Computers and Electronics in Agriculture最新文献_第10页

Classifying grain and impurity to assess maize cleaning loss using time–frequency images of vibro-piezoelectric signals coupling machine learning 利用振动压电信号的时频图像耦合机器学习对谷物和杂质进行分类，以评估玉米的清洁损失

IF 7.7 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2024-11-06 DOI: 10.1016/j.compag.2024.109583

Yibo Li , Yuxin Hou , Tao Cui , Danielle S Tan , Yang Xu , Dongxing Zhang , Mengmeng Qiao , Lijian Xiong

Accurately differentiating maize mixtures and assessing grain cleaning loss contributes to improving the efficiency and sustainability of agricultural systems. This study proposes a novel detection method integrating time–frequency images of particle vibro-piezoelectric signals and machine learning to classify grain and impurity and assess maize cleaning loss. Specifically, an indie-developed vibro-piezoelectric detection setup is employed to capture the time-domain response signals of grain and impurity for building a database of maize collision signals. Using the Short-Time Fourier Transform (STFT) and Weighted Average Algorithm (WAA), 1D time-domain signals characterizing only the time-varying properties are converted into 2D time–frequency images possessing rich spectral feature information and energy distribution. Subsequently, 15 texture features are extracted from 2D time–frequency images with the Grey-Level-Gradient Co-ccurrence Matrix (GLGCM). After eliminating weakly-correlated features, eleven texture features are chosen and consolidated within the first four Principal Components (PCs). These four PCs and the traditional 1D time-domain signals are pre-processed and input into the Naive Bayes (NB), Support Vector Machine (SVM), Decision Tree (DT), and Random Forest (RF) classifiers. The NB model with Savitzky-Golay (SG) pre-processing utilizing 2D time–frequency image features exhibits the highest accuracy of 95.74%, surpassing the optimal 1D time-domain classification model by 5.31 percentage points. Bench tests verify that the piezoelectric detection unit with the optimal NB model can control the absolute error in grain loss rate to within 0.43%. Notably, the proposed method also applies to the classification and cleaning loss detection of other typical crops by replacing the collision signal database.

准确区分玉米混合物和评估谷物清洗损失有助于提高农业系统的效率和可持续性。本研究提出了一种新型检测方法，将颗粒振动压电信号的时频图像与机器学习相结合，对谷物和杂质进行分类，并评估玉米的清洁损失。具体而言，利用自主研发的振动压电检测装置捕捉谷物和杂质的时域响应信号，建立玉米碰撞信号数据库。利用短时傅里叶变换 (STFT) 和加权平均算法 (WAA)，将仅描述时变特性的一维时域信号转换为具有丰富频谱特征信息和能量分布的二维时频图像。随后，利用灰色-梯度共生矩阵（GLGCM）从二维时频图像中提取 15 个纹理特征。剔除弱相关特征后，选出 11 个纹理特征，并将其合并到前四个主成分（PC）中。这四个 PC 和传统的一维时域信号经过预处理后，输入到 Naive Bayes（NB）、支持向量机（SVM）、决策树（DT）和随机森林（RF）分类器中。利用二维时频图像特征进行萨维茨基-戈莱（SG）预处理的 NB 模型准确率最高，达到 95.74%，比最佳一维时域分类模型高出 5.31 个百分点。工作台测试证实，采用最佳 NB 模型的压电检测单元可将晶粒损耗率的绝对误差控制在 0.43% 以内。值得注意的是，通过替换碰撞信号数据库，所提出的方法也适用于其他典型作物的分类和清选损失检测。

{"title":"Classifying grain and impurity to assess maize cleaning loss using time–frequency images of vibro-piezoelectric signals coupling machine learning","authors":"Yibo Li , Yuxin Hou , Tao Cui , Danielle S Tan , Yang Xu , Dongxing Zhang , Mengmeng Qiao , Lijian Xiong","doi":"10.1016/j.compag.2024.109583","DOIUrl":"10.1016/j.compag.2024.109583","url":null,"abstract":"<div><div>Accurately differentiating maize mixtures and assessing grain cleaning loss contributes to improving the efficiency and sustainability of agricultural systems. This study proposes a novel detection method integrating time–frequency images of particle vibro-piezoelectric signals and machine learning to classify grain and impurity and assess maize cleaning loss. Specifically, an indie-developed vibro-piezoelectric detection setup is employed to capture the time-domain response signals of grain and impurity for building a database of maize collision signals. Using the Short-Time Fourier Transform (STFT) and Weighted Average Algorithm (WAA), 1D time-domain signals characterizing only the time-varying properties are converted into 2D time–frequency images possessing rich spectral feature information and energy distribution. Subsequently, 15 texture features are extracted from 2D time–frequency images with the Grey-Level-Gradient Co-ccurrence Matrix (GLGCM). After eliminating weakly-correlated features, eleven texture features are chosen and consolidated within the first four Principal Components (PCs). These four PCs and the traditional 1D time-domain signals are pre-processed and input into the Naive Bayes (NB), Support Vector Machine (SVM), Decision Tree (DT), and Random Forest (RF) classifiers. The NB model with Savitzky-Golay (SG) pre-processing utilizing 2D time–frequency image features exhibits the highest accuracy of 95.74%, surpassing the optimal 1D time-domain classification model by 5.31 percentage points. Bench tests verify that the piezoelectric detection unit with the optimal NB model can control the absolute error in grain loss rate to within 0.43%. Notably, the proposed method also applies to the classification and cleaning loss detection of other typical crops by replacing the collision signal database.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"227 ","pages":"Article 109583"},"PeriodicalIF":7.7,"publicationDate":"2024-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142593816","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

Autonomous net inspection and cleaning in sea-based fish farms: A review 海基养鱼场的自主鱼网检查和清洁：综述

IF 7.7 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2024-11-06 DOI: 10.1016/j.compag.2024.109609

Jiaying Fu, Da Liu, Yingchao He, Fang Cheng

In sea-based fish farms, biofouling and net damage are unavoidable challenges. To ensure safe, reliable, and sustainable fish production, timely monitoring of nets is crucial for detecting biofouling and net damage, along with providing decision support for subsequent maintenance and cleaning. In recent years, technological advancements have driven the automation of production processes, with a growing trend toward using robots instead of human labor for net operations in sea-based fish farms. However, there is a lack of a systematic review of autonomous net inspection and cleaning. This paper addresses this gap by reviewing and analyzing the current state of autonomous net inspection and cleaning in sea-based fish farms. Key technologies, including robot control, net inspection, and net cleaning, are summarized, along with their future development in practical applications. This paper also emphasizes Industry 4.0 technologies that support these advancements, such as sensors, robotics, artificial intelligence (AI), the Internet of Things (IoT), big data analytics, and the digital twin (DT). Furthermore, advanced robotic solutions currently used for autonomous net inspection and cleaning, as well as their potential benefits and drawbacks, are presented. Finally, the challenges and future research directions are highlighted, offering valuable insights for institutions and companies working to enhance the autonomy and intelligence of net operations in sea-based fish farms.

