Pub Date : 2024-09-06DOI: 10.1109/TAFE.2024.3444730
Moteaal Asadi Shirzi;Mehrdad R. Kermani
In this article, we propose a new algorithm to improve plant recognition through the use of feature descriptors. The accurate results from this identification method are essential for enabling autonomous tasks, such as stem-stake coupling, in precision agriculture. The proposed method divides the input seedling color image into subimages within the International Commission on Illumination, for three color axes, L for lightness, A for the green-red component, and B for the blue-yellow component, color space and extracts seven key feature descriptors for each subimage. It then uses feature descriptors to create a matrix, which is employed to train an artificial neural network to determine optimized cutoff values. This network suggests cutoff values for a multilevel threshold segmentation for plant recognition. The method provides robust and real-time adaptive segmentation adaptable to various seedlings, backgrounds, and lighting conditions. By enabling accurate segmentation of the plant, morphological image processing can more effectively eliminate leaves to locate the seedling stem. This methodology automates image analysis in seedling propagation facilities and greenhouses and enables a wide range of precision agricultural tasks.
{"title":"Adaptive Feature-Based Plant Recognition","authors":"Moteaal Asadi Shirzi;Mehrdad R. Kermani","doi":"10.1109/TAFE.2024.3444730","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3444730","url":null,"abstract":"In this article, we propose a new algorithm to improve plant recognition through the use of feature descriptors. The accurate results from this identification method are essential for enabling autonomous tasks, such as stem-stake coupling, in precision agriculture. The proposed method divides the input seedling color image into subimages within the International Commission on Illumination, for three color axes, L for lightness, A for the green-red component, and B for the blue-yellow component, color space and extracts seven key feature descriptors for each subimage. It then uses feature descriptors to create a matrix, which is employed to train an artificial neural network to determine optimized cutoff values. This network suggests cutoff values for a multilevel threshold segmentation for plant recognition. The method provides robust and real-time adaptive segmentation adaptable to various seedlings, backgrounds, and lighting conditions. By enabling accurate segmentation of the plant, morphological image processing can more effectively eliminate leaves to locate the seedling stem. This methodology automates image analysis in seedling propagation facilities and greenhouses and enables a wide range of precision agricultural tasks.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"335-346"},"PeriodicalIF":0.0,"publicationDate":"2024-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142430880","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-05DOI: 10.1109/TAFE.2024.3445119
Baofeng Ji;Haoyu Li;Xin Jin;Ji Zhang;Fazhan Tao;Peng Li;Jianhua Wang;Huitao Fan
Timely detection and prevention of tomato leaf diseases are crucial for improving tomato yields. To address the issue of low efficiency in detecting tomato leaf diseases, this article proposes a lightweight tomato leaf disease recognition method. First, enhanced intersection over union is introduced in the you only look once v8 (YOLOv8) model to replace the complete intersection over union loss function, enhancing the accuracy of bounding box localization. To solve the problem of fixed sample shapes and square convolution kernels not adapting well to different targets, lightweight alterable Kernel convolution (AKConv) is introduced, providing arbitrary parameters and shapes for the convolution kernel. Inspired by the lightweight characteristics of AKConv, the C2f module is improved by integrating AKConv to reduce floating-point operations and computational complexity during the convolution process. Second, as it is not feasible to construct a lightweight model with a large depth to achieve sufficient accuracy, a new lightweight convolution technique is introduced. GSConv, combining the GS bottleneck and the efficient cross stage partial block (VoV-GSCSP), replaces the feature fusion layer to achieve lightweight feature enrichment. To test and train the model, a tomato leaf disease dataset was constructed. The improved model demonstrated higher accuracy and fewer parameters on the tomato leaf disease dataset. The improved model achieved an mean average precision 50 (mAP50) of 94.9 $%$ and an mAP50:95 of 75.6 $%$ , representing increases of 1.9 $%$ and 2.8 $%$ over the original model, respectively. The number of parameters is only 2 322 262, a reduction of 22.8 $%$ compared to the original model. This method meets the daily needs of tomato leaf disease detection, providing technical support for agricultural spraying robots to quickly and accurately detect tomato leaf diseases and precisely spray pesticides.
