Pub Date : 2024-10-01DOI: 10.1109/TAFE.2024.3433348
M. Nagaraju;Priyanka Chawla
One of India's main crops, maize, accounts for 2–3% of global production. Disease detection in maize fields has become increasingly difficult due to a lack of knowledge about disease symptoms. Furthermore, manual disease detection methods take a lot of time and are not effective. Recent developments in convolutional neural networks (CNNs) have exhibited remarkable performance in disease recognition and classification. A CNN is a deep learning technique that extracts the features from an image and performs the disease classification effectively. The optimization of hyperparameters is a tedious problem that impacts the performance of a model. The main purpose of the present research is to support future research to configure suitable hyperparameters to a model. In the present work, a deep CNN is proposed for the classification of seven different diseases of maize crop. Several hyperparameters, such as image size, batch size, number of epochs, optimizers, learning rate, kernel size, and number of hidden layers, were tested with various values in the experimental approach. The obtained results show that running the model for 200 epochs improved the classification accuracy with 87.44%. It also states that choosing input image sizes of 168 × 168 and 224 × 224 resulted in a good classification accuracy of 84.66% and 85.23%, respectively. The proposed deep CNN model has attained 85.83% classification accuracy with the Adam optimizer and a learning rate of 0.001. However, the results achieved by other optimizers, such as root-mean-square propagation (81.95%) and stochastic gradient descent (79.66%), are not better when compared with the Adam optimizer. Finally, the results have provided a better knowledge in selecting appropriate hyperparameters to the application of plant disease classification.
{"title":"Deep Learning-Based Maize Crop Disease Classification Model in Telangana Region of South India","authors":"M. Nagaraju;Priyanka Chawla","doi":"10.1109/TAFE.2024.3433348","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3433348","url":null,"abstract":"One of India's main crops, maize, accounts for 2–3% of global production. Disease detection in maize fields has become increasingly difficult due to a lack of knowledge about disease symptoms. Furthermore, manual disease detection methods take a lot of time and are not effective. Recent developments in convolutional neural networks (CNNs) have exhibited remarkable performance in disease recognition and classification. A CNN is a deep learning technique that extracts the features from an image and performs the disease classification effectively. The optimization of hyperparameters is a tedious problem that impacts the performance of a model. The main purpose of the present research is to support future research to configure suitable hyperparameters to a model. In the present work, a deep CNN is proposed for the classification of seven different diseases of maize crop. Several hyperparameters, such as image size, batch size, number of epochs, optimizers, learning rate, kernel size, and number of hidden layers, were tested with various values in the experimental approach. The obtained results show that running the model for 200 epochs improved the classification accuracy with 87.44%. It also states that choosing input image sizes of 168 × 168 and 224 × 224 resulted in a good classification accuracy of 84.66% and 85.23%, respectively. The proposed deep CNN model has attained 85.83% classification accuracy with the Adam optimizer and a learning rate of 0.001. However, the results achieved by other optimizers, such as root-mean-square propagation (81.95%) and stochastic gradient descent (79.66%), are not better when compared with the Adam optimizer. Finally, the results have provided a better knowledge in selecting appropriate hyperparameters to the application of plant disease classification.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"627-637"},"PeriodicalIF":0.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142430799","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 : 2024-10-01DOI: 10.1109/TAFE.2024.3460970
Claudio Tomazzoli;Davide Quaglia;Sara Migliorini
Assuming climatic homogeneity is no longer acceptable in greenhouse farming since it can result in less-than-ideal agronomic decisions. Indeed, several approaches have been proposed based on installing sensors in predefined points of interest (PoIs) to obtain a better mapping of climatic conditions. However, these approaches suffer from two main problems, i.e., identifying the most significant PoIs inside the greenhouse and placing a sensor at each PoI, which may be costly and incompatible with field operations. As regards the first problem, we propose a genetic algorithm to identify the best sensing places based on the agronomic definition of zones of interest. As regards the second problem, we exploit agricultural robots to collect climatic information to train a set of virtual sensors based on recurrent neural networks. The proposed solution has been tested on a real-world dataset regarding a greenhouse in Verona (Italy).