在海基养鱼场，生物污损和鱼网损坏是不可避免的挑战。为确保安全、可靠和可持续的鱼类生产，及时监测渔网对于检测生物污损和渔网损坏以及为后续维护和清洁提供决策支持至关重要。近年来，技术进步推动了生产流程的自动化，在海基养鱼场中使用机器人代替人工进行网箱作业的趋势日益明显。然而，目前还缺乏对自主鱼网检查和清洁的系统回顾。本文针对这一空白，回顾并分析了海上养鱼场自主网具检查和清洁的现状。本文总结了包括机器人控制、鱼网检查和鱼网清洁在内的关键技术，以及这些技术在实际应用中的未来发展。本文还强调了支持这些进步的工业 4.0 技术，如传感器、机器人技术、人工智能 (AI)、物联网 (IoT)、大数据分析和数字孪生 (DT)。此外，还介绍了目前用于自主网络检测和清洁的先进机器人解决方案及其潜在的优点和缺点。最后，重点介绍了面临的挑战和未来的研究方向，为致力于提高海基养鱼场网作业的自主性和智能性的机构和公司提供了宝贵的见解。

{"title":"Autonomous net inspection and cleaning in sea-based fish farms: A review","authors":"Jiaying Fu, Da Liu, Yingchao He, Fang Cheng","doi":"10.1016/j.compag.2024.109609","DOIUrl":"10.1016/j.compag.2024.109609","url":null,"abstract":"<div><div>In sea-based fish farms, biofouling and net damage are unavoidable challenges. To ensure safe, reliable, and sustainable fish production, timely monitoring of nets is crucial for detecting biofouling and net damage, along with providing decision support for subsequent maintenance and cleaning. In recent years, technological advancements have driven the automation of production processes, with a growing trend toward using robots instead of human labor for net operations in sea-based fish farms. However, there is a lack of a systematic review of autonomous net inspection and cleaning. This paper addresses this gap by reviewing and analyzing the current state of autonomous net inspection and cleaning in sea-based fish farms. Key technologies, including robot control, net inspection, and net cleaning, are summarized, along with their future development in practical applications. This paper also emphasizes Industry 4.0 technologies that support these advancements, such as sensors, robotics, artificial intelligence (AI), the Internet of Things (IoT), big data analytics, and the digital twin (DT). Furthermore, advanced robotic solutions currently used for autonomous net inspection and cleaning, as well as their potential benefits and drawbacks, are presented. Finally, the challenges and future research directions are highlighted, offering valuable insights for institutions and companies working to enhance the autonomy and intelligence of net operations in sea-based fish farms.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"227 ","pages":"Article 109609"},"PeriodicalIF":7.7,"publicationDate":"2024-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142592560","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

WE-DeepLabV3+: A lightweight segmentation model for Panax notoginseng leaf diseases WE-DeepLabV3+：三七叶病轻量级分割模型

IF 7.7 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2024-11-05 DOI: 10.1016/j.compag.2024.109612

Zilong Wang , Ling Yang , Ruoxi Wang , Lian Lei , Hao Ding , Qiliang Yang

Panax notoginseng plays an important role in traditional Chinese medicine. However, diseases pose a significant threat to the quality and yield of P. notoginseng. The main challenge related to the identification of P. notoginseng leaf diseases is how to achieve good performance in the case of small diseased spots on P. notoginseng leaf, overlapping edges of diseased leaf and the difficulty in mobile deployment. A lightweight semantic segmentation model Window Efficient-DeepLabv3+ is proposed for segmentation and quantification of P. notoginseng leaf diseases. We propose the Window Attention-ASPP module and present a hierarchical stacking of features, which improves model accuracy for minor target lesions while reducing parameters. In addition, a lightweight backbone network MobileNetV2 is utilized as a feature extraction module. The decoding stage introduces the Efficient Channel Attention module, which effectively improves the accuracy of the segmentation of the blade contour. Experimental results yielded the Mean Intersection Over Union, Mean Precision, and Mean Recall metrics of WE-DeepLabV3+ network to be 82.0 %, 87.6 %, and 92.4 % respectively, outperforming other segmentation models such as UNet, PSPNet, CaraNet, SegNet, and BiSeNetV2. Moreover, the number of parameters has been reduced by 90.6 % with only 5.1 M parameters. Finally, the method is used to quantify the disease of P. notoginseng leaf, the error is only 1.15 % and 0.82 %, which proves that it can quantify the disease severity accurately. Thus, the proposed method holds great significance for raising the yield and quality of P. notoginseng, also providing reliable guidance for precise fertilization and drug control.

三七在传统中药中发挥着重要作用。然而，病害对三七的质量和产量构成了重大威胁。三七叶片病害识别的主要挑战是如何在三七叶片病斑较小、病叶边缘重叠以及移动部署困难的情况下实现良好的性能。我们提出了一种轻量级语义分割模型 Window Efficient-DeepLabv3+ 用于分割和量化田七叶片病害。我们提出了 Window Attention-ASPP 模块，并对特征进行了分层堆叠，在减少参数的同时提高了模型对轻微目标病变的准确性。此外，我们还利用轻量级骨干网络 MobileNetV2 作为特征提取模块。解码阶段引入了高效通道关注模块，有效提高了叶片轮廓分割的准确性。实验结果表明，WE-DeepLabV3+ 网络的平均联合交叉率（Mean Intersection Over Union）、平均精确率（Mean Precision）和平均召回率（Mean Recall）指标分别为 82.0%、87.6% 和 92.4%，优于 UNet、PSPNet、CaraNet、SegNet 和 BiSeNetV2 等其他分割模型。此外，参数数量减少了 90.6%，只有 5.1 M 个参数。最后，将该方法用于量化田七叶片的病害，误差仅为 1.15 % 和 0.82 %，证明该方法能准确量化病害严重程度。因此，该方法对提高田七的产量和质量具有重要意义，同时也为精确施肥和药物控制提供了可靠的指导。

