影响。下冲对铺展性能的主要负面影响是有效铺展宽度的减小。根据仿真结果，选择四旋翼x形布置（RD = 19.45%）、飞行速度≥5m /s （RD≤19.85%）、肥料粒度≥2mm （RD≤11.60%）的参数进行操作较为合理。该研究为预测无人机肥料沉积模式提供了有价值的理论框架，但其更广泛的适用性需要在更广泛的参数空间上进一步验证，从而有助于推进精准航空在农业中的应用。

{"title":"Modeling the impact of multi-rotor UAV downwash on granular fertilizer distribution in precision agriculture","authors":"Wang Xunwei , Zhou Zhiyan , Chen Boqian , Deng Konghong , Lin Jianqin","doi":"10.1016/j.compag.2025.111389","DOIUrl":"10.1016/j.compag.2025.111389","url":null,"abstract":"<div><div>Research on unmanned aerial vehicle (UAV) downwash has largely focused on spray applications, with millimeter-scale granular fertilizer spreading receiving considerably less attention. This study developed a coupled computational fluid dynamics and discrete element method (CFD-DEM) model to simulate wind field-particle interactions during fertilizer spreading, and its correctness was verified under specific parameters. Using this model, UAV downwash characteristics and their relationship with resultant deposition patterns were investigated, involving the three key parameters of the UAV rotor layout, flight speed, and fertilizer particle size. The results indicated that the axisymmetric rotor structure generated symmetric helical vortex tubes, which helped prevent skewness in the deposition pattern. The downwash wind field of four-rotor X-shaped under low-speed flight conditions manifested as four helical downward air columns. As flight speed increased, these columns tilted backward and subsequently detached from the ground, forming wake vortices. The symmetrical downwash airflow caused central deposition accumulation, reducing spreading range while inducing irregular patterns. Both reduced flight speeds and smaller particle sizes amplified these distortions: deposition patterns evolved from elliptical forms to irregular polygon at lower speeds/small size particles. Compared to no-wind conditions, the downwash reduced maximum effective width. Quantitative analysis via the relative deviation (<em>RD</em>) of lateral distribution curves evaluated the impact of wind field. The primary negative impact of downwash on spreading performance was a reduction in effective swath width. Based on the simulation results, it was reasonable to select the parameters of four-rotor X-shaped layout (<em>RD</em> = 19.45 %), flight speeds ≥ 5 m/s (<em>RD</em> ≤ 19.85 %) and fertilizer particle size ≥ 2 mm (<em>RD</em> ≤ 11.60 %) for operation. This study provides a valuable theoretical framework for predicting UAV fertilizer deposition patterns, while its broader applicability requires further validation across a wider parameter space, thereby contributing to the advancement of precision aerial application in agriculture.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"243 ","pages":"Article 111389"},"PeriodicalIF":8.9,"publicationDate":"2026-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979071","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Transformer based embedding in query-response generation system for agriculture 基于变压器的农业查询-响应生成系统嵌入

IF 8.9 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2026-01-16 DOI: 10.1016/j.compag.2026.111406

Y.Y.Narayana Reddy , Thulasiram Narayanan , Adusumalli Balaji

Information science can play a major role in promoting the sustainable agriculture goals of the country. The optimism is to design a query-response generation system that offers instant assistance to the farmers by fulfilling their demands. It is very difficult to develop a background of knowledge that can answer the questions of different farmers on plant protection. To handle this problem, the records of the previous eight years of the countrywide farmers helpline network in terms of the call logs are gathered and processed to produce the required knowledge base. The proposed Convolutional Progressive attention-based Bidirectional Encoder Representations of Transformers (CPA-BERT) model retrieves detailed features and gives highly contextualized representations, which allows to get a deeper insight into the text. Lastly, the latent representations enhance the oversight of the system in determining the responses that are significant to the query and responses by prioritizing the most significant features of queries and answers through a new long short-term memory enclosed Cross Variational Autoencoder (LSTM-CVA) model. The sample data applied in the study consists of the analysis of 34 million call records collected on the government operated Kisan Call Centre (KCC). The suggested solution shows performance in different metrics such as accuracy of 98.04, precision of 97.52, recall of 97.02 and the F1-score of 97.98. The suggested solution assists farmers in making proper choices regarding their farming processes because the solution provides quick and pertinent responses to their questions.

信息科学可以在促进国家可持续农业目标方面发挥重要作用。乐观的是设计一个查询-响应生成系统，通过满足农民的需求，为他们提供即时帮助。很难建立一个知识背景，可以回答不同农民在植物保护方面的问题。为了解决这一问题，收集全国农民求助热线网络前8年的通话记录，并对其进行处理，生成所需的知识库。所提出的基于卷积渐进注意的变形器双向编码器表示（CPA-BERT）模型检索详细特征并给出高度上下文化的表示，从而可以更深入地了解文本。最后，潜在表征通过一种新的长短期记忆封装交叉变分自编码器（LSTM-CVA）模型对查询和答案的最重要特征进行优先排序，从而增强了系统在确定对查询和响应重要的响应方面的监督。研究中使用的样本数据包括对政府运营的Kisan呼叫中心（KCC）收集的3400万条呼叫记录的分析。该方案在正确率为98.04，精密度为97.52，召回率为97.02，f1得分为97.98等不同指标上均表现出色。建议的解决方案有助于农民对其耕作过程做出正确的选择，因为该解决方案为他们的问题提供了快速和相关的回答。

{"title":"Transformer based embedding in query-response generation system for agriculture","authors":"Y.Y.Narayana Reddy , Thulasiram Narayanan , Adusumalli Balaji","doi":"10.1016/j.compag.2026.111406","DOIUrl":"10.1016/j.compag.2026.111406","url":null,"abstract":"<div><div>Information science can play a major role in promoting the sustainable agriculture goals of the country. The optimism is to design a query-response generation system that offers instant assistance to the farmers by fulfilling their demands. It is very difficult to develop a background of knowledge that can answer the questions of different farmers on plant protection. To handle this problem, the records of the previous eight years of the countrywide farmers helpline network in terms of the call logs are gathered and processed to produce the required knowledge base. The proposed Convolutional Progressive attention-based Bidirectional Encoder Representations of Transformers (CPA-BERT) model retrieves detailed features and gives highly contextualized representations, which allows to get a deeper insight into the text. Lastly, the latent representations enhance the oversight of the system in determining the responses that are significant to the query and responses by prioritizing the most significant features of queries and answers through a new long short-term memory enclosed Cross Variational Autoencoder (LSTM-CVA) model. The sample data applied in the study consists of the analysis of 34 million call records collected on the government operated Kisan Call Centre (KCC). The suggested solution shows performance in different metrics such as accuracy of 98.04, precision of 97.52, recall of 97.02 and the F1-score of 97.98. The suggested solution assists farmers in making proper choices regarding their farming processes because the solution provides quick and pertinent responses to their questions.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"244 ","pages":"Article 111406"},"PeriodicalIF":8.9,"publicationDate":"2026-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145982033","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