{"title":"Planning the Greenhouse Climatic Mapping Using an Agricultural Robot and Recurrent-Neural- Network-Based Virtual Sensors","authors":"Claudio Tomazzoli;Davide Quaglia;Sara Migliorini","doi":"10.1109/TAFE.2024.3460970","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3460970","url":null,"abstract":"Assuming climatic homogeneity is no longer acceptable in greenhouse farming since it can result in less-than-ideal agronomic decisions. Indeed, several approaches have been proposed based on installing sensors in predefined points of interest (PoIs) to obtain a better mapping of climatic conditions. However, these approaches suffer from two main problems, i.e., identifying the most significant PoIs inside the greenhouse and placing a sensor at each PoI, which may be costly and incompatible with field operations. As regards the first problem, we propose a genetic algorithm to identify the best sensing places based on the agronomic definition of zones of interest. As regards the second problem, we exploit agricultural robots to collect climatic information to train a set of virtual sensors based on recurrent neural networks. The proposed solution has been tested on a real-world dataset regarding a greenhouse in Verona (Italy).","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"617-626"},"PeriodicalIF":0.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10701545","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142409064","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}
The codling moth pest poses a significant threat to global crop production, with potential losses of up to 80% in apple orchards. Special camera-based sensor nodes are deployed in the field to record and transmit images of trapped insects to monitor the presence of the pest. This article investigates the embedding of computer vision algorithms in the sensor node using a novel state-of-the-art microcontroller unit (MCU), the GreenWaves Technologies' GAP9 system-on-chip, which combines 10 RISC-V general purposes cores with a convolution hardware accelerator. We compare the performance of a lightweight Viola–Jones detector algorithm with a convolutional neural network (CNN), MobileNetV3-SSDLite, trained for the pest detection task. On two datasets that differentiate for the distance between the camera sensor and the pest targets, the CNN generalizes better than the other method and achieves a detection accuracy between 83% and 72%. Thanks to the GAP9’s CNN accelerator, the CNN inference task takes only $text{147 ms}$ to process a 320 × 240 pixel image. Compared to the GAP8 MCU, which only relies on general-purpose cores for processing, we achieved 9.5× faster inference speed. When running on a 1000 mAh battery at 3.7 V, the estimated lifetime is approximately 199 days, processing an image every 30 s. Our study demonstrates that the novel heterogeneous MCU can perform end-to-end CNN inference with an energy consumption of just 4.85 mJ, matching the efficiency of the simpler Viola–Jones algorithm and offering power consumption up to 15× lower than previous methods.
{"title":"Accelerating Image-based Pest Detection on a Heterogeneous Multicore Microcontroller","authors":"Luca Bompani;Luca Crupi;Daniele Palossi;Olmo Baldoni;Davide Brunelli;Francesco Conti;Manuele Rusci;Luca Benini","doi":"10.1109/TAFE.2024.3451888","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3451888","url":null,"abstract":"The codling moth pest poses a significant threat to global crop production, with potential losses of up to 80% in apple orchards. Special camera-based sensor nodes are deployed in the field to record and transmit images of trapped insects to monitor the presence of the pest. This article investigates the embedding of computer vision algorithms in the sensor node using a novel state-of-the-art microcontroller unit (MCU), the GreenWaves Technologies' GAP9 system-on-chip, which combines 10 RISC-V general purposes cores with a convolution hardware accelerator. We compare the performance of a lightweight Viola–Jones detector algorithm with a convolutional neural network (CNN), MobileNetV3-SSDLite, trained for the pest detection task. On two datasets that differentiate for the distance between the camera sensor and the pest targets, the CNN generalizes better than the other method and achieves a detection accuracy between 83% and 72%. Thanks to the GAP9’s CNN accelerator, the CNN inference task takes only \u0000<inline-formula><tex-math>$text{147 ms}$</tex-math></inline-formula>\u0000 to process a 320 × 240 pixel image. Compared to the GAP8 MCU, which only relies on general-purpose cores for processing, we achieved 9.5× faster inference speed. When running on a 1000 mAh battery at 3.7 V, the estimated lifetime is approximately 199 days, processing an image every 30 s. Our study demonstrates that the novel heterogeneous MCU can perform end-to-end CNN inference with an energy consumption of just 4.85 mJ, matching the efficiency of the simpler Viola–Jones algorithm and offering power consumption up to 15× lower than previous methods.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"170-180"},"PeriodicalIF":0.0,"publicationDate":"2024-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142430798","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 : 2024-09-24DOI: 10.1109/TAFE.2024.3451489
Lin Jiao;Qihuang Liu;Haiyun Liu;Peng Chen;Rujing Wang;Kang Liu;Shifeng Dong