{"title":"WE-DeepLabV3+: A lightweight segmentation model for Panax notoginseng leaf diseases","authors":"Zilong Wang , Ling Yang , Ruoxi Wang , Lian Lei , Hao Ding , Qiliang Yang","doi":"10.1016/j.compag.2024.109612","DOIUrl":"10.1016/j.compag.2024.109612","url":null,"abstract":"<div><div><em>Panax notoginseng</em> plays an important role in traditional Chinese medicine. However, diseases pose a significant threat to the quality and yield of <em>P. notoginseng</em>. The main challenge related to the identification of <em>P. notoginseng</em> leaf diseases is how to achieve good performance in the case of small diseased spots on <em>P. notoginseng</em> leaf, overlapping edges of diseased leaf and the difficulty in mobile deployment. A lightweight semantic segmentation model Window Efficient-DeepLabv3+ is proposed for segmentation and quantification of <em>P. notoginseng</em> leaf diseases. We propose the Window Attention-ASPP module and present a hierarchical stacking of features, which improves model accuracy for minor target lesions while reducing parameters. In addition, a lightweight backbone network MobileNetV2 is utilized as a feature extraction module. The decoding stage introduces the Efficient Channel Attention module, which effectively improves the accuracy of the segmentation of the blade contour. Experimental results yielded the Mean Intersection Over Union, Mean Precision, and Mean Recall metrics of WE-DeepLabV3+ network to be 82.0 %, 87.6 %, and 92.4 % respectively, outperforming other segmentation models such as UNet, PSPNet, CaraNet, SegNet, and BiSeNetV2. Moreover, the number of parameters has been reduced by 90.6 % with only 5.1 M parameters. Finally, the method is used to quantify the disease of <em>P. notoginseng</em> leaf, the error is only 1.15 % and 0.82 %, which proves that it can quantify the disease severity accurately. Thus, the proposed method holds great significance for raising the yield and quality of <em>P. notoginseng</em>, also providing reliable guidance for precise fertilization and drug control.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"227 ","pages":"Article 109612"},"PeriodicalIF":7.7,"publicationDate":"2024-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142587287","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

Data value creation in agriculture: A review 农业数据价值创造：综述

IF 7.7 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2024-11-05 DOI: 10.1016/j.compag.2024.109602

Havva Uyar , Ioannis Karvelas , Stamatia Rizou , Spyros Fountas

Agricultural data have great potential to improve decision-making, enhance operational efficiency, and drive innovation. Despite the growing acknowledgment of their value, there remains a gap in understanding how data value creation is perceived and implemented in agriculture. This study addresses this gap by investigating data value creation mechanisms, targets, and impacts through a structured literature review of 80 articles, including 13 core articles retrieved via targeted database searches and 67 additional articles identified through cross-reference snowballing. Key “value creation mechanisms” are categorized as transparency and access, discovery and experimentation, prediction and optimization, customization and targeting, learning and crowdsourcing, and monitoring and adaptation. The value creation mechanisms aim to enhance key “targets”, namely organizational performance, business process improvement, product and service innovation, and consumer and market experience. Organization performance was the most frequently addressed value target, appearing in approximately 85% of the core articles, followed by business process improvement, highlighted in approximately 77% of the articles. Together, the mechanisms and targets create “impact”, constructing the value of data. The findings reveal that all core articles (100%) emphasize the functional value of agricultural data, while 54% also explore their symbolic value, which enhances reputation and market positioning. A key takeaway is that, unlike many other assets, the value of agricultural data increases with reuse, which calls for a shift in focus from data ownership to ownership of the value derived from them. This study highlights the need for robust frameworks to fully realize the potential of agricultural data and calls for future research to further characterize and assess this value. These insights are essential for developing tools and methodologies that enhance productivity, sustainability, and profitability in agriculture.

农业数据在改善决策、提高运营效率和推动创新方面具有巨大潜力。尽管人们日益认识到数据的价值，但在了解农业领域如何看待和实施数据价值创造方面仍存在差距。本研究针对这一空白，通过对 80 篇文章进行结构化文献综述，调查数据价值创造机制、目标和影响，其中包括通过定向数据库搜索检索到的 13 篇核心文章，以及通过交叉引用滚雪球法确定的 67 篇其他文章。主要的 "价值创造机制 "分为透明度与获取、发现与实验、预测与优化、定制与目标定位、学习与众包，以及监测与适应。价值创造机制旨在提高关键 "目标"，即组织绩效、业务流程改进、产品和服务创新以及消费者和市场体验。组织绩效是最常涉及的价值目标，出现在约 85% 的核心文章中，其次是业务流程改进，在约 77% 的文章中得到强调。这些机制和目标共同产生了 "影响"，构建了数据的价值。研究结果显示，所有核心文章（100%）都强调了农业数据的功能价值，54%的文章还探讨了其象征价值，即提高声誉和市场定位。一项重要启示是，与许多其他资产不同，农业数据的价值会随着重复使用而增加，这就要求将重点从数据所有权转移到数据衍生价值的所有权上。这项研究强调，要充分发挥农业数据的潜力，就必须建立健全的框架，并呼吁今后开展研究，进一步描述和评估这种价值。这些见解对于开发提高农业生产力、可持续性和盈利能力的工具和方法至关重要。