PMT-GAN: Pineapple maturity transformation GAN for enhanced classification network PMT-GAN：菠萝成熟度变换GAN增强分类网络

IF 8.9 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2026-01-16 DOI: 10.1016/j.compag.2026.111435

Jiehao Li , Yaowen Liu , Jiahuan Lu , Shan Zeng , Xiwen Luo , C.L. Philip Chen , Chenguang Yang

Accurate identification of pineapple maturity is of paramount significance for enhancing its market value and consumer satisfaction, and it also aids fruit growers in pinpointing the optimal harvesting time. However, in real-world cultivation settings, fluctuations in environmental factors such as light, temperature, and humidity give rise to high variability in the maturity characteristics of pineapple fruits. This results in an extremely limited amount of data for certain maturity stages of pineapple and, consequently, a highly uneven distribution of data across different maturity stages. These factors collectively impede the generalization capability of traditional convolutional neural network-based object detection. To enhance the accuracy and generalization performance of pineapple maturity classification algorithms, this paper introduces an improved Generative Adversarial Network (GAN) model, namely the Pineapple Maturity Transformation GAN (PMT-GAN). By optimizing the generator architecture and incorporating a multiscale feature extraction module, the model’s performance in handling substantial geometric transformations is significantly bolstered. Additionally, the integration of the Swin Transformer module further augments the model’s explicit semantic modeling capabilities. Moreover, the adoption of the SSIM as the cycle consistency loss effectively preserves the structural coherence of images. Experimental results demonstrate that the optimized cycle consistency loss is markedly lower than the conventional loss, with reduced oscillation amplitude. Through UMAP and Grad-CAM visualization techniques, a high degree of similarity between original and generated pineapple images is observed. Compared to the original CycleGAN network, PMT-GAN achieves a maximum reduction of 30.43 percentage points in FID and 38.29 percentage points in KID. Furthermore, the network exhibits commendable generalization performance in maturity transformation experiments with other fruit species. Ultimately, the dataset size for the four maturity stages of pineapple is successfully expanded, with sample numbers becoming more balanced across stages (Stage 1: 1787 images, Stage 2: 1847 images, Stage 3: 1848 images, Stage 4: 1825 images). Detection network experiments based on the expanded dataset indicate that the expanded dataset outperforms the original dataset in key performance metrics such as precision, recall, and mAP. Research findings confirm that PMT-GAN effectively enhances the generation of pineapple maturity data, possesses robust generalization capabilities, and generates data that better serve detection networks. This provides an important methodological reference for the fine-grained classification of fruit maturity and the generation and recognition of relevant data.

准确识别菠萝成熟度对于提高其市场价值和消费者满意度具有至关重要的意义，也有助于果农确定最佳采收时间。然而，在现实的栽培环境中，环境因素（如光、温度和湿度）的波动会导致菠萝果实成熟特征的高度变化。这导致菠萝某些成熟阶段的数据量极其有限，因此，不同成熟阶段的数据分布极不均匀。这些因素共同阻碍了传统的基于卷积神经网络的目标检测的泛化能力。为了提高菠萝成熟度分类算法的准确率和泛化性能，本文引入了一种改进的生成对抗网络（GAN）模型，即菠萝成熟度变换GAN （PMT-GAN）。通过优化生成器架构并结合多尺度特征提取模块，该模型在处理大量几何变换方面的性能得到了显著增强。此外，Swin Transformer模块的集成进一步增强了模型的显式语义建模功能。此外，采用SSIM作为周期一致性损失，有效地保持了图像的结构一致性。实验结果表明，优化后的周期一致性损耗明显低于常规损耗，振荡幅度减小。通过UMAP和Grad-CAM可视化技术，观察到原始菠萝图像与生成的菠萝图像高度相似。与原来的CycleGAN网络相比，PMT-GAN在FID和KID方面最大降低了30.43个百分点和38.29个百分点。此外，该网络在其他水果品种的成熟度变换实验中表现出良好的泛化性能。最终，成功地扩展了菠萝四个成熟阶段的数据集大小，各个阶段的样本数量变得更加平衡（阶段1:1787张图像，阶段2:1847张图像，阶段3:1848张图像，阶段4:1825张图像）。基于扩展数据集的检测网络实验表明，扩展数据集在精度、召回率和mAP等关键性能指标上优于原始数据集。研究结果证实，PMT-GAN有效增强了菠萝成熟度数据的生成，具有鲁棒的泛化能力，生成的数据能够更好地服务于检测网络。这为果实成熟度的细粒度分类及相关数据的生成和识别提供了重要的方法论参考。