Achieving the precise and real-time detection of wheat spikes play a crucial role in wheat growth monitoring for precision agriculture community. Machine-learning methods are commonly introduced to automatically detect and count the wheat spikes, which need carefully selected hand-crafted feature descriptors, leading to time-consuming and poor performance. The deep learning has become a promising technology for the accurate detection wheat spikes, owing to its powerful ability of feature extraction. However, the obtained wheat spike images from UAV still have serious overlap, dense distribution, various orientations, and large aspect ratios, leading to poor performance of recent wheat spike detection method. To address the demand of precise and fast detection and counting of wheat spike with dense distribution and arbitrary-orientation, a novel deep learning-based method, WheatNet, has been proposed. The attention mechanism has been introduced the process of feature fusing to highlight the important features of wheat spike as well as inhibit the useless information. Additionally, to optimize the parameters of the network, a loss function with soft dynamic label assignment is adopted to reduce the number of low-quality matches, which provides significant performance gains over other wheat spike detectors. Furthermore, to achieve the precise detection of wheat spike with multi-orientations, a large-scale oriented wheat spike dataset has been constructed, named RoWheat, including 900 images and 50419 annotations with dense distribution and various orientation. Experimental studies demonstrate that the proposed WheatNet achieves a recall of 99.7% and mAP of 91.8%, showing its promising performance gain compared to other state-of-the-art methods.
{"title":"WheatNet: Attentional Path Aggregation Feature Pyramid Network for Precise Detection and Counting of Dense and Arbitrary-Oriented Wheat Spikes","authors":"Lin Jiao;Qihuang Liu;Haiyun Liu;Peng Chen;Rujing Wang;Kang Liu;Shifeng Dong","doi":"10.1109/TAFE.2024.3451489","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3451489","url":null,"abstract":"Achieving the precise and real-time detection of wheat spikes play a crucial role in wheat growth monitoring for precision agriculture community. Machine-learning methods are commonly introduced to automatically detect and count the wheat spikes, which need carefully selected hand-crafted feature descriptors, leading to time-consuming and poor performance. The deep learning has become a promising technology for the accurate detection wheat spikes, owing to its powerful ability of feature extraction. However, the obtained wheat spike images from UAV still have serious overlap, dense distribution, various orientations, and large aspect ratios, leading to poor performance of recent wheat spike detection method. To address the demand of precise and fast detection and counting of wheat spike with dense distribution and arbitrary-orientation, a novel deep learning-based method, WheatNet, has been proposed. The attention mechanism has been introduced the process of feature fusing to highlight the important features of wheat spike as well as inhibit the useless information. Additionally, to optimize the parameters of the network, a loss function with soft dynamic label assignment is adopted to reduce the number of low-quality matches, which provides significant performance gains over other wheat spike detectors. Furthermore, to achieve the precise detection of wheat spike with multi-orientations, a large-scale oriented wheat spike dataset has been constructed, named RoWheat, including 900 images and 50419 annotations with dense distribution and various orientation. Experimental studies demonstrate that the proposed WheatNet achieves a recall of 99.7% and mAP of 91.8%, showing its promising performance gain compared to other state-of-the-art methods.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"606-616"},"PeriodicalIF":0.0,"publicationDate":"2024-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142408886","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}
Food security, producing enough food for every person on the planet, is becoming a significant issue. Increasing world population and climate change are setting new challenges to food production. Water stress can cause severe damage to crops, and detecting and preventing this threat is crucial. Smart agriculture and the use of sensors directly on the field is a promising and rapidly evolving solution. Data collected by a large number of sensors must be analyzed and efficiently interpreted. In this context, machine learning is an effective solution. This article conducts a comparative analysis of several well-established machine learning models, all trained on a dataset enriched with a novel parameter for the assessment of plant health, the stem electrical impedance (modulus and phase). This feature gives promising results since it is a direct parameter of the plant itself. Moreover, the inclusion of the stem impedance parameter significantly boosted the model's performance, notably enhancing the effectiveness, particularly evident in the case of the top-performing model in this study, the random forest algorithm. When incorporating stem electrical impedance, this model achieved an impressive F1 score of 98%, markedly surpassing the 88% obtained in its absence. As a complementary analysis, a permutation feature performance analysis was conducted, highlighting the potential of stem impedance modulus as a promising feature for evaluating plant watering conditions. The removal of impedance modulus from the training model resulted in an average classification performance loss of 25% in terms of F1 score, suggesting how impedance monitoring is a promising approach for plant health management.