{"title":"Data value creation in agriculture: A review","authors":"Havva Uyar , Ioannis Karvelas , Stamatia Rizou , Spyros Fountas","doi":"10.1016/j.compag.2024.109602","DOIUrl":"10.1016/j.compag.2024.109602","url":null,"abstract":"<div><div>Agricultural data have great potential to improve decision-making, enhance operational efficiency, and drive innovation. Despite the growing acknowledgment of their value, there remains a gap in understanding how data value creation is perceived and implemented in agriculture. This study addresses this gap by investigating data value creation mechanisms, targets, and impacts through a structured literature review of 80 articles, including 13 core articles retrieved via targeted database searches and 67 additional articles identified through cross-reference snowballing. Key “value creation mechanisms” are categorized as transparency and access, discovery and experimentation, prediction and optimization, customization and targeting, learning and crowdsourcing, and monitoring and adaptation. The value creation mechanisms aim to enhance key “targets”, namely organizational performance, business process improvement, product and service innovation, and consumer and market experience. Organization performance was the most frequently addressed value target, appearing in approximately 85% of the core articles, followed by business process improvement, highlighted in approximately 77% of the articles. Together, the mechanisms and targets create “impact”, constructing the value of data. The findings reveal that all core articles (100%) emphasize the functional value of agricultural data, while 54% also explore their symbolic value, which enhances reputation and market positioning. A key takeaway is that, unlike many other assets, the value of agricultural data increases with reuse, which calls for a shift in focus from data ownership to ownership of the value derived from them. This study highlights the need for robust frameworks to fully realize the potential of agricultural data and calls for future research to further characterize and assess this value. These insights are essential for developing tools and methodologies that enhance productivity, sustainability, and profitability in agriculture.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"227 ","pages":"Article 109602"},"PeriodicalIF":7.7,"publicationDate":"2024-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142587288","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Efficient three-dimensional reconstruction and skeleton extraction for intelligent pruning of fruit trees 用于果树智能修剪的高效三维重建和骨架提取技术

IF 7.7 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2024-11-05 DOI: 10.1016/j.compag.2024.109554

Xiaojuan Li , Bo Liu , Yinggang Shi , Mingming Xiong , Dongyu Ren , Letian Wu , Xiangjun Zou

The three-dimensional reconstruction of fruit trees plays a crucial role in assessing their growth status, analyzing agronomic traits, and categorizing their organs. This is vital for implementing intelligent orchard management. This study aims to develop a cost-effective and efficient method for the three-dimensional reconstruction and skeleton extraction of fruit trees. The proposed method leverages the 3D geometric structure captured by Time-of-Flight (TOF) sensors and addresses common issues such as occlusion and perspective ambiguity. Firstly, the TOF sensor and its supporting components are used to build an acquisition platform to collect the full range point cloud of fruit trees in the key growth period. The noise information is filtered through the point cloud preprocessing operation to obtain the complete target point cloud and extract its structural invariant features. The IWOA-RANSAC-NDT algorithm is introduced for 3D model registration. Secondly, the Delaunay triangulation algorithm and Dijkstra shortest path algorithm are used to calculate the Minimum Spanning Tree. Branch segmentation is expedited using the Kd-tree data structure. The Levenberg Marquardt algorithm and the cylindrical fitting method are used to obtain the full fruit tree skeleton model. Finally, taking walnut tree as the experimental object, a high-precision fruit tree point cloud model is constructed, and the actual verification is carried out based on the measured data. Findings indicate that the proposed methodology can accurately construct both 3D point cloud and skeleton models of fruit trees with accuracy deviations from the measured data remaining within 7 %. The proposed method offers valuable data and technical support for the future development of highly autonomous, practical, and user-oriented fruit tree pruning systems.

果树的三维重建在评估果树生长状况、分析农艺性状以及对果树器官进行分类方面发挥着至关重要的作用。这对于实施果园智能化管理至关重要。本研究旨在开发一种经济高效的果树三维重建和骨架提取方法。所提出的方法利用了飞行时间（TOF）传感器捕捉到的三维几何结构，并解决了遮挡和透视模糊等常见问题。首先，利用 TOF 传感器及其配套组件搭建采集平台，采集果树生长关键期的全范围点云。通过点云预处理操作过滤噪声信息，获得完整的目标点云，并提取其结构不变特征。引入 IWOA-RANSAC-NDT 算法进行三维模型配准。其次，利用 Delaunay 三角测量算法和 Dijkstra 最短路径算法计算最小生成树。利用 Kd 树数据结构加快了分支分割。使用 Levenberg Marquardt 算法和圆柱拟合方法获得完整的果树骨架模型。最后，以核桃树为实验对象，构建了高精度果树点云模型，并根据测量数据进行了实际验证。研究结果表明，所提出的方法可以精确地构建果树的三维点云和骨架模型，与测量数据的精度偏差保持在 7%以内。所提出的方法为未来开发高度自主、实用和面向用户的果树修剪系统提供了宝贵的数据和技术支持。

{"title":"Efficient three-dimensional reconstruction and skeleton extraction for intelligent pruning of fruit trees","authors":"Xiaojuan Li , Bo Liu , Yinggang Shi , Mingming Xiong , Dongyu Ren , Letian Wu , Xiangjun Zou","doi":"10.1016/j.compag.2024.109554","DOIUrl":"10.1016/j.compag.2024.109554","url":null,"abstract":"<div><div>The three-dimensional reconstruction of fruit trees plays a crucial role in assessing their growth status, analyzing agronomic traits, and categorizing their organs. This is vital for implementing intelligent orchard management. This study aims to develop a cost-effective and efficient method for the three-dimensional reconstruction and skeleton extraction of fruit trees. The proposed method leverages the 3D geometric structure captured by Time-of-Flight (TOF) sensors and addresses common issues such as occlusion and perspective ambiguity. Firstly, the TOF sensor and its supporting components are used to build an acquisition platform to collect the full range point cloud of fruit trees in the key growth period. The noise information is filtered through the point cloud preprocessing operation to obtain the complete target point cloud and extract its structural invariant features. The IWOA-RANSAC-NDT algorithm is introduced for 3D model registration. Secondly, the Delaunay triangulation algorithm and Dijkstra shortest path algorithm are used to calculate the Minimum Spanning Tree. Branch segmentation is expedited using the Kd-tree data structure. The Levenberg Marquardt algorithm and the cylindrical fitting method are used to obtain the full fruit tree skeleton model. Finally, taking walnut tree as the experimental object, a high-precision fruit tree point cloud model is constructed, and the actual verification is carried out based on the measured data. Findings indicate that the proposed methodology can accurately construct both 3D point cloud and skeleton models of fruit trees with accuracy deviations from the measured data remaining within 7 %. The proposed method offers valuable data and technical support for the future development of highly autonomous, practical, and user-oriented fruit tree pruning systems.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"227 ","pages":"Article 109554"},"PeriodicalIF":7.7,"publicationDate":"2024-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142587286","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