{"title":"PMT-GAN: Pineapple maturity transformation GAN for enhanced classification network","authors":"Jiehao Li , Yaowen Liu , Jiahuan Lu , Shan Zeng , Xiwen Luo , C.L. Philip Chen , Chenguang Yang","doi":"10.1016/j.compag.2026.111435","DOIUrl":"10.1016/j.compag.2026.111435","url":null,"abstract":"<div><div>Accurate identification of pineapple maturity is of paramount significance for enhancing its market value and consumer satisfaction, and it also aids fruit growers in pinpointing the optimal harvesting time. However, in real-world cultivation settings, fluctuations in environmental factors such as light, temperature, and humidity give rise to high variability in the maturity characteristics of pineapple fruits. This results in an extremely limited amount of data for certain maturity stages of pineapple and, consequently, a highly uneven distribution of data across different maturity stages. These factors collectively impede the generalization capability of traditional convolutional neural network-based object detection. To enhance the accuracy and generalization performance of pineapple maturity classification algorithms, this paper introduces an improved Generative Adversarial Network (GAN) model, namely the Pineapple Maturity Transformation GAN (PMT-GAN). By optimizing the generator architecture and incorporating a multiscale feature extraction module, the model’s performance in handling substantial geometric transformations is significantly bolstered. Additionally, the integration of the Swin Transformer module further augments the model’s explicit semantic modeling capabilities. Moreover, the adoption of the SSIM as the cycle consistency loss effectively preserves the structural coherence of images. Experimental results demonstrate that the optimized cycle consistency loss is markedly lower than the conventional loss, with reduced oscillation amplitude. Through UMAP and Grad-CAM visualization techniques, a high degree of similarity between original and generated pineapple images is observed. Compared to the original CycleGAN network, PMT-GAN achieves a maximum reduction of 30.43 percentage points in FID and 38.29 percentage points in KID. Furthermore, the network exhibits commendable generalization performance in maturity transformation experiments with other fruit species. Ultimately, the dataset size for the four maturity stages of pineapple is successfully expanded, with sample numbers becoming more balanced across stages (Stage 1: 1787 images, Stage 2: 1847 images, Stage 3: 1848 images, Stage 4: 1825 images). Detection network experiments based on the expanded dataset indicate that the expanded dataset outperforms the original dataset in key performance metrics such as precision, recall, and mAP. Research findings confirm that PMT-GAN effectively enhances the generation of pineapple maturity data, possesses robust generalization capabilities, and generates data that better serve detection networks. This provides an important methodological reference for the fine-grained classification of fruit maturity and the generation and recognition of relevant data.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"243 ","pages":"Article 111435"},"PeriodicalIF":8.9,"publicationDate":"2026-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979160","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Pha-YOLO: A multi-view Phalaenopsis flower detection and counting method based on improved YOLO11 with dynamic reference selection and adaptive thresholding pa - yolo：基于改进的YOLO11动态参考选择和自适应阈值法的蝴蝶兰多视角花检测计数方法

IF 8.9 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2026-01-15 DOI: 10.1016/j.compag.2026.111446

Dawei Xu , Xiaopeng Huang , Zihe Zhao , Zhenyuan Zhao , Dongyang Hu , Chao Yuan

Phalaenopsis orchids are high-value ornamental plants and their commercial value depends on accurate flower counts for quality grading. However, manual counting is labor-intensive, inconsistent, and costly. Automated counting is challenging due to severe occlusion among overlapping flowers, dynamic scale changes during conveyor transport, and the inherent limitations of single-view detectors, which cannot capture complete spatial information. To address these challenges, we propose a multi-view flower-counting framework based on an enhanced YOLO architecture, termed Pha-YOLO. First, to counteract detection errors arising from dynamic scale variations on a conveyor belt, Pha-YOLO incorporates the lightweight FasterNet backbone, adds a P2 small-object detection layer, and integrates an Efficient Multi-scale Attention module, collectively boosting its ability to detect flowers across scales. Second, to overcome the limitations of single-view counting, we develop a multi-view fusion strategy that uses four synchronized cameras, dynamically selects the optimal reference view for coordinate alignment, and adaptively adjusts clustering thresholds based on scene attributes, thereby mitigating occlusions and preventing duplicate counts. Ablation experiments show that Pha-YOLO achieves 98.2% mean Average Precision (mAP), a 2.0% improvement over the baseline, while using only 2.454M parameters. In counting experiments, the proposed multi-view method reduces the average missing detection rate by 19.95% relative to single-view approaches and lowers the overall counting error by 4.40% compared with fixed-reference, fixed-threshold fusion. These results underscore the proposed framework’s superior accuracy and robustness for automated Phalaenopsis flower counting.

蝴蝶兰是一种高价值的观赏植物，其商业价值取决于准确的花数来进行质量分级。然而，手工计数是劳动密集型的，不一致的，而且成本很高。由于重叠花朵之间的严重遮挡，传送带运输过程中的动态尺度变化以及单视图检测器的固有局限性，无法捕获完整的空间信息，因此自动计数具有挑战性。为了解决这些挑战，我们提出了一个基于增强型YOLO架构的多视图花计数框架，称为Pha-YOLO。首先，为了抵消传送带上动态尺度变化引起的检测误差，Pha-YOLO结合了轻量级的FasterNet主干，增加了P2小物体检测层，并集成了一个高效的多尺度注意力模块，共同提高了其跨尺度检测花卉的能力。其次，为了克服单视图计数的局限性，我们开发了一种多视图融合策略，该策略使用四个同步摄像机，动态选择最佳参考视图进行坐标对齐，并根据场景属性自适应调整聚类阈值，从而减轻遮挡并防止重复计数。烧蚀实验表明，在仅使用2.454万个参数的情况下，Pha-YOLO的平均精度（mAP）达到98.2%，比基线提高2.0%。在计数实验中，与单视图方法相比，多视图方法的平均缺失检测率降低了19.95%，与固定参考、固定阈值融合相比，该方法的总计数误差降低了4.40%。这些结果强调了所提出的框架对蝴蝶兰花的自动计数具有优越的准确性和鲁棒性。

{"title":"Pha-YOLO: A multi-view Phalaenopsis flower detection and counting method based on improved YOLO11 with dynamic reference selection and adaptive thresholding","authors":"Dawei Xu , Xiaopeng Huang , Zihe Zhao , Zhenyuan Zhao , Dongyang Hu , Chao Yuan","doi":"10.1016/j.compag.2026.111446","DOIUrl":"10.1016/j.compag.2026.111446","url":null,"abstract":"<div><div>Phalaenopsis orchids are high-value ornamental plants and their commercial value depends on accurate flower counts for quality grading. However, manual counting is labor-intensive, inconsistent, and costly. Automated counting is challenging due to severe occlusion among overlapping flowers, dynamic scale changes during conveyor transport, and the inherent limitations of single-view detectors, which cannot capture complete spatial information. To address these challenges, we propose a multi-view flower-counting framework based on an enhanced YOLO architecture, termed Pha-YOLO. First, to counteract detection errors arising from dynamic scale variations on a conveyor belt, Pha-YOLO incorporates the lightweight FasterNet backbone, adds a P2 small-object detection layer, and integrates an Efficient Multi-scale Attention module, collectively boosting its ability to detect flowers across scales. Second, to overcome the limitations of single-view counting, we develop a multi-view fusion strategy that uses four synchronized cameras, dynamically selects the optimal reference view for coordinate alignment, and adaptively adjusts clustering thresholds based on scene attributes, thereby mitigating occlusions and preventing duplicate counts. Ablation experiments show that Pha-YOLO achieves 98.2% mean Average Precision (mAP), a 2.0% improvement over the baseline, while using only 2.454M parameters. In counting experiments, the proposed multi-view method reduces the average missing detection rate by 19.95% relative to single-view approaches and lowers the overall counting error by 4.40% compared with fixed-reference, fixed-threshold fusion. These results underscore the proposed framework’s superior accuracy and robustness for automated Phalaenopsis flower counting.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"243 ","pages":"Article 111446"},"PeriodicalIF":8.9,"publicationDate":"2026-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979159","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