及时发现和预防番茄叶病对提高番茄产量至关重要。针对番茄叶病检测效率低的问题，本文提出了一种轻量级番茄叶病识别方法。首先，在 You only look once v8（YOLOv8）模型中引入了增强的交乘联合（enhanced intersection over union）损失函数，取代了完全交乘联合损失函数，提高了边界框定位的准确性。为了解决固定样本形状和方形卷积核不能很好地适应不同目标的问题，引入了轻量级可改变卷积核（AKConv），为卷积核提供任意参数和形状。受 AKConv 轻量级特性的启发，C2f 模块通过整合 AKConv 进行了改进，以减少卷积过程中的浮点运算和计算复杂度。其次，由于构建大深度的轻量级模型无法达到足够的精度，因此引入了一种新的轻量级卷积技术。GSConv 结合了 GS 瓶颈和高效的跨阶段部分块（VoV-GSCSP），取代了特征融合层，实现了轻量级的特征丰富。为了测试和训练该模型，构建了一个番茄叶病数据集。改进后的模型在番茄叶病数据集上表现出更高的精确度和更少的参数。改进模型的平均精确度 50 (mAP50) 为 94.9%，mAP50:95 为 75.6%，分别比原始模型提高了 1.9%和 2.8%。参数数仅为 2 322 262，比原始模型减少了 22.8%。该方法满足了番茄叶片病害检测的日常需求，为农业喷洒机器人快速准确地检测番茄叶片病害、精准喷洒农药提供了技术支持。
{"title":"Lightweight Tomato Leaf Intelligent Disease Detection Model Based on Adaptive Kernel Convolution and Feature Fusion","authors":"Baofeng Ji;Haoyu Li;Xin Jin;Ji Zhang;Fazhan Tao;Peng Li;Jianhua Wang;Huitao Fan","doi":"10.1109/TAFE.2024.3445119","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3445119","url":null,"abstract":"Timely detection and prevention of tomato leaf diseases are crucial for improving tomato yields. To address the issue of low efficiency in detecting tomato leaf diseases, this article proposes a lightweight tomato leaf disease recognition method. First, enhanced intersection over union is introduced in the you only look once v8 (YOLOv8) model to replace the complete intersection over union loss function, enhancing the accuracy of bounding box localization. To solve the problem of fixed sample shapes and square convolution kernels not adapting well to different targets, lightweight alterable Kernel convolution (AKConv) is introduced, providing arbitrary parameters and shapes for the convolution kernel. Inspired by the lightweight characteristics of AKConv, the C2f module is improved by integrating AKConv to reduce floating-point operations and computational complexity during the convolution process. Second, as it is not feasible to construct a lightweight model with a large depth to achieve sufficient accuracy, a new lightweight convolution technique is introduced. GSConv, combining the GS bottleneck and the efficient cross stage partial block (VoV-GSCSP), replaces the feature fusion layer to achieve lightweight feature enrichment. To test and train the model, a tomato leaf disease dataset was constructed. The improved model demonstrated higher accuracy and fewer parameters on the tomato leaf disease dataset. The improved model achieved an mean average precision 50 (mAP50) of 94.9\u0000<inline-formula><tex-math>$%$</tex-math></inline-formula>\u0000 and an mAP50:95 of 75.6\u0000<inline-formula><tex-math>$%$</tex-math></inline-formula>\u0000, representing increases of 1.9\u0000<inline-formula><tex-math>$%$</tex-math></inline-formula>\u0000 and 2.8\u0000<inline-formula><tex-math>$%$</tex-math></inline-formula>\u0000 over the original model, respectively. The number of parameters is only 2 322 262, a reduction of 22.8\u0000<inline-formula><tex-math>$%$</tex-math></inline-formula>\u0000 compared to the original model. This method meets the daily needs of tomato leaf disease detection, providing technical support for agricultural spraying robots to quickly and accurately detect tomato leaf diseases and precisely spray pesticides.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"563-575"},"PeriodicalIF":0.0,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142408745","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The assumption of climate homogeneity is no longer acceptable in greenhouse farming since it can result in less-than-ideal decisions. At the same time, installing a sensor in each area of interest is costly and unsuitable for field operations. In this article, we address this problem by putting forth the idea of virtual sensors; their behavior is modeled by a context-aware recurrent neural network trained through the contextual relationships between a small set of permanent monitoring stations and a set of temporary sensors placed in specific points of interest for a short period. More precisely, we consider not only space location but also temporal features and distance with respect to the permanent sensors. This article shows the complete pipeline to configure the recurrent neural network, perform training, and deploy the resulting model into an embedded system for on-site application execution.