Multisite evaluation of microtensiometer and osmotic cell stem water potential sensors in almond orchards 杏园微张力计和渗透细胞茎水势传感器的多点评估

IF 7.7 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2024-11-05 DOI: 10.1016/j.compag.2024.109547

Isaya Kisekka , Srinivasa Rao Peddinti , Peter Savchik , Liyuan Yang , Mae Culumber , Khalid Bali , Luke Milliron , Erica Edwards , Mallika Nocco , Clarissa A. Reyes , Robert J. Mahoney , Kenneth Shackel , Allan Fulton

In the face of climate change, optimization of almond irrigation management is critical for ensuring the long-term sustainability of nut production and water resources. To achieve optimal irrigation management, continuous monitoring of the plant water status is critical in scheduling irrigation. It is a widely accepted practice to use stem water potential (SWP) as a measure of plant water status in woody perennials like almonds. However, the pressure chamber (PC) commonly used to make these measurements is labor-intensive and does not provide continuous data without significant additional labor. In this study, we evaluated two recently developed stem water potential sensors (Microtensiometer [MT], and Osmotic Cell [OC]), both of which can measure the SWP nearly continuously when embedded in stem sapwood tissue (typically in the trunk or branch of a tree). SWP sensors were evaluated in nine commercial almond orchards in the Central Valley of California. The SWP values obtained from both sensors were compared to the values measured using a PC using statistical software called FITEVAL. Overall, sensor performance varied from good to acceptable and from acceptable to unacceptable for MT and OC sensors respectively. The MT sensors demonstrated higher accuracy with a Nash-Sutcliff Coefficient of Efficiency (NSE) of 0.84 (95 % CI: 0.78–0.88) and a Root Mean Square Error (RMSE) of −0.24 MPa (95 % CI: −0.21 to −0.28 MPa), while the OC sensor had an NSE of 0.68 (95 % CI: 0.61–0.74) and an RMSE of −0.32 MPa (95 % CI: −0.29 to −0.35 MPa). MT sensors exhibited the added advantage of providing sub-hourly data and displaying tree recovery from water stress following irrigation, positioning them as potentially superior for precision almond orchard water management. If widely adopted, SWP sensors have the potential to optimize water use in almond production.

面对气候变化，优化杏仁灌溉管理对于确保坚果生产和水资源的长期可持续性至关重要。要实现优化灌溉管理，持续监测植物水分状况对安排灌溉至关重要。使用茎干水势（SWP）来衡量杏仁等多年生木本植物的水分状况是一种广为接受的做法。然而，通常用来进行这些测量的压力室（PC）是劳动密集型的，如果不付出大量额外的劳动，就无法提供连续的数据。在这项研究中，我们评估了最近开发的两种茎干水势传感器（微张力计 [MT] 和渗透细胞 [OC]），这两种传感器嵌入茎干边材组织（通常在树干或树枝中）后几乎可以连续测量 SWP。在加利福尼亚中央谷地的九个商业杏仁园中对 SWP 传感器进行了评估。使用名为 FITEVAL 的统计软件将两种传感器获得的 SWP 值与 PC 测量值进行了比较。总体而言，MT 和 OC 传感器的性能分别从良好到可接受以及从可接受到不可接受。MT 传感器的精度更高，其纳什-苏特克利夫效率系数 (NSE) 为 0.84（95 % CI：0.78-0.88），均方根误差 (RMSE) 为 -0.24 兆帕（95 % CI：-0.21 至 -0.28 兆帕），而 OC 传感器的 NSE 为 0.68（95 % CI：0.61-0.74），RMSE 为 -0.32 兆帕（95 % CI：-0.29 至 -0.35 兆帕）。MT 传感器具有提供亚小时数据和显示灌溉后树木从水胁迫中恢复的额外优势，因此在杏园精确水分管理方面具有潜在优势。如果得到广泛应用，SWP 传感器有可能优化杏仁生产中的用水。

{"title":"Multisite evaluation of microtensiometer and osmotic cell stem water potential sensors in almond orchards","authors":"Isaya Kisekka , Srinivasa Rao Peddinti , Peter Savchik , Liyuan Yang , Mae Culumber , Khalid Bali , Luke Milliron , Erica Edwards , Mallika Nocco , Clarissa A. Reyes , Robert J. Mahoney , Kenneth Shackel , Allan Fulton","doi":"10.1016/j.compag.2024.109547","DOIUrl":"10.1016/j.compag.2024.109547","url":null,"abstract":"<div><div>In the face of climate change, optimization of almond irrigation management is critical for ensuring the long-term sustainability of nut production and water resources. To achieve optimal irrigation management, continuous monitoring of the plant water status is critical in scheduling irrigation. It is a widely accepted practice to use stem water potential (SWP) as a measure of plant water status in woody perennials like almonds. However, the pressure chamber (PC) commonly used to make these measurements is labor-intensive and does not provide continuous data without significant additional labor. In this study, we evaluated two recently developed stem water potential sensors (Microtensiometer [MT], and Osmotic Cell [OC]), both of which can measure the SWP nearly continuously when embedded in stem sapwood tissue (typically in the trunk or branch of a tree). SWP sensors were evaluated in nine commercial almond orchards in the Central Valley of California. The SWP values obtained from both sensors were compared to the values measured using a PC using statistical software called FITEVAL. Overall, sensor performance varied from good to acceptable and from acceptable to unacceptable for MT and OC sensors respectively. The MT sensors demonstrated higher accuracy with a Nash-Sutcliff Coefficient of Efficiency (NSE) of 0.84 (95 % CI: 0.78–0.88) and a Root Mean Square Error (RMSE) of −0.24 MPa (95 % CI: −0.21 to −0.28 MPa), while the OC sensor had an NSE of 0.68 (95 % CI: 0.61–0.74) and an RMSE of −0.32 MPa (95 % CI: −0.29 to −0.35 MPa). MT sensors exhibited the added advantage of providing sub-hourly data and displaying tree recovery from water stress following irrigation, positioning them as potentially superior for precision almond orchard water management. If widely adopted, SWP sensors have the potential to optimize water use in almond production.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"227 ","pages":"Article 109547"},"PeriodicalIF":7.7,"publicationDate":"2024-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142593815","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Review of weed recognition: A global agriculture perspective 杂草识别回顾：全球农业视角