UAV remote sensing-driven precision variable management in cotton: technological framework, applications, and research outlook 无人机遥感驱动的棉花精准变量管理：技术框架、应用与研究展望

IF 8.9 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2026-01-14 DOI: 10.1016/j.compag.2026.111426

Lechun Zhang , Yingkuan Wang , Xinyu Xue , Wenjiang Huang , Tianye Yang , Hang Zhu , Yubin Lan

This review focuses on unmanned aerial vehicle (UAV) remote sensing–driven precision variable management in cotton fields. It systematically examines the key processes and engineering constraints from observable indicators to executable prescriptions within the closed loop of sensing, prescription, execution, and quality feedback. First, the review back-derives observable quantities and payload configurations from task requirements. It compares RGB, multispectral, hyperspectral, thermal infrared, and LiDAR sensing modes in terms of coverage, resolution, and temporal window, and highlights the value of multisource information in reducing prescription uncertainty under strict geometric co-registration. Second, it summarizes the scale transformation from ground sample distance (GSD) to control units, proposing that denoising and connectivity purification should be completed within the computational domain before aggregation to achieve an effective prescription resolution. Spatial foresight compensation is further introduced to mitigate end-to-end latency, preventing high-frequency toggling and striping coverage. Furthermore, the review summarizes three prescription modeling paradigms (index, mechanism, and learning) and clarifies their complementarity, strengths, limitations, and suitable use cases. On the execution side, the review analyzes how prescription encoding interacts with onboard decoding constraints. It covers RTK/GNSS positioning, PWM frequency and duty-cycle stability ranges, pressure and flow dynamics, droplet spectrum categories, and operational meteorological windows, and provides practical recommendations on the minimum executable patch size and dual-threshold hysteresis. Overall, UAVs can reliably support monitoring, decision-making, and execution in cotton fields characterized by short operational windows and fine spatial heterogeneity. Establishing prescription-level metadata and standardized quality feedback mechanisms is essential to enable cross-field, cross-season generalization and large-scale implementation.

本文对无人机遥感驱动的棉田精准变量管理技术进行了综述。它系统地检查了关键过程和工程约束，从可观察的指标到可执行的处方，在传感、处方、执行和质量反馈的闭环中。首先，回顾从任务需求中得出可观察的数量和有效载荷配置。比较了RGB、多光谱、高光谱、热红外和LiDAR传感模式的覆盖范围、分辨率和时间窗口，强调了多源信息在严格几何共配下降低处方不确定性的价值。其次，总结了从地面样本距离（ground sample distance， GSD）到控制单元的尺度转换，提出在聚集前，应在计算域内完成去噪和连通性净化，以实现有效的处方分解。进一步引入空间前瞻补偿来缓解端到端延迟，防止高频切换和条带覆盖。此外，综述总结了三种处方建模范式（索引、机制和学习），并阐明了它们的互补性、优势、局限性和合适的用例。在执行方面，本文分析了处方编码如何与板载解码约束相互作用。它涵盖了RTK/GNSS定位、PWM频率和占空比稳定范围、压力和流量动力学、液滴频谱类别和业务气象窗口，并提供了关于最小可执行补丁大小和双阈值滞后的实用建议。总体而言，无人机可以可靠地支持棉花田的监测、决策和执行，其特点是操作窗口短，空间异质性好。建立处方级元数据和标准化质量反馈机制是实现跨领域、跨季节推广和大规模实施的必要条件。

{"title":"UAV remote sensing-driven precision variable management in cotton: technological framework, applications, and research outlook","authors":"Lechun Zhang , Yingkuan Wang , Xinyu Xue , Wenjiang Huang , Tianye Yang , Hang Zhu , Yubin Lan","doi":"10.1016/j.compag.2026.111426","DOIUrl":"10.1016/j.compag.2026.111426","url":null,"abstract":"<div><div>This review focuses on unmanned aerial vehicle (UAV) remote sensing–driven precision variable management in cotton fields. It systematically examines the key processes and engineering constraints from observable indicators to executable prescriptions within the closed loop of sensing, prescription, execution, and quality feedback. First, the review back-derives observable quantities and payload configurations from task requirements. It compares RGB, multispectral, hyperspectral, thermal infrared, and LiDAR sensing modes in terms of coverage, resolution, and temporal window, and highlights the value of multisource information in reducing prescription uncertainty under strict geometric co-registration. Second, it summarizes the scale transformation from ground sample distance (GSD) to control units, proposing that denoising and connectivity purification should be completed within the computational domain before aggregation to achieve an effective prescription resolution. Spatial foresight compensation is further introduced to mitigate end-to-end latency, preventing high-frequency toggling and striping coverage. Furthermore, the review summarizes three prescription modeling paradigms (index, mechanism, and learning) and clarifies their complementarity, strengths, limitations, and suitable use cases. On the execution side, the review analyzes how prescription encoding interacts with onboard decoding constraints. It covers RTK/GNSS positioning, PWM frequency and duty-cycle stability ranges, pressure and flow dynamics, droplet spectrum categories, and operational meteorological windows, and provides practical recommendations on the minimum executable patch size and dual-threshold hysteresis. Overall, UAVs can reliably support monitoring, decision-making, and execution in cotton fields characterized by short operational windows and fine spatial heterogeneity. Establishing prescription-level metadata and standardized quality feedback mechanisms is essential to enable cross-field, cross-season generalization and large-scale implementation.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"243 ","pages":"Article 111426"},"PeriodicalIF":8.9,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979074","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

MWG-YOLO: Multi-Path weighted guided YOLOv8 lightweight model for rose detection in field MWG-YOLO：多路径加权制导YOLOv8轻型玫瑰田间检测模型

IF 8.9 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2026-01-14 DOI: 10.1016/j.compag.2025.111384

Weijie Peng , Jianneng Chen , Zhiwei Chen , Zezhong Ding , Mengjie Wang , Zhengrui Tian , Dewen Wang , Chunwang Dong