{"title":"Estimating Greenhouse Climate Through Context-Aware Recurrent Neural Networks Over an Embedded System","authors":"Claudio Tomazzoli;Elia Brentarolli;Davide Quaglia;Sara Migliorini","doi":"10.1109/TAFE.2024.3441470","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3441470","url":null,"abstract":"The assumption of climate homogeneity is no longer acceptable in greenhouse farming since it can result in less-than-ideal decisions. At the same time, installing a sensor in each area of interest is costly and unsuitable for field operations. In this article, we address this problem by putting forth the idea of virtual sensors; their behavior is modeled by a context-aware recurrent neural network trained through the contextual relationships between a small set of permanent monitoring stations and a set of temporary sensors placed in specific points of interest for a short period. More precisely, we consider not only space location but also temporal features and distance with respect to the permanent sensors. This article shows the complete pipeline to configure the recurrent neural network, perform training, and deploy the resulting model into an embedded system for on-site application execution.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"554-562"},"PeriodicalIF":0.0,"publicationDate":"2024-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142408786","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Crop yield prediction is a crucial task in agricultural science, involving the classification of potential yield into various levels. This is vital for both farmers and policymakers. The features considered for this task are diverse, including weather, soil, and historical yield data. Recently, plant images captured in different modalities, such as red–green–blue, infrared, and multispectral bands, have also been utilized. Most of these data are inherently temporal. Integrating such multimodal and temporal data is advantageous for yield classification. In this work, a deep learning framework based on meta-transformers and temporal graph neural networks has been proposed to achieve this goal. Meta-Transformers allow the modeling of multimodal interactions, while temporayel graph neural networks enable the utilization of time sequences. Experimental results on the publicly available EPFL multimodal dataset demonstrate that the proposed framework achieves a high classification accuracy of nearly 97%, surpassing other state-of-the-art models, such as long short-term memory networks, 1-D convolutional neural networks, and Transformers. In addition, the proposed model excels in accuracy metrics, with a precision of approximately 98%, an F1-Score of 91%, and a recall of 94% in crop yield prediction.