IF 7.7 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2024-11-04 DOI: 10.1016/j.compag.2024.109499

Madeleine Darbyshire , Shaun Coutts , Petra Bosilj , Elizabeth Sklar , Simon Parsons

Recent years have seen the emergence of various precision weed management technologies in both research and commercial contexts. These technologies better target weed management interventions to provide weed control that is more efficient and environmentally friendly. To support this effort, a significant amount of research has focused on machine vision to recognize weeds in a variety of crops. In this work, we systematically survey recent literature on weed recognition in crops and evaluate its relevance based on the status of global agriculture as presented in FAO statistics. Our findings indicate a notable emphasis on crops like sugar beet, carrot, and maize, while wheat and rice, despite their substantial contribution to global cropland and food supply, are relatively understudied. We conduct an in-depth analysis of the 12 most researched crop categories to discern trends in weed recognition research, and to understand why some crops are studied more intensively than others. This analysis reveals that the trajectory of research varies significantly between crops. We find that weed recognition in some globally critical crops is at an early stage of development, and lacks implementation and testing in real-world environments. Additionally, we find the differences in approach to weed recognition are not explained solely by the requirements of precision weed management for a given crop. Instead, the approaches taken, like with the choice of crop, often appear expedient, influenced by factors such as readily available annotated data, rather than by the crop-specific requirements of a precision weed management system.

近年来，在研究和商业领域出现了各种精准杂草管理技术。这些技术更有针对性地进行杂草管理干预，以提供更高效、更环保的杂草控制。为支持这一努力，大量研究集中于机器视觉识别各种作物中的杂草。在这项工作中，我们系统地调查了近期有关农作物杂草识别的文献，并根据联合国粮农组织（FAO）统计的全球农业状况对其相关性进行了评估。我们的研究结果表明，甜菜、胡萝卜和玉米等作物明显受到重视，而小麦和水稻尽管对全球耕地和粮食供应贡献巨大，但研究相对不足。我们对研究最多的 12 类作物进行了深入分析，以发现杂草识别研究的趋势，并了解为什么有些作物的研究比其他作物更深入。分析结果表明，不同作物的研究轨迹差异很大。我们发现，一些全球重要作物的杂草识别处于早期发展阶段，缺乏在实际环境中的实施和测试。此外，我们还发现，杂草识别方法的差异并不能完全归因于特定作物对杂草精准管理的要求。相反，与选择作物一样，所采取的方法往往是权宜之计，受到现成的注释数据等因素的影响，而不是受到精确杂草管理系统对特定作物的要求的影响。

{"title":"Review of weed recognition: A global agriculture perspective","authors":"Madeleine Darbyshire , Shaun Coutts , Petra Bosilj , Elizabeth Sklar , Simon Parsons","doi":"10.1016/j.compag.2024.109499","DOIUrl":"10.1016/j.compag.2024.109499","url":null,"abstract":"<div><div>Recent years have seen the emergence of various precision weed management technologies in both research and commercial contexts. These technologies better target weed management interventions to provide weed control that is more efficient and environmentally friendly. To support this effort, a significant amount of research has focused on machine vision to recognize weeds in a variety of crops. In this work, we systematically survey recent literature on weed recognition in crops and evaluate its relevance based on the status of global agriculture as presented in FAO statistics. Our findings indicate a notable emphasis on crops like sugar beet, carrot, and maize, while wheat and rice, despite their substantial contribution to global cropland and food supply, are relatively understudied. We conduct an in-depth analysis of the 12 most researched crop categories to discern trends in weed recognition research, and to understand why some crops are studied more intensively than others. This analysis reveals that the trajectory of research varies significantly between crops. We find that weed recognition in some globally critical crops is at an early stage of development, and lacks implementation and testing in real-world environments. Additionally, we find the differences in approach to weed recognition are not explained solely by the requirements of precision weed management for a given crop. Instead, the approaches taken, like with the choice of crop, often appear expedient, influenced by factors such as readily available annotated data, rather than by the crop-specific requirements of a precision weed management system.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"227 ","pages":"Article 109499"},"PeriodicalIF":7.7,"publicationDate":"2024-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142594113","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A crop’s spectral signature is worth a compressive text 作物的光谱特征值得压缩文本

IF 7.7 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2024-11-04 DOI: 10.1016/j.compag.2024.109576

Wei Cheng , Hongrui Ye , Xiao Wen , Qi Su , Huanran Hu , Jiachen Zhang , Feifan Zhang

The accuracy of crop mapping based on remotely sensed hyperspectral imagery has been significantly improved through the use of deep learning. However, traditional deep learning can be computationally intensive, requiring millions of parameters, which can make it ‘expensive’ to deploy and optimize. Inspired by studies in natural language processing, we consider the spectral signature corresponding to each pixel as text. Specifically, we first feed the hyperspectral image (HSI) data into the Channel2Vec module to generate channel embeddings. Based on the channel embeddings, we use a lossless compressor and Normalized Compression Distance (NCD) to create a spectral tokenizer. It can segment the spectral signature corresponding to each pixel into multiple windows along the channel dimension, and then extract local sequence information from each window. By combining the local sequence information with the original HSI data, we construct spectral embeddings. Finally, we again use the lossless compressor to compute the NCD between the spectral embeddings, and then classify using only the

k

-nearest-neighbor classifier (

k

NN). The proposed framework is ready-to-use and lightweight. Without any training, it achieves results competitive with deep learning models on three benchmark datasets. It outperforms the average of 11 advanced deep learning methods trained at scale. Moreover, it outperforms more than half of these models in the few-shot scenario, where there are not enough labels to effectively train a neural network.