The detection and classification of roses are esse growth increase the difficulty of rose detection in the field. At the same time, the complex model will limit the deployment of the identification model in the field environment with low computational power, and affect the detection efficiency of the model for roses of different maturity. Therefore, a lightweight rose maturity detection model based on the improved YOLOv8 is proposed. Firstly, to address the issues of small and overlapping target recognition, a plug-and-play MWG-FPN structure is proposed to replace the Concat layer, which fuses features from different levels and enhances the feature expression ability. The MWCA mechanism is designed as the feature extraction module of MWG-FPN to capture feature information from different directions and improve the extraction effect of target details. Secondly, a model pruning algorithm based on sparse regularization is used to prune redundant channels and reduce model complexity and volume.Furthermore, the channel distillation strategy (CWD) was adopted to address the decrease in detection accuracy caused by pruning, while maintaining the inference speed and model size. Finally, Bayesian optimization of hyperparameters was used to further explore the potential of the model. The results showed that the improved model achieved an mAP50 of 92.3 % in field detection, which was 4.4 % higher than that of the original model. The parameters and FLOPs were reduced by 16.98 M and 65.4G respectively, and the detection speed reached 141.9 fps. Additionally, to prove the effectiveness of the model in actual field operations, the model was deployed and generalized performance was tested. The test results indicated that the model exhibited excellent performance on edge devices and unfamiliar datasets, providing a theoretical basis for actual rose picking.

玫瑰品种的检测和分类日益复杂，增加了田间玫瑰品种检测的难度。同时，复杂的模型会限制识别模型在计算能力较低的野外环境下的部署，影响模型对不同成熟度玫瑰的检测效率。为此，提出了一种基于改进的YOLOv8的玫瑰成熟度轻量级检测模型。首先，针对小目标和重叠目标识别的问题，提出了一种即插即用的MWG-FPN结构来取代Concat层，融合了不同层次的特征，增强了特征表达能力；设计MWCA机制作为MWG-FPN的特征提取模块，从不同方向捕获特征信息，提高目标细节的提取效果。其次，采用基于稀疏正则化的模型剪枝算法对冗余信道进行剪枝，降低模型复杂度和体积；此外，在保持推理速度和模型大小的同时，采用通道蒸馏策略（CWD）解决了剪枝导致的检测精度下降的问题。最后，利用贝叶斯超参数优化进一步挖掘模型的潜力。结果表明，改进模型的田间检测mAP50为92.3%，比原模型提高了4.4个百分点。参数和FLOPs分别降低16.98 M和65.4G，检测速度达到141.9 fps。此外，为了证明该模型在实际现场作业中的有效性，对该模型进行了部署和广义性能测试。实验结果表明，该模型在边缘设备和陌生数据集上均表现出优异的性能，为实际玫瑰采摘提供了理论依据。

{"title":"MWG-YOLO: Multi-Path weighted guided YOLOv8 lightweight model for rose detection in field","authors":"Weijie Peng , Jianneng Chen , Zhiwei Chen , Zezhong Ding , Mengjie Wang , Zhengrui Tian , Dewen Wang , Chunwang Dong","doi":"10.1016/j.compag.2025.111384","DOIUrl":"10.1016/j.compag.2025.111384","url":null,"abstract":"<div><div>The detection and classification of roses are esse growth increase the difficulty of rose detection in the field. At the same time, the complex model will limit the deployment of the identification model in the field environment with low computational power, and affect the detection efficiency of the model for roses of different maturity. Therefore, a lightweight rose maturity detection model based on the improved YOLOv8 is proposed. Firstly, to address the issues of small and overlapping target recognition, a plug-and-play MWG-FPN structure is proposed to replace the Concat layer, which fuses features from different levels and enhances the feature expression ability. The MWCA mechanism is designed as the feature extraction module of MWG-FPN to capture feature information from different directions and improve the extraction effect of target details. Secondly, a model pruning algorithm based on sparse regularization is used to prune redundant channels and reduce model complexity and volume.Furthermore, the channel distillation strategy (CWD) was adopted to address the decrease in detection accuracy caused by pruning, while maintaining the inference speed and model size. Finally, Bayesian optimization of hyperparameters was used to further explore the potential of the model. The results showed that the improved model achieved an mAP50 of 92.3 % in field detection, which was 4.4 % higher than that of the original model. The parameters and FLOPs were reduced by 16.98 M and 65.4G respectively, and the detection speed reached 141.9 fps. Additionally, to prove the effectiveness of the model in actual field operations, the model was deployed and generalized performance was tested. The test results indicated that the model exhibited excellent performance on edge devices and unfamiliar datasets, providing a theoretical basis for actual rose picking.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"243 ","pages":"Article 111384"},"PeriodicalIF":8.9,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979076","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SIFNet: A spectrum-image fusion neural network in hyperspectral imaging for intelligent quality assessment of crude tea to ensure tea product availability SIFNet：高光谱成像中用于粗茶质量智能评估的光谱图像融合神经网络，以确保茶叶产品的可获得性

IF 8.9 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2026-01-14 DOI: 10.1016/j.compag.2026.111441

Dengshan Li , Zhenhao Xu , Rongxi Lin , Yixing Huang , Qin Ouyang , Zhonghua Liu

Accurately assessing the quality of crude tea (primary processed tea) is critical for ensuring the standardization and consistent supply of tea products. Hyperspectral imaging (HSI) has emerged as a powerful data-driven technology that utilizes molecular fingerprint spectra and spatial image information for rapid quality assessment of crude tea. However, spectral overlap and image similarity between different grades hinder the performance of traditional classification methods. To address these challenges, this study proposes a novel spectrum-image fusion neural network (SIFNet) that tackles spectral dependency and subtle inter-grade differences in image features. SIFNet consists of a spectral feature extraction branch and an image feature extraction branch. The spectral feature extraction combines one-dimensional convolutional neural networks (1D-CNN) and bidirectional long short-term memory networks (Bi-LSTM) to capture spectral dependencies. While the image feature extraction branch employs two-dimensional CNN (2D-CNN) and fuzzy operation modules to capture both precise and fuzzy image features. Finally, cross-modal features are fused using fully connected layers to improve classification performance. Five-fold cross-validation on the crude Wuyi rock tea hyperspectral image dataset demonstrated that SIFNet achieved an accuracy of 96.75 ± 2.27%, outperforming other state-of-the-art machine learning and deep learning algorithms. This study proposed a data-driven approach for intelligent quality assessment of crude tea, providing robust technical support for enhancing the sustainable supply capacity of the tea industry.