{"title":"Crop Yield Prediction Using Multimodal Meta-Transformer and Temporal Graph Neural Networks","authors":"Somrita Sarkar;Anamika Dey;Ritam Pradhan;Upendra Mohan Sarkar;Chandranath Chatterjee;Arijit Mondal;Pabitra Mitra","doi":"10.1109/TAFE.2024.3438330","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3438330","url":null,"abstract":"Crop yield prediction is a crucial task in agricultural science, involving the classification of potential yield into various levels. This is vital for both farmers and policymakers. The features considered for this task are diverse, including weather, soil, and historical yield data. Recently, plant images captured in different modalities, such as red–green–blue, infrared, and multispectral bands, have also been utilized. Most of these data are inherently temporal. Integrating such multimodal and temporal data is advantageous for yield classification. In this work, a deep learning framework based on meta-transformers and temporal graph neural networks has been proposed to achieve this goal. Meta-Transformers allow the modeling of multimodal interactions, while temporayel graph neural networks enable the utilization of time sequences. Experimental results on the publicly available EPFL multimodal dataset demonstrate that the proposed framework achieves a high classification accuracy of nearly 97%, surpassing other state-of-the-art models, such as long short-term memory networks, 1-D convolutional neural networks, and Transformers. In addition, the proposed model excels in accuracy metrics, with a precision of approximately 98%, an F1-Score of 91%, and a recall of 94% in crop yield prediction.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"545-553"},"PeriodicalIF":0.0,"publicationDate":"2024-08-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142408785","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In the evolving landscape of precision agriculture, the integration of remote pest traps with deep learning technologies marks a critical step forward in remote pest detection, with the potential to substantially improve traditional pest monitoring methods. This article provides a comprehensive review of the developments, challenges, and innovative solutions in creating sensor-based electronic traps and applying deep learning for efficient and autonomous pest identification. By addressing the complexities of sensor integration, data collection, and the need for adaptive algorithms capable of classifying a wide range of insect pests, this review highlights the effective combination of electronic trap advancements with the precision offered by convolutional neural networks. An in-depth analysis of the technological advancements in electronic pest trap development is presented, highlighting improvements in design, efficiency, and sustainability while referring to ongoing and future challenges. Moreover, this article explores deep learning techniques, emphasizing on dataset enhancement and model optimization to overcome traditional challenges such as data scarcity and to improve the robustness of pest detection models. A thorough evaluation of various trap types against 85 unique pests is conducted, with the delta trap emerging as the most versatile, showcasing compatibility with multiple sensors and effectiveness against various pests. This review equips researchers, practitioners, and agricultural developers with critical insights and methodologies that can significantly enhance pest monitoring efficiency, reduce pesticide usage, and support sustainable agricultural practices.
{"title":"Insect Pest Trap Development and DL-Based Pest Detection: A Comprehensive Review","authors":"Athanasios Passias;Karolos-Alexandros Tsakalos;Nick Rigogiannis;Dionisis Voglitsis;Nick Papanikolaou;Maria Michalopoulou;George Broufas;Georgios Ch. Sirakoulis","doi":"10.1109/TAFE.2024.3436470","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3436470","url":null,"abstract":"In the evolving landscape of precision agriculture, the integration of remote pest traps with deep learning technologies marks a critical step forward in remote pest detection, with the potential to substantially improve traditional pest monitoring methods. This article provides a comprehensive review of the developments, challenges, and innovative solutions in creating sensor-based electronic traps and applying deep learning for efficient and autonomous pest identification. By addressing the complexities of sensor integration, data collection, and the need for adaptive algorithms capable of classifying a wide range of insect pests, this review highlights the effective combination of electronic trap advancements with the precision offered by convolutional neural networks. An in-depth analysis of the technological advancements in electronic pest trap development is presented, highlighting improvements in design, efficiency, and sustainability while referring to