{"title":"Evaluation of Machine Learning Models for Water Stress Detection Using Stem Impedance","authors":"Federico Cum;Stefano Calvo;Alessandro Sanginario;Umberto Garlando","doi":"10.1109/TAFE.2024.3457156","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3457156","url":null,"abstract":"Food security, producing enough food for every person on the planet, is becoming a significant issue. Increasing world population and climate change are setting new challenges to food production. Water stress can cause severe damage to crops, and detecting and preventing this threat is crucial. Smart agriculture and the use of sensors directly on the field is a promising and rapidly evolving solution. Data collected by a large number of sensors must be analyzed and efficiently interpreted. In this context, machine learning is an effective solution. This article conducts a comparative analysis of several well-established machine learning models, all trained on a dataset enriched with a novel parameter for the assessment of plant health, the stem electrical impedance (modulus and phase). This feature gives promising results since it is a direct parameter of the plant itself. Moreover, the inclusion of the stem impedance parameter significantly boosted the model's performance, notably enhancing the effectiveness, particularly evident in the case of the top-performing model in this study, the random forest algorithm. When incorporating stem electrical impedance, this model achieved an impressive F1 score of 98%, markedly surpassing the 88% obtained in its absence. As a complementary analysis, a permutation feature performance analysis was conducted, highlighting the potential of stem impedance modulus as a promising feature for evaluating plant watering conditions. The removal of impedance modulus from the training model resulted in an average classification performance loss of 25% in terms of F1 score, suggesting how impedance monitoring is a promising approach for plant health management.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"314-322"},"PeriodicalIF":0.0,"publicationDate":"2024-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142430788","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 : 2024-09-23DOI: 10.1109/TAFE.2024.3449214
David J. Daniels;Frank Podd;Anthony J. Peyton;Qiao Cheng
Optimization of the yield of crops is essential for the security of the food supply and the efficiency of farming. This paper examines some of the issues and challenges involved with the measurement of the potato tubers within the soil using ground penetrating radar (GPR) in the U.K. An order of magnitude assessment of the received signal levels from single or multiple groups of potatoes is provided. The antenna configurations are based on loaded dipole antennas near the potato ridge surface. Measurements of potato tubers at two test sites in the U.K. are described, as well as an approach to signal processing to optimize detectability. The article provides a systematic study of GPR techniques for the monitoring of tuber growth.
{"title":"Application of Ground Penetrating Radar to Potato Crop Assessment","authors":"David J. Daniels;Frank Podd;Anthony J. Peyton;Qiao Cheng","doi":"10.1109/TAFE.2024.3449214","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3449214","url":null,"abstract":"Optimization of the yield of crops is essential for the security of the food supply and the efficiency of farming. This paper examines some of the issues and challenges involved with the measurement of the potato tubers within the soil using ground penetrating radar (GPR) in the U.K. An order of magnitude assessment of the received signal levels from single or multiple groups of potatoes is provided. The antenna configurations are based on loaded dipole antennas near the potato ridge surface. Measurements of potato tubers at two test sites in the U.K. are described, as well as an approach to signal processing to optimize detectability. The article provides a systematic study of GPR techniques for the monitoring of tuber growth.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"596-605"},"PeriodicalIF":0.0,"publicationDate":"2024-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142408756","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 : 2024-09-18DOI: 10.1109/TAFE.2024.3454109
Tanushree Dey;Somnath Bera;Lakshman Prasad Latua;Milan Parua;Anwesha Mukherjee;Debashis De
This article proposes a crop yield prediction and recommendation system for agriculture 5.0 based on edge computing, machine learning (ML), and steganography. In comparison with the existing crop yield prediction and recommendation frameworks, for the first time we are integrating steganography with edge computing and ML to provide a secure crop yield prediction and recommendation system. In the proposed system, an edge device is used for data preprocessing, and the private cloud server referred to as agri-server is maintained for data analysis and storage. For protecting data privacy during transmission, modified least significant bit-based image steganography is used. For data analysis, six ML approaches are used and compared based on their performance. The experimental results demonstrate that each ML approach achieves above 90% accuracy in crop yield prediction. The results also present that the proposed framework achieves highest prediction accuracy of 99.9% which is better than the existing crop yield prediction frameworks. The results also demonstrate that the proposed framework reduces the latency and energy consumption by $sim$ 10% compared to the remote cloud-based crop yield prediction framework.