准确评估粗茶（初加工茶）的质量对于保证茶叶产品的标准化和一致性供应至关重要。高光谱成像（HSI）是一种利用分子指纹光谱和空间图像信息对粗茶进行快速质量评价的强大数据驱动技术。然而，不同等级之间的光谱重叠和图像相似度影响了传统分类方法的性能。为了解决这些挑战，本研究提出了一种新的光谱图像融合神经网络（SIFNet），该网络可以处理光谱依赖性和图像特征的微妙等级差异。SIFNet由光谱特征提取分支和图像特征提取分支组成。光谱特征提取结合一维卷积神经网络（1D-CNN）和双向长短期记忆网络（Bi-LSTM）捕获光谱依赖关系。而图像特征提取分支则采用二维CNN （2D-CNN）和模糊运算模块来捕获精确和模糊图像特征。最后，利用全连通层融合跨模态特征，提高分类性能。在武夷岩茶原始高光谱图像数据集上的5次交叉验证表明，SIFNet的准确率为96.75±2.27%，优于其他先进的机器学习和深度学习算法。本研究提出了一种数据驱动的粗茶质量智能评价方法，为提升茶产业可持续供应能力提供有力的技术支撑。

{"title":"SIFNet: A spectrum-image fusion neural network in hyperspectral imaging for intelligent quality assessment of crude tea to ensure tea product availability","authors":"Dengshan Li , Zhenhao Xu , Rongxi Lin , Yixing Huang , Qin Ouyang , Zhonghua Liu","doi":"10.1016/j.compag.2026.111441","DOIUrl":"10.1016/j.compag.2026.111441","url":null,"abstract":"<div><div>Accurately assessing the quality of crude tea (primary processed tea) is critical for ensuring the standardization and consistent supply of tea products. Hyperspectral imaging (HSI) has emerged as a powerful data-driven technology that utilizes molecular fingerprint spectra and spatial image information for rapid quality assessment of crude tea. However, spectral overlap and image similarity between different grades hinder the performance of traditional classification methods. To address these challenges, this study proposes a novel spectrum-image fusion neural network (SIFNet) that tackles spectral dependency and subtle inter-grade differences in image features. SIFNet consists of a spectral feature extraction branch and an image feature extraction branch. The spectral feature extraction combines one-dimensional convolutional neural networks (1D-CNN) and bidirectional long short-term memory networks (Bi-LSTM) to capture spectral dependencies. While the image feature extraction branch employs two-dimensional CNN (2D-CNN) and fuzzy operation modules to capture both precise and fuzzy image features. Finally, cross-modal features are fused using fully connected layers to improve classification performance. Five-fold cross-validation on the crude Wuyi rock tea hyperspectral image dataset demonstrated that SIFNet achieved an accuracy of 96.75 ± 2.27%, outperforming other state-of-the-art machine learning and deep learning algorithms. This study proposed a data-driven approach for intelligent quality assessment of crude tea, providing robust technical support for enhancing the sustainable supply capacity of the tea industry.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"243 ","pages":"Article 111441"},"PeriodicalIF":8.9,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979077","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A novel approach to monitor peanut equivalent water thickness through modular training and transfer learning of an improved PROSAIL model using a Wasserstein generative adversarial network 基于Wasserstein生成对抗网络的改进PROSAIL模型的模块化训练和迁移学习，提出了一种监测花生等效水分厚度的新方法

IF 8.9 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2026-01-14 DOI: 10.1016/j.compag.2026.111437

Shiyuan Liu , Yumeng Zhou , Weiguang Yang , Jiangtao Tan , Xi Li , Zhenhui Xiong , Zewu Fang , Hong Li , Yifei Chen , Yubin Lan , Shubo Wan , Jianguo Wang , Tingting Chen , Lei Zhang

Empirical and physical models are widely used for monitoring equivalent water thickness (EWT) to adjust plant moisture management. However, model transferability to different times and locations, and insufficient training data remain the two key challenges of field spectroscopy analysis. Therefore, this study aims to construct a hybrid model, which combines the physical models optimized by Wasserstein Generative Adversarial Nets (WGAN) and empirical models for performing hyperparameter searches (the process of finding optimal model settings) to monitor the peanut EWT. Specifically, we develop a large spectral dataset consisting of field-measured data which including 246 peanut varieties in five peanut farms across China and synthetic datasets generated from the physical models optimized by WGAN. Furthermore, the PWLEH was constructed by hyperparameter tuning and pre-training which using synthetic datasets, and then fine-tuned by modular training with field data of peanut canopy water content. Comparing the model constructed with field data (R² = 0.5618, mean squared error (MSE) = 0.0725) and PROSAIL (a widely used canopy radiative transfer model) (R² = 0.7105, MSE = 0.0473), PWLEH achieved high accuracy in predicting peanut water content (R² = 0.7650, MSE = 0.0519). Unlike pure data-driven approaches, the new hybrid model incorporated radiative transfer knowledge and obtained higher predictive performance with fewer field data. This study demonstrates the potential of applying an optimized PROSAIL, hyperparameter search and modular training to improve the accuracy and transferability of the EWT prediction model, providing a new approach for sustainable agricultural management.

经验模型和物理模型被广泛用于监测等效水厚（EWT），以调整植物的水分管理。然而，模型在不同时间和地点的可转移性以及训练数据的不足仍然是现场光谱分析的两个主要挑战。因此，本研究旨在构建一个混合模型，将Wasserstein生成对抗网络（WGAN）优化的物理模型与进行超参数搜索（寻找最优模型设置的过程）的经验模型相结合，以监测花生EWT。具体而言，我们开发了一个大型光谱数据集，包括中国五个花生农场的246个花生品种的实地测量数据和由WGAN优化的物理模型生成的合成数据集。在此基础上，利用合成数据集进行超参数整定和预训练，然后利用花生冠层含水量实测数据进行模块化训练，对PWLEH进行微调。与实测数据（R2 = 0.5618，均方误差(MSE) = 0.0725）和PROSAIL（广泛应用的冠层辐射传输模型）（R2 = 0.7105, MSE = 0.0473）相比，PWLEH对花生水分含量的预测精度较高（R2 = 0.7650, MSE = 0.0519）。与纯数据驱动的方法不同，新的混合模型结合了辐射传递知识，并以更少的现场数据获得了更高的预测性能。该研究证明了应用优化的PROSAIL、超参数搜索和模块化训练来提高EWT预测模型的准确性和可移植性的潜力，为可持续农业管理提供了新的途径。