ongoing and future challenges. Moreover, this article explores deep learning techniques, emphasizing on dataset enhancement and model optimization to overcome traditional challenges such as data scarcity and to improve the robustness of pest detection models. A thorough evaluation of various trap types against 85 unique pests is conducted, with the \u0000<italic>delta trap</i>\u0000 emerging as the most versatile, showcasing compatibility with multiple sensors and effectiveness against various pests. This review equips researchers, practitioners, and agricultural developers with critical insights and methodologies that can significantly enhance pest monitoring efficiency, reduce pesticide usage, and support sustainable agricultural practices.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"323-334"},"PeriodicalIF":0.0,"publicationDate":"2024-08-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142430749","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-06DOI: 10.1109/TAFE.2024.3429537
Lennart Almstedt;Francesco Betti Sorbelli;Bas Boom;Rosalba Calvini;Elena Costi;Alexandru Dinca;Veronica Ferrari;Daniele Giannetti;Loretta Ichim;Amin Kargar;Catalin Lazar;Lara Maistrello;Alfredo Navarra;David Niederprüm;Peter Offermans;Brendan O'Flynn;Lorenzo Palazzetti;Niccolò Patelli;Cristina M. Pinotti;Dan Popescu;Aravind K. Rangarajan;Liviu Serghei;Alessandro Ulrici;Lars Wolf;Dimitrios Zorbas;Leonard Zurek
In this article, we introduce machine learning (ML) techniques developed for the monitoring of the brown marmorated stink bug (BMSB), a significant agricultural pest responsible for considerable crop damage worldwide. The Haly.ID project, initiated in early 2021, aims to enhance BMSB monitoring through the utilization of information and communication technology methods. We employ computer vision techniques on RGB images captured by drones and investigate the performance of deep neural networks to evaluate the impact of this invasive species on crop yields in orchards around Europe. Specifically, we evaluate the single shot multibox detector, detection transformer, YOLOv5, YOLOv9, and YOLOv10 architectures for full-level and patch-level image analysis, respectively. To improve detection accuracy, we experiment with shortwave infrared hyperspectral imaging (SWIR-HSI) in laboratory settings. Given that pheromone baited traps are the most accepted tools for pest detection by field operators, we also propose an Internet of Things sticky trap with an integrated camera equipped with lightweight convolutional neural networks models operating “on the edge” in this resource constrained system. In addition, we develop a client–server application for real-time bug detection, integrating the ML models to provide accessible results to farmers. Lastly, we explore effective postharvesting strategies using SWIR-HSI images to detect insect punctures invisible to the naked eye, thereby enhancing the quality of marketable fruit.
{"title":"Beyond the Naked Eye: Computer Vision for Detecting Brown Marmorated Stink Bug and Its Punctures","authors":"Lennart Almstedt;Francesco Betti Sorbelli;Bas Boom;Rosalba Calvini;Elena Costi;Alexandru Dinca;Veronica Ferrari;Daniele Giannetti;Loretta Ichim;Amin Kargar;Catalin Lazar;Lara Maistrello;Alfredo Navarra;David Niederprüm;Peter Offermans;Brendan O'Flynn;Lorenzo Palazzetti;Niccolò Patelli;Cristina M. Pinotti;Dan Popescu;Aravind K. Rangarajan;Liviu Serghei;Alessandro Ulrici;Lars Wolf;Dimitrios Zorbas;Leonard Zurek","doi":"10.1109/TAFE.2024.3429537","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3429537","url":null,"abstract":"In this article, we introduce machine learning (ML) techniques developed for the monitoring of the brown marmorated stink bug (BMSB), a significant agricultural pest responsible for considerable crop damage worldwide. The <sc>Haly.ID</small> project, initiated in early 2021, aims to enhance BMSB monitoring through the utilization of information and communication technology methods. We employ computer vision techniques on RGB images captured by drones and investigate the performance of deep neural networks to evaluate the impact of this invasive species on crop yields in orchards around Europe. Specifically, we evaluate the single shot multibox detector, detection transformer, <sc>YOLOv5</small>, <sc>YOLOv9</small>, and <sc>YOLOv10</small> architectures for full-level and patch-level image analysis, respectively. To improve detection accuracy, we experiment with shortwave infrared hyperspectral imaging (SWIR-HSI) in laboratory settings. Given that pheromone baited traps are the most accepted tools for pest detection by field operators, we also propose an Internet of Things sticky trap with an integrated camera equipped with lightweight convolutional neural networks models operating “on the edge” in this resource constrained system. In addition, we develop a client–server application for real-time bug detection, integrating the ML models to provide accessible results to farmers. Lastly, we explore effective postharvesting strategies using SWIR-HSI images to detect insect punctures invisible to the naked eye, thereby enhancing the quality of marketable fruit.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"3 1","pages":"6-17"},"PeriodicalIF":0.0,"publicationDate":"2024-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10623862","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143820326","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-05DOI: 10.1109/TAFE.2024.3434309