通过使用深度学习，基于遥感高光谱图像的作物测绘精度得到了显著提高。然而，传统的深度学习计算密集，需要数百万个参数，因此部署和优化成本 "昂贵"。受自然语言处理研究的启发，我们将每个像素对应的光谱特征视为文本。具体来说，我们首先将高光谱图像（HSI）数据输入 Channel2Vec 模块，生成通道嵌入。在通道嵌入的基础上，我们使用无损压缩器和归一化压缩距离（NCD）来创建光谱标记器。它可以将每个像素对应的光谱特征沿信道维度分割成多个窗口，然后从每个窗口中提取局部序列信息。通过将局部序列信息与原始 HSI 数据相结合，我们构建了光谱嵌入。最后，我们再次使用无损压缩器计算光谱嵌入之间的 NCD，然后仅使用 k-nearest-neighbor 分类器（kNN）进行分类。所提出的框架是即用型和轻量级的。无需任何训练，它就能在三个基准数据集上取得与深度学习模型相当的结果。它的表现优于经过大规模训练的 11 种高级深度学习方法的平均水平。此外，在没有足够标签来有效训练神经网络的 "少数几个镜头 "场景中，它的表现优于一半以上的模型。

{"title":"A crop’s spectral signature is worth a compressive text","authors":"Wei Cheng , Hongrui Ye , Xiao Wen , Qi Su , Huanran Hu , Jiachen Zhang , Feifan Zhang","doi":"10.1016/j.compag.2024.109576","DOIUrl":"10.1016/j.compag.2024.109576","url":null,"abstract":"<div><div>The accuracy of crop mapping based on remotely sensed hyperspectral imagery has been significantly improved through the use of deep learning. However, traditional deep learning can be computationally intensive, requiring millions of parameters, which can make it ‘expensive’ to deploy and optimize. Inspired by studies in natural language processing, we consider the spectral signature corresponding to each pixel as text. Specifically, we first feed the hyperspectral image (HSI) data into the Channel2Vec module to generate channel embeddings. Based on the channel embeddings, we use a lossless compressor and Normalized Compression Distance (NCD) to create a spectral tokenizer. It can segment the spectral signature corresponding to each pixel into multiple windows along the channel dimension, and then extract local sequence information from each window. By combining the local sequence information with the original HSI data, we construct spectral embeddings. Finally, we again use the lossless compressor to compute the NCD between the spectral embeddings, and then classify using only the <span><math><mi>k</mi></math></span>-nearest-neighbor classifier (<span><math><mi>k</mi></math></span>NN). The proposed framework is ready-to-use and lightweight. Without any training, it achieves results competitive with deep learning models on three benchmark datasets. It outperforms the average of 11 advanced deep learning methods trained at scale. Moreover, it outperforms more than half of these models in the few-shot scenario, where there are not enough labels to effectively train a neural network.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"227 ","pages":"Article 109576"},"PeriodicalIF":7.7,"publicationDate":"2024-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142578238","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

Upscaling and downscaling approaches for early season rice yield prediction using Sentinel-2 and machine learning for precision nitrogen fertilisation 利用哨兵-2 和机器学习进行早稻产量预测的升尺度和降尺度方法，促进精准氮肥施用

IF 7.7 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2024-11-03 DOI: 10.1016/j.compag.2024.109603

Giorgio Impollonia, Michele Croci, Stefano Amaducci

Early season yield prediction could support rice farmers in adopting precision agriculture for nitrogen fertilisation management. Remote sensing and machine learning (ML) can be used to predict and map crop yield during phenological stages relevant to nitrogen application, like tillering in rice, at both within-field and field scales. This study evaluated the transferability of ML models in early season yield prediction through upscaling and downscaling approaches. The effects of two prediction times (tillering and ripening stages) and training/testing set sizes on ML models performance were evaluated over five rice growing seasons (from 2018 to 2022) in northern Italy, using whole-field-average yields and yield maps. Vegetation indices from Sentinel-2 imagery using the Google Earth Engine platform fed five ML algorithms (Cubist-CUB, Gaussian Process Regression-GPR, Neural Network-NNET, Random Forest-RF, and Support Vector Machines-SVM). ML algorithms were trained with yield maps and tested with whole-field-average yields to obtain a downscaling approach, while the opposite was done to obtain an upscaling approach. The downscaling approach showed higher accuracy than upscaling approach. Ripening stage predictions were more accurate than tillering stages, although the downscaling approach showed small differences between tillering and ripening stages. The highest tillering stage accuracy was achieved by SVM for both downscaling and upscaling approaches with 20 % and 27.8 % of Normalized Root Mean Square Error (NRMSE), and 0.99 and 0.99 of Simple Additive Weighting (SAW) score, respectively. Set size and data distribution effected ML models accuracy, with the highest performance achieved by RF and GPR with 0.80 and 1.00 of SAW score for the downscaling and upscaling approaches, respectively. This study demonstrated how ML models and downscaling approach could support rice farmers to calculate the nitrogen dose using the predicted yield at the tillering stage, enabling them to apply a site-specific nitrogen fertilisation based on the within-field yield prediction variability.

早季产量预测可帮助稻农采用精准农业进行氮肥管理。遥感和机器学习（ML）可用于预测和绘制与氮肥施用相关的物候期（如水稻分蘖期）的作物产量，既可用于田间，也可用于田间尺度。本研究通过升尺度和降尺度方法评估了 ML 模型在早季产量预测中的可转移性。利用全田平均产量和产量图，评估了意大利北部五个水稻生长季（2018 年至 2022 年）中两个预测时间（分蘖期和成熟期）和训练/测试集大小对 ML 模型性能的影响。使用谷歌地球引擎平台从哨兵-2 图像中获取的植被指数为五种 ML 算法（立方体-CUB、高斯过程回归-GPR、神经网络-NNET、随机森林-RF 和支持向量机-SVM）提供了支持。用产量图训练 ML 算法，并用全场平均产量进行测试，以获得降尺度方法，反之则获得升尺度方法。降尺度方法比升尺度方法显示出更高的准确性。尽管降尺度方法在分蘖期和成熟期之间显示出的差异很小，但成熟期预测比分蘖期预测更准确。在降尺度和升尺度方法中，SVM 的分蘖期预测准确率最高，归一化均方根误差（NRMSE）分别为 20 % 和 27.8 %，简单加权（SAW）得分分别为 0.99 和 0.99。集合大小和数据分布影响了 ML 模型的准确性，其中 RF 和 GPR 的性能最高，降尺度和升尺度方法的 SAW 分数分别为 0.80 和 1.00。这项研究表明，ML 模型和降尺度方法可帮助稻农利用分蘖期的预测产量计算氮肥剂量，使他们能够根据田间产量预测的变异性进行因地制宜的氮肥施用。