{"title":"A novel approach to monitor peanut equivalent water thickness through modular training and transfer learning of an improved PROSAIL model using a Wasserstein generative adversarial network","authors":"Shiyuan Liu , Yumeng Zhou , Weiguang Yang , Jiangtao Tan , Xi Li , Zhenhui Xiong , Zewu Fang , Hong Li , Yifei Chen , Yubin Lan , Shubo Wan , Jianguo Wang , Tingting Chen , Lei Zhang","doi":"10.1016/j.compag.2026.111437","DOIUrl":"10.1016/j.compag.2026.111437","url":null,"abstract":"<div><div>Empirical and physical models are widely used for monitoring equivalent water thickness (EWT) to adjust plant moisture management. However, model transferability to different times and locations, and insufficient training data remain the two key challenges of field spectroscopy analysis. Therefore, this study aims to construct a hybrid model, which combines the physical models optimized by Wasserstein Generative Adversarial Nets (WGAN) and empirical models for performing hyperparameter searches (the process of finding optimal model settings) to monitor the peanut EWT. Specifically, we develop a large spectral dataset consisting of field-measured data which including 246 peanut varieties in five peanut farms across China and synthetic datasets generated from the physical models optimized by WGAN. Furthermore, the PWLEH was constructed by hyperparameter tuning and pre-training which using synthetic datasets, and then fine-tuned by modular training with field data of peanut canopy water content. Comparing the model constructed with field data (R<sup>2</sup> = 0.5618, mean squared error (MSE) = 0.0725) and PROSAIL (a widely used canopy radiative transfer model) (R<sup>2</sup> = 0.7105, MSE = 0.0473), PWLEH achieved high accuracy in predicting peanut water content (R<sup>2</sup> = 0.7650, MSE = 0.0519). Unlike pure data-driven approaches, the new hybrid model incorporated radiative transfer knowledge and obtained higher predictive performance with fewer field data. This study demonstrates the potential of applying an optimized PROSAIL, hyperparameter search and modular training to improve the accuracy and transferability of the EWT prediction model, providing a new approach for sustainable agricultural management.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"243 ","pages":"Article 111437"},"PeriodicalIF":8.9,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979075","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Precision yield estimation and mapping in manual strawberry harvesting with instrumented picking carts and a robust data processing pipeline 精确产量估计和地图在人工草莓收获与仪器采摘车和一个强大的数据处理管道

IF 8.9 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2026-01-13 DOI: 10.1016/j.compag.2025.111302

Uddhav Bhattarai , Rajkishan Arikapudi , Chen Peng , Steven A. Fennimore , Frank N. Martin , Stavros G. Vougioukas

High-resolution yield maps for manually harvested crops are impractical to generate on commercial scales because yield monitors are available only for mechanical harvesters. However, precision crop management relies on accurately determining spatial and temporal yield variability. This study presents the development of an integrated system for precision yield estimation and mapping for manually harvested strawberries. Conventional strawberry picking carts were instrumented with a Global Positioning System (GPS) receiver, an Inertial Measurement Unit (IMU), and load cells to record real-time geo-tagged harvest data and cart motion. Extensive data were collected in two strawberry fields in California, USA, during a harvest season. To address the inconsistencies and errors caused by the sensors and the manual harvesting process, a robust data processing pipeline was developed by integrating supervised deep learning model with unsupervised algorithms. The pipeline was used to estimate the yield distribution and generate yield maps for season-long harvests at the desired grid resolution. The estimated yield distributions were used to calculate two metrics: the total mass harvested over specific row segments and the total mass of trays harvested. The metrics were compared to ground truth and achieved accuracies of 90.48% and 94.05%, respectively. Additionally, the accuracy of the estimated yield based on the number of trays harvested per cart for season-long harvest was better than 94% achieving a strong correlation (Pearson r = 0.99) with the actual number of counted trays in both fields. The proposed system provides a scalable and practical solution for specialty crops, assisting in efficient yield estimation and mapping, field management, and labor management for sustainable crop production. The dataset and code supporting this study are available at: https://doi.org/10.5061/dryad.v6wwpzh7h and https://github.com/uddhavbhattarai/iCarritoYieldEstimationandMapping.git.

人工收割作物的高分辨率产量图在商业规模上是不切实际的，因为产量监测器只能用于机械收割。然而，精确的作物管理依赖于准确地确定产量的时空变化。本研究提出了一种用于人工收获草莓的精确产量估算和制图的集成系统的开发。传统的草莓采摘车配备了全球定位系统（GPS）接收器、惯性测量单元（IMU）和称重传感器，以记录实时地理标记的收获数据和推车运动。在一个收获季节，在美国加利福尼亚州的两个草莓田收集了大量数据。为了解决传感器和人工采集过程中产生的不一致和错误，将有监督深度学习模型与无监督算法相结合，开发了鲁棒的数据处理管道。该管道用于估计产量分布，并在所需的网格分辨率下生成季节性收获的产量图。估计的产量分布用于计算两个指标：特定行段收获的总质量和收获的托盘总质量。将这些指标与地面真实度进行比较，准确率分别为90.48%和94.05%。此外，基于每辆车收获的托盘数量的估计产量的准确性优于94%，与两个领域的实际计算托盘数量具有很强的相关性（Pearson r = 0.99）。该系统为特种作物提供了可扩展和实用的解决方案，有助于有效的产量估算和制图、田间管理和可持续作物生产的劳动力管理。支持本研究的数据集和代码可在https://doi.org/10.5061/dryad.v6wwpzh7h和https://github.com/uddhavbhattarai/iCarritoYieldEstimationandMapping.git上获得。