Kamlesh S. Patle;Neha Sharma;Priyanka Khaparde;Harsh Varshney;Gulafsha Bhatti;Yash Agrawal;Vinay S. Palaparthy
Prediction of plant diseases is essential to reduce crop loss. Early disease prediction models have been investigated for this purpose, where data on leaf wetness duration (LWD) is one of the key components. Leaf wetness sensors (LWSs) are used to better understand how foliar wetness affects plant disease cycles and epidemic development. LWS can be fabricated on printed circuit boards (PCBs), where interdigitated electrode patterns are widely used. However, it is important to understand the efficacy of these patterns for in-situ measurements. For this purpose, in this work, we have fabricated three different patterns viz. circular, oval, and rectangular on the PCB and tested their efficacy during lab and field measurements. Lab measurements indicate that the circular patterned LWS offers a sensitivity of about 1600% over the dry-to-wet range, which is about 2 and 1.5 times more than oval and rectangular patterns, respectively. Besides this, circular patterned LWS offers the hysteresis of about 2%, whereas the oval and rectangular patterned LWS show about 3% and 7%, respectively. Field measurement results specify that circular patterned LWS and commercial LWS Phytos 31 indicate the same number of LWD events. However, oval and rectangular patterned LWS shows extra false events.
{"title":"Impact of Electrode Patterns Variation on the Response Characteristic of Leaf Wetness Sensors","authors":"Kamlesh S. Patle;Neha Sharma;Priyanka Khaparde;Harsh Varshney;Gulafsha Bhatti;Yash Agrawal;Vinay S. Palaparthy","doi":"10.1109/TAFE.2024.3434309","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3434309","url":null,"abstract":"Prediction of plant diseases is essential to reduce crop loss. Early disease prediction models have been investigated for this purpose, where data on leaf wetness duration (LWD) is one of the key components. Leaf wetness sensors (LWSs) are used to better understand how foliar wetness affects plant disease cycles and epidemic development. LWS can be fabricated on printed circuit boards (PCBs), where interdigitated electrode patterns are widely used. However, it is important to understand the efficacy of these patterns for in-situ measurements. For this purpose, in this work, we have fabricated three different patterns viz. circular, oval, and rectangular on the PCB and tested their efficacy during lab and field measurements. Lab measurements indicate that the circular patterned LWS offers a sensitivity of about 1600% over the dry-to-wet range, which is about 2 and 1.5 times more than oval and rectangular patterns, respectively. Besides this, circular patterned LWS offers the hysteresis of about 2%, whereas the oval and rectangular patterned LWS show about 3% and 7%, respectively. Field measurement results specify that circular patterned LWS and commercial LWS Phytos 31 indicate the same number of LWD events. However, oval and rectangular patterned LWS shows extra false events.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"536-544"},"PeriodicalIF":0.0,"publicationDate":"2024-08-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142408784","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-31DOI: 10.1109/TAFE.2024.3432837
Marwane Rezzouki;Guillaume Ferré
In recent years, connecting fishing nets underwater has received much interest in carrying out fishing activities efficiently and protecting the ocean from pollution, especially from ghost fishing. In this article, we propose a hybrid acoustic system for communication and localization underwater. This system offers fishers the ability to enhance their fishing activities by establishing a reliable data link and facilitating the tracking of the fishing nets. The proposed system is based on a technique called differential chirp spread spectrum to connect fishing nets underwater. Moreover, multiple synchronized hydrophones are used at the receiver to calculate the time differential of arrival and then estimate the location of the acoustic sources. The communication performance of the proposed system is evaluated in terms of bit error rate using simulation and ocean experiments, whereas the localization performance is presented as a root-mean-square error using Bellhop-based channel modeling for network simulation.