本文提出了一种基于边缘计算、机器学习（ML）和隐写技术的农业 5.0 农作物产量预测和推荐系统。与现有的作物产量预测和推荐框架相比，我们首次将隐写术与边缘计算和 ML 相结合，提供了一个安全的作物产量预测和推荐系统。在提议的系统中，边缘设备用于数据预处理，而被称为农业服务器的私有云服务器则用于数据分析和存储。为了在传输过程中保护数据隐私，使用了基于最小有效位的修正图像隐写术。在数据分析方面，使用了六种 ML 方法，并根据其性能进行了比较。实验结果表明，每种 ML 方法在作物产量预测方面的准确率都超过了 90%。结果还表明，拟议框架的预测准确率最高，达到 99.9%，优于现有的作物产量预测框架。结果还表明，与基于云的远程作物产量预测框架相比，拟议框架减少了 10% 的延迟和能耗。
{"title":"iCrop: An Intelligent Crop Recommendation System for Agriculture 5.0","authors":"Tanushree Dey;Somnath Bera;Lakshman Prasad Latua;Milan Parua;Anwesha Mukherjee;Debashis De","doi":"10.1109/TAFE.2024.3454109","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3454109","url":null,"abstract":"This article proposes a crop yield prediction and recommendation system for agriculture 5.0 based on edge computing, machine learning (ML), and steganography. In comparison with the existing crop yield prediction and recommendation frameworks, for the first time we are integrating steganography with edge computing and ML to provide a secure crop yield prediction and recommendation system. In the proposed system, an edge device is used for data preprocessing, and the private cloud server referred to as agri-server is maintained for data analysis and storage. For protecting data privacy during transmission, modified least significant bit-based image steganography is used. For data analysis, six ML approaches are used and compared based on their performance. The experimental results demonstrate that each ML approach achieves above 90% accuracy in crop yield prediction. The results also present that the proposed framework achieves highest prediction accuracy of 99.9% which is better than the existing crop yield prediction frameworks. The results also demonstrate that the proposed framework reduces the latency and energy consumption by \u0000<inline-formula><tex-math>$sim$</tex-math></inline-formula>\u000010% compared to the remote cloud-based crop yield prediction framework.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"587-595"},"PeriodicalIF":0.0,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142409063","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}
Internal browning (IB)-related disorders in apples are causing significant economic losses, as they undermine consumer trust and market acceptability. Especially for susceptible cultivars, a comprehensive assessment of IB across the whole supply chain is crucial to meet consumer demand and trust, reduce food waste, and improve the profit margins of producers. To address these objectives, there is an urgent need for a fast, reliable, and portable nondestructive technique that enables real-time decisionmaking. In this study, apples were harvested early and late from two orchards and stored under two different conditions. After seven months storage, a representative sample of apples were analyzed using electrical impedance spectroscopy (EIS) to assess IB, categorizing the samples into healthy, slight brown, and severe brown. To validate the EIS results, a standard quality parameter, fruit firmness, was analyzed. The EIS spectrum shows that the magnitude in the lower frequency range (40 Hz to 1.4 kHz) and phase in mid-frequency range (1.4 to 15 kHz) yields the most promising results, with statistically significant differences (p$leq$0.001) and (p$leq$0.005), respectively. Contrarily, firmness measurement did not exhibit promising discrimination between healthy and internally browned apples (p-value of 0.21). Furthermore, the EIS spectrum of the three different classes were fitted using a single Cole equivalent model, revealing its efficacy as the best-fit equivalent circuit and offered valuable insights into the physio-chemical changes in biological cells. This work solidifies the EIS potential as a powerful tool for real-time, nondestructive, user-friendly, and cost-effective method in sustainable precision agriculture and food security assessment.