{"title":"Precision yield estimation and mapping in manual strawberry harvesting with instrumented picking carts and a robust data processing pipeline","authors":"Uddhav Bhattarai , Rajkishan Arikapudi , Chen Peng , Steven A. Fennimore , Frank N. Martin , Stavros G. Vougioukas","doi":"10.1016/j.compag.2025.111302","DOIUrl":"10.1016/j.compag.2025.111302","url":null,"abstract":"<div><div>High-resolution yield maps for manually harvested crops are impractical to generate on commercial scales because yield monitors are available only for mechanical harvesters. However, precision crop management relies on accurately determining spatial and temporal yield variability. This study presents the development of an integrated system for precision yield estimation and mapping for manually harvested strawberries. Conventional strawberry picking carts were instrumented with a Global Positioning System (GPS) receiver, an Inertial Measurement Unit (IMU), and load cells to record real-time geo-tagged harvest data and cart motion. Extensive data were collected in two strawberry fields in California, USA, during a harvest season. To address the inconsistencies and errors caused by the sensors and the manual harvesting process, a robust data processing pipeline was developed by integrating supervised deep learning model with unsupervised algorithms. The pipeline was used to estimate the yield distribution and generate yield maps for season-long harvests at the desired grid resolution. The estimated yield distributions were used to calculate two metrics: the total mass harvested over specific row segments and the total mass of trays harvested. The metrics were compared to ground truth and achieved accuracies of 90.48% and 94.05%, respectively. Additionally, the accuracy of the estimated yield based on the number of trays harvested per cart for season-long harvest was better than 94% achieving a strong correlation (Pearson r = 0.99) with the actual number of counted trays in both fields. The proposed system provides a scalable and practical solution for specialty crops, assisting in efficient yield estimation and mapping, field management, and labor management for sustainable crop production. The dataset and code supporting this study are available at: <span><span>https://doi.org/10.5061/dryad.v6wwpzh7h</span><svg><path></path></svg></span> and <span><span>https://github.com/uddhavbhattarai/iCarritoYieldEstimationandMapping.git</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"243 ","pages":"Article 111302"},"PeriodicalIF":8.9,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145978374","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Explainability and privacy in AI-enabled crop monitoring: Trends and future directions in soybean research 人工智能作物监测中的可解释性和隐私性：大豆研究的趋势和未来方向

IF 8.9 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2026-01-13 DOI: 10.1016/j.compag.2025.111392

Jayme Garcia Arnal Barbedo , Marcelo Santos da Silva , Mirela Teixeira Cazzolato , Lucas Pascotti Valem , Renato Tinós , Roseli Aparecida Francelin Romero , Luiz Otavio Murta Junior , Adriano de Jesus Holanda , Joaquim Cezar Felipe , José Baldin Pinheiro , José Tiago Barroso Chagas , Roberto Fray da Silva , Everton Castelão Tetila , Lucio Andre de Castro Jorge , Huaqiang Yuan , Weiling Li , Ketan Kotecha , Liang Zhao

AI is playing an increasingly central role in crop monitoring, driven by rapid advances in deep learning that now tackle recognition and prediction tasks once out of reach. However, translating these gains into soybean production is increasingly constrained by two intertwined requirements, explainability (to support expert scrutiny and responsible use of black-box models) and privacy (to protect sensitive farm data and enable collaboration across stakeholders). This review synthesizes recent advances in interpretable and privacy preserving machine learning, emphasizing soybean related applications where empirical evidence is solid, and covering both post hoc and inherently interpretable approaches alongside privacy mechanisms such as federated learning with secure aggregation and differential privacy. Across the literature, recurring deployment barriers are identified, most notably variability across farms and seasons, the need for explanations that remain meaningful both locally and globally, infrastructure limitations in rural settings, risks of information leakage through explanations, and the scarcity of multi-season validation under real-world conditions. These findings suggest that field-ready soybean monitoring systems should be designed with explainability and privacy as major goals, rather than add-ons, and evaluated under realistic variability and governance requirements. The ultimate goal is to help bridge the gap between academic innovation and practical, deployable solutions that protect farmer data while supporting decision-making where it matters most.

在深度学习快速发展的推动下，人工智能在作物监测中发挥着越来越重要的作用，深度学习现在可以解决曾经遥不可及的识别和预测任务。然而，将这些成果转化为大豆生产越来越受到两个相互交织的要求的制约，即可解释性（支持专家审查和负责任地使用黑盒模型）和隐私性（保护敏感的农场数据并使利益相关者之间能够合作）。本综述综合了可解释和保护隐私的机器学习的最新进展，强调了经验证据确凿的大豆相关应用，并涵盖了事后和内在可解释的方法以及隐私机制，如具有安全聚合和差分隐私的联邦学习。在文献中，发现了反复出现的部署障碍，最明显的是农场和季节之间的差异，对本地和全球都有意义的解释的需求，农村环境中的基础设施限制，解释带来的信息泄露风险，以及现实世界条件下多季节验证的稀缺性。这些发现表明，田间大豆监测系统的设计应以可解释性和隐私性为主要目标，而不是附加目标，并在现实的可变性和治理要求下进行评估。最终目标是帮助弥合学术创新与实际、可部署的解决方案之间的差距，以保护农民数据，同时支持最重要的决策。

{"title":"Explainability and privacy in AI-enabled crop monitoring: Trends and future directions in soybean research","authors":"Jayme Garcia Arnal Barbedo , Marcelo Santos da Silva , Mirela Teixeira Cazzolato , Lucas Pascotti Valem , Renato Tinós , Roseli Aparecida Francelin Romero , Luiz Otavio Murta Junior , Adriano de Jesus Holanda , Joaquim Cezar Felipe , José Baldin Pinheiro , José Tiago Barroso Chagas , Roberto Fray da Silva , Everton Castelão Tetila , Lucio Andre de Castro Jorge , Huaqiang Yuan , Weiling Li , Ketan Kotecha , Liang Zhao","doi":"10.1016/j.compag.2025.111392","DOIUrl":"10.1016/j.compag.2025.111392","url":null,"abstract":"<div><div>AI is playing an increasingly central role in crop monitoring, driven by rapid advances in deep learning that now tackle recognition and prediction tasks once out of reach. However, translating these gains into soybean production is increasingly constrained by two intertwined requirements, explainability (to support expert scrutiny and responsible use of black-box models) and privacy (to protect sensitive farm data and enable collaboration across stakeholders). This review synthesizes recent advances in interpretable and privacy preserving machine learning, emphasizing soybean related applications where empirical evidence is solid, and covering both post hoc and inherently interpretable approaches alongside privacy mechanisms such as federated learning with secure aggregation and differential privacy. Across the literature, recurring deployment barriers are identified, most notably variability across farms and seasons, the need for explanations that remain meaningful both locally and globally, infrastructure limitations in rural settings, risks of information leakage through explanations, and the scarcity of multi-season validation under real-world conditions. These findings suggest that field-ready soybean monitoring systems should be designed with explainability and privacy as major goals, rather than add-ons, and evaluated under realistic variability and governance requirements. The ultimate goal is to help bridge the gap between academic innovation and practical, deployable solutions that protect farmer data while supporting decision-making where it matters most.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"243 ","pages":"Article 111392"},"PeriodicalIF":8.9,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979073","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0