{"title":"Sustainable Fishing: Chirp-Based Signals for Underwater Acoustic Communication and Localization","authors":"Marwane Rezzouki;Guillaume Ferré","doi":"10.1109/TAFE.2024.3432837","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3432837","url":null,"abstract":"In recent years, connecting fishing nets underwater has received much interest in carrying out fishing activities efficiently and protecting the ocean from pollution, especially from ghost fishing. In this article, we propose a hybrid acoustic system for communication and localization underwater. This system offers fishers the ability to enhance their fishing activities by establishing a reliable data link and facilitating the tracking of the fishing nets. The proposed system is based on a technique called differential chirp spread spectrum to connect fishing nets underwater. Moreover, multiple synchronized hydrophones are used at the receiver to calculate the time differential of arrival and then estimate the location of the acoustic sources. The communication performance of the proposed system is evaluated in terms of bit error rate using simulation and ocean experiments, whereas the localization performance is presented as a root-mean-square error using Bellhop-based channel modeling for network simulation.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"527-535"},"PeriodicalIF":0.0,"publicationDate":"2024-07-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142408781","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-25DOI: 10.1109/TAFE.2024.3415124
Basma Alsaid;Tracy Saroufil;Romaissa Berim;Sohaib Majzoub;Abir J. Hussain
Physical contamination of food is a prevalent issue within the food production industry. Contamination can occur at any stage of the food processing line. Many techniques are used in the literature for the detection of physical contamination in food. However, these techniques have some limitations when applied to fresh food products, particularly, when samples are characterized by diverse shapes and sizes. In addition, some of these techniques fail to detect hidden contaminants. In this work, we propose a novel approach to detect hidden physical contamination in fresh food products, including plastic fragments, stone fragments, and other foreign food objects, such as different food types that might inadvertently contaminate the sample. Electrical impedance tomography (EIT) is utilized to capture the impedance image of the sample to be used for contamination detection. Four deep learning models are trained using the EIT images to perform binary classification to identify contaminated samples. Three of the models are developed to detect the contaminants, each on its own, while the fourth model is used to detect any of the contaminates put together. The trained models achieved promising results with the accuracy of 85%, 92.9%, and 85.7% detecting plastic, stones, and other food types, respectively. The obtained accuracy when all contaminants put together was 78%. This performance shows the efficacy of the proposed approach over the existing techniques in the field.
食品的物理污染是食品生产行业普遍存在的问题。污染可能发生在食品加工生产线的任何阶段。文献中使用了许多检测食品物理污染的技术。然而，当这些技术应用于新鲜食品时，尤其是当样品的形状和大小各不相同时，就会受到一些限制。此外，其中一些技术还无法检测到隐藏的污染物。在这项工作中，我们提出了一种新方法来检测新鲜食品中隐藏的物理污染，包括塑料碎片、石块碎片和其他异物，如可能无意中污染样品的不同食物类型。电阻抗层析成像（EIT）被用来捕捉样品的阻抗图像，以用于污染检测。使用 EIT 图像训练了四个深度学习模型，以执行二元分类来识别受污染的样品。其中三个模型用于单独检测污染物，而第四个模型则用于综合检测任何污染物。经过训练的模型在检测塑料、石块和其他食物类型方面分别取得了 85%、92.9% 和 85.7% 的准确率，取得了可喜的成果。当所有杂质加在一起时，准确率为 78%。这一结果表明，与该领域的现有技术相比，所提出的方法非常有效。
{"title":"Food Physical Contamination Detection Using AI-Enhanced Electrical Impedance Tomography","authors":"Basma Alsaid;Tracy Saroufil;Romaissa Berim;Sohaib Majzoub;Abir J. Hussain","doi":"10.1109/TAFE.2024.3415124","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3415124","url":null,"abstract":"Physical contamination of food is a prevalent issue within the food production industry. Contamination can occur at any stage of the food processing line. Many techniques are used in the literature for the detection of physical contamination in food. However, these techniques have some limitations when applied to fresh food products, particularly, when samples are characterized by diverse shapes and sizes. In addition, some of these techniques fail to detect hidden contaminants. In this work, we propose a novel approach to detect hidden physical contamination in fresh food