{"title":"Upscaling and downscaling approaches for early season rice yield prediction using Sentinel-2 and machine learning for precision nitrogen fertilisation","authors":"Giorgio Impollonia, Michele Croci, Stefano Amaducci","doi":"10.1016/j.compag.2024.109603","DOIUrl":"10.1016/j.compag.2024.109603","url":null,"abstract":"<div><div>Early season yield prediction could support rice farmers in adopting precision agriculture for nitrogen fertilisation management. Remote sensing and machine learning (ML) can be used to predict and map crop yield during phenological stages relevant to nitrogen application, like tillering in rice, at both within-field and field scales. This study evaluated the transferability of ML models in early season yield prediction through upscaling and downscaling approaches. The effects of two prediction times (tillering and ripening stages) and training/testing set sizes on ML models performance were evaluated over five rice growing seasons (from 2018 to 2022) in northern Italy, using whole-field-average yields and yield maps. Vegetation indices from Sentinel-2 imagery using the Google Earth Engine platform fed five ML algorithms (Cubist-CUB, Gaussian Process Regression-GPR, Neural Network-NNET, Random Forest-RF, and Support Vector Machines-SVM). ML algorithms were trained with yield maps and tested with whole-field-average yields to obtain a downscaling approach, while the opposite was done to obtain an upscaling approach. The downscaling approach showed higher accuracy than upscaling approach. Ripening stage predictions were more accurate than tillering stages, although the downscaling approach showed small differences between tillering and ripening stages. The highest tillering stage accuracy was achieved by SVM for both downscaling and upscaling approaches with 20 % and 27.8 % of Normalized Root Mean Square Error (NRMSE), and 0.99 and 0.99 of Simple Additive Weighting (SAW) score, respectively. Set size and data distribution effected ML models accuracy, with the highest performance achieved by RF and GPR with 0.80 and 1.00 of SAW score for the downscaling and upscaling approaches, respectively. This study demonstrated how ML models and downscaling approach could support rice farmers to calculate the nitrogen dose using the predicted yield at the tillering stage, enabling them to apply a site-specific nitrogen fertilisation based on the within-field yield prediction variability.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"227 ","pages":"Article 109603"},"PeriodicalIF":7.7,"publicationDate":"2024-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142571282","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Capped honey segmentation in honey combs based on deep learning approach 基于深度学习方法的蜂巢封盖蜜细分

IF 7.7 1区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Computers and Electronics in Agriculture

Pub Date : 2024-11-02 DOI: 10.1016/j.compag.2024.109573

Francisco J. Rodriguez-Lozano , Sergio R. Geninatti , José M. Flores , Francisco J. Quiles-Latorre , Manuel Ortiz-Lopez

Honey is the food stored by honey bees for periods when it is scarce in the field as well as being a product that is consumed worldwide by humans. Each hive generates different amounts of honey depending on the population of the bee hive, health state or environmental factors. In fact, the reserves of honey provide beekeepers with a double function: to predict the amount of honey that can be obtained and to analyze the state of the bee colonies. The assessment of honey reserves is commonplace in scientific research related to the health of bee colonies, genetic improvement or environmental issues, and emerging technologies can provide useful tools to evaluate honey stored in hives. In this context, this work proposes a methodology to detect the honey areas in high resolution photographs automatically using methods based on deep learning. Specifically, the methodology follows a “divide and conquer” approach where the images are separated into tiles with overlapping areas that are used by a semantic segmentation algorithm based on Feature Pyramid Network (FPN), detecting the honey in each tile to finally merge the tiles back into the complete image. The proposal has been compared with different feature extractors (backbones) and other semantic segmentation models, obtaining on average accurate results above 90% and 87% in the Dice score and IOU metrics respectively.

蜂蜜是蜜蜂在田间缺乏食物时储存的食物，也是全世界人类消费的产品。根据蜂群数量、健康状况或环境因素的不同，每个蜂巢产生的蜂蜜量也不同。事实上，蜂蜜储量为养蜂人提供了双重功能：预测可获得的蜂蜜量和分析蜂群状况。在与蜂群健康、遗传改良或环境问题相关的科学研究中，对蜂蜜储量进行评估已是司空见惯的事，而新兴技术则可为评估蜂巢中储存的蜂蜜提供有用的工具。在此背景下，这项工作提出了一种方法，利用基于深度学习的方法自动检测高分辨率照片中的蜂蜜区域。具体来说，该方法采用了一种 "分而治之 "的方法，即把图像分割成具有重叠区域的瓦片，由基于特征金字塔网络（FPN）的语义分割算法使用，检测每个瓦片中的蜂蜜，最后将瓦片合并成完整的图像。该建议与不同的特征提取器（骨干）和其他语义分割模型进行了比较，在 Dice score 和 IOU 指标上分别获得了平均高于 90% 和 87% 的准确结果。

{"title":"Capped honey segmentation in honey combs based on deep learning approach","authors":"Francisco J. Rodriguez-Lozano , Sergio R. Geninatti , José M. Flores , Francisco J. Quiles-Latorre , Manuel Ortiz-Lopez","doi":"10.1016/j.compag.2024.109573","DOIUrl":"10.1016/j.compag.2024.109573","url":null,"abstract":"<div><div>Honey is the food stored by honey bees for periods when it is scarce in the field as well as being a product that is consumed worldwide by humans. Each hive generates different amounts of honey depending on the population of the bee hive, health state or environmental factors. In fact, the reserves of honey provide beekeepers with a double function: to predict the amount of honey that can be obtained and to analyze the state of the bee colonies. The assessment of honey reserves is commonplace in scientific research related to the health of bee colonies, genetic improvement or environmental issues, and emerging technologies can provide useful tools to evaluate honey stored in hives. In this context, this work proposes a methodology to detect the honey areas in high resolution photographs automatically using methods based on deep learning. Specifically, the methodology follows a <em>“divide and conquer”</em> approach where the images are separated into tiles with overlapping areas that are used by a semantic segmentation algorithm based on Feature Pyramid Network (FPN), detecting the honey in each tile to finally merge the tiles back into the complete image. The proposal has been compared with different feature extractors (backbones) and other semantic segmentation models, obtaining on average accurate results above 90% and 87% in the Dice score and IOU metrics respectively.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"227 ","pages":"Article 109573"},"PeriodicalIF":7.7,"publicationDate":"2024-11-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142571484","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