{"title":"Electrical Impedance Spectroscopy-Based Detection of Internal Browning Disorder in Apples","authors":"Sundus Riaz;Pietro Ibba;Nadja Sadar;Ahmed Rasheed;Stefan Stürz;Angelo Zanella;Luisa Petti;Paolo Lugli","doi":"10.1109/TAFE.2024.3454529","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3454529","url":null,"abstract":"Internal browning (IB)-related disorders in apples are causing significant economic losses, as they undermine consumer trust and market acceptability. Especially for susceptible cultivars, a comprehensive assessment of IB across the whole supply chain is crucial to meet consumer demand and trust, reduce food waste, and improve the profit margins of producers. To address these objectives, there is an urgent need for a fast, reliable, and portable nondestructive technique that enables real-time decisionmaking. In this study, apples were harvested early and late from two orchards and stored under two different conditions. After seven months storage, a representative sample of apples were analyzed using electrical impedance spectroscopy (EIS) to assess IB, categorizing the samples into healthy, slight brown, and severe brown. To validate the EIS results, a standard quality parameter, fruit firmness, was analyzed. The EIS spectrum shows that the magnitude in the lower frequency range (40 Hz to 1.4 kHz) and phase in mid-frequency range (1.4 to 15 kHz) yields the most promising results, with statistically significant differences (p<inline-formula><tex-math>$leq$</tex-math></inline-formula>0.001) and (p<inline-formula><tex-math>$leq$</tex-math></inline-formula>0.005), respectively. Contrarily, firmness measurement did not exhibit promising discrimination between healthy and internally browned apples (p-value of 0.21). Furthermore, the EIS spectrum of the three different classes were fitted using a single Cole equivalent model, revealing its efficacy as the best-fit equivalent circuit and offered valuable insights into the physio-chemical changes in biological cells. This work solidifies the EIS potential as a powerful tool for real-time, nondestructive, user-friendly, and cost-effective method in sustainable precision agriculture and food security assessment.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"3 1","pages":"26-33"},"PeriodicalIF":0.0,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10681631","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143820327","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}
Determining soil organic carbon (SOC) stock and its changes over time is crucial for understanding carbon cycling. This study evaluates the reliability of visible and near-infrared (Vis-NIR) spectroscopy as a cost-effective method for detecting SOC within monitoring, reporting, and verification (MRV) systems. Soil samples from a long-term field experiment (LTE) in northern Italy, comparing maize-based forage systems were used as a case study. Three sampling campaigns (2003, 2012, and 2018) were utilized for a total of 162 soil samples collected in the LTE (54 each). Soil samples archived were retrieved and scanned using a Vis-NIR spectrometer to create a site-specific soil spectral library (Site-SSL). Aiming to implement a local prediction model samples collected in 2003 were used as a training dataset to estimate the SOC of the soil samples collected in 2012 and 2018. Concurrently, a second prediction model was run adding 172 regional soil samples (Reg-SSL) collected the same soil-landscape as the LTE. N.4 model strategies were compared, including random forest (RF), cubist (CU), memory based learning (MBL) and support vector machine (SVM) on Site-SSL and Reg-SSL. A sensitivity analysis was performed to evaluate the impact of training sample size, followed by an assessment of the cost-benefit of spectroscopic approach compared to conventional analysis. The results showed that the Vis-NIR spectral libraries, along with the CU and SVM models, were able to detect changes in SOC in the Site-SSL dataset, yielding the best results. To maintain optimal performance, it is advisable to include the standard analyses of at least 10 percent of the subsequent monitoring samples in the training set.