products, including plastic fragments, stone fragments, and other foreign food objects, such as different food types that might inadvertently contaminate the sample. Electrical impedance tomography (EIT) is utilized to capture the impedance image of the sample to be used for contamination detection. Four deep learning models are trained using the EIT images to perform binary classification to identify contaminated samples. Three of the models are developed to detect the contaminants, each on its own, while the fourth model is used to detect any of the contaminates put together. The trained models achieved promising results with the accuracy of 85%, 92.9%, and 85.7% detecting plastic, stones, and other food types, respectively. The obtained accuracy when all contaminants put together was 78%. This performance shows the efficacy of the proposed approach over the existing techniques in the field.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"518-526"},"PeriodicalIF":0.0,"publicationDate":"2024-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142408760","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this work, modeling, simulation, and experimental measurements of a LoRaWAN network aimed at implementing a dense farm-area network (FAN) in the agrifood industry are presented. First, the network is modeled for a farm of the future, with as many sensors as would be useful, for the four main productive chains in Uruguay as a study case: livestock, timber, agriculture, and dairy industries. To this end, a survey of commercial sensors was conducted, a few farms were visited, and managers and partners in agrocompanies were interviewed. A LoRaWAN network with a single gateway was simulated to estimate the efficiency (related to data packets lost), in the case of a 1000 ha cattle field with more than 1500 sensors and some cameras sharing the network. Finally, the network efficiency was measured, using 30–40 LoRa modules @ 915 MHz, transmitting at pseudorandom times to emulate up to thousands of LoRa sensor nodes. The simulated and measured results are very similar, reaching > 92% efficiency in all cases. Sites bigger than 1000 ha on the four main productive chains were also simulated. Additionally, energy consumption and transmission distance measurements of LoRaWAN modules are presented, as well as an overview of the economic aspects related to the deployment of the network to corroborate them fit the requirements of a FAN in the agribusiness.
{"title":"A Model for a Dense LoRaWAN Farm-Area Network in the Agribusiness","authors":"Alfredo Arnaud;Matías Miguez;María Eugenia Araújo;Ariel Dagnino;Joel Gak;Aarón Jimenz;José Job Flores;Nicolas Calarco;Luis Arturo Soriano","doi":"10.1109/TAFE.2024.3422843","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3422843","url":null,"abstract":"In this work, modeling, simulation, and experimental measurements of a LoRaWAN network aimed at implementing a dense farm-area network (FAN) in the agrifood industry are presented. First, the network is modeled for a farm of the future, with as many sensors as would be useful, for the four main productive chains in Uruguay as a study case: livestock, timber, agriculture, and dairy industries. To this end, a survey of commercial sensors was conducted, a few farms were visited, and managers and partners in agrocompanies were interviewed. A LoRaWAN network with a single gateway was simulated to estimate the efficiency (related to data packets lost), in the case of a 1000 ha cattle field with more than 1500 sensors and some cameras sharing the network. Finally, the network efficiency was measured, using 30–40 LoRa modules @ 915 MHz, transmitting at pseudorandom times to emulate up to thousands of LoRa sensor nodes. The simulated and measured results are very similar, reaching > 92% efficiency in all cases. Sites bigger than 1000 ha on the four main productive chains were also simulated. Additionally, energy consumption and transmission distance measurements of LoRaWAN modules are presented, as well as an overview of the economic aspects related to the deployment of the network to corroborate them fit the requirements of a FAN in the agribusiness.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"284-292"},"PeriodicalIF":0.0,"publicationDate":"2024-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142430881","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：