{"title":"Can Soil Organic Carbon in Long-Term Experiments Be Detected Using Vis-NIR Spectroscopy?","authors":"Roberto Barbetti;Francesco Palazzi;Pier Mario Chiarabaglio;Carlos Lozano Fondon;Daniele Rizza;Alessandro Rocci;Carlo Grignani;Laura Zavattaro;Barbara Moretti;Maria Fantappiè;Stefano Monaco","doi":"10.1109/TAFE.2024.3449215","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3449215","url":null,"abstract":"Determining soil organic carbon (SOC) stock and its changes over time is crucial for understanding carbon cycling. This study evaluates the reliability of visible and near-infrared (Vis-NIR) spectroscopy as a cost-effective method for detecting SOC within monitoring, reporting, and verification (MRV) systems. Soil samples from a long-term field experiment (LTE) in northern Italy, comparing maize-based forage systems were used as a case study. Three sampling campaigns (2003, 2012, and 2018) were utilized for a total of 162 soil samples collected in the LTE (54 each). Soil samples archived were retrieved and scanned using a Vis-NIR spectrometer to create a site-specific soil spectral library (Site-SSL). Aiming to implement a local prediction model samples collected in 2003 were used as a training dataset to estimate the SOC of the soil samples collected in 2012 and 2018. Concurrently, a second prediction model was run adding 172 regional soil samples (Reg-SSL) collected the same soil-landscape as the LTE. N.4 model strategies were compared, including random forest (RF), cubist (CU), memory based learning (MBL) and support vector machine (SVM) on Site-SSL and Reg-SSL. A sensitivity analysis was performed to evaluate the impact of training sample size, followed by an assessment of the cost-benefit of spectroscopic approach compared to conventional analysis. The results showed that the Vis-NIR spectral libraries, along with the CU and SVM models, were able to detect changes in SOC in the Site-SSL dataset, yielding the best results. To maintain optimal performance, it is advisable to include the standard analyses of at least 10 percent of the subsequent monitoring samples in the training set.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"3 1","pages":"43-48"},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143821551","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}
Climate change, adverse weather conditions, and illegitimate farming practices have caused severe damage to the agricultural ecosystem, resulting in significant crop loss in the last decade. One of the major challenges is the breakout of plant diseases that harm the crop in the field. To address this issue, several artificial intelligence and Internet of Things-based systems have been developed for crop monitoring and containment of plant diseases at early stages. In this article, a real-time plant disease identification system is designed using drone-based surveillance and farmer's input. A lightweight plant disease classification model is deployed in the proposed system using a fusion of a vision transformer and a convolutional neural network. The proposed model deploys conditional attention with a statistical squeeze-and-excitation module to efficiently learn the plant disease patterns from images captured under normal and challenging weather conditions. With only 0.95 million trainable parameters, the performance of the proposed plant disease classification model surpasses that of seven state-of-the-art techniques on five public datasets and an in-house developed maize dataset from drone camera-captured images under varying environmental conditions. To provide a better learning experience of real-world data to the model, a generative adversarial network, C3GAN, inspired by cycleGAN, is proposed for data augmentation of the collected maize dataset. The system keeps updating the model parameters based on the feedback of agriculture experts and farmers when new diseases break out or the model's performance deteriorates on unseen data during the surveillance over a period of time.
{"title":"Real-Time Plant Disease Identification: Fusion of Vision Transformer and Conditional Convolutional Network With C3GAN-Based Data Augmentation","authors":"Poornima Singh Thakur;Shubhangi Chaturvedi;Pritee Khanna;Tanuja Sheorey;Aparajita Ojha","doi":"10.1109/TAFE.2024.3447792","DOIUrl":"https://doi.org/10.1109/TAFE.2024.3447792","url":null,"abstract":"Climate change, adverse weather conditions, and illegitimate farming practices have caused severe damage to the agricultural ecosystem, resulting in significant crop loss in the last decade. One of the major challenges is the breakout of plant diseases that harm the crop in the field. To address this issue, several artificial intelligence and Internet of Things-based systems have been developed for crop monitoring and containment of plant diseases at early stages. In this article, a real-time plant disease identification system is designed using drone-based surveillance and farmer's input. A lightweight plant disease classification model is deployed in the proposed system using a fusion of a vision transformer and a convolutional neural network. The proposed model deploys conditional attention with a statistical squeeze-and-excitation module to efficiently learn the plant disease patterns from images captured under normal and challenging weather conditions. With only 0.95 million trainable parameters, the performance of the proposed plant disease classification model surpasses that of seven state-of-the-art techniques on five public datasets and an in-house developed maize dataset from drone camera-captured images under varying environmental conditions. To provide a better learning experience of real-world data to the model, a generative adversarial network, C3GAN, inspired by cycleGAN, is proposed for data augmentation of the collected maize dataset. The system keeps updating the model parameters based on the feedback of agriculture experts and farmers when new diseases break out or the model's performance deteriorates on unseen data during the surveillance over a period of time.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"2 2","pages":"576-586"},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142408783","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：