IEEE Transactions on AgriFood Electronics最新文献

{"title":"Efficient Monitoring Framework for Spoilage Prediction in Wheat Grain Warehouse","authors":"Jyoti Devi;Anita Singhrova","doi":"10.1109/TAFE.2026.3685969","DOIUrl":"https://doi.org/10.1109/TAFE.2026.3685969","url":null,"abstract":"Efficient monitoring of grain warehouse is critical in reducing postharvest losses and preserving grain quality during storage and transport. This study presents analytical framework for real-time spoilage assessment using temperature (T), relative humidity (RH), and carbon dioxide (CO<inline-formula><tex-math>$_{2}$</tex-math></inline-formula>) data. A distributed network of calibrated high-quality sensors are deployed to continuously record environmental and biochemical variations within grain bulks. The collected multivariate time-series data are analyzed through 22 deep learning models, capable of extracting both spatial and temporal dependencies associated with spoilage progression. Spatial–temporal feature mapping enabled the identification of localized hotspots of spoilage, while sequential modeling captured dynamic respiration trends. The dataset represents a multi-sensor spatio-temporal monitoring system that captures ambient conditions and deploys 14 sensors across 19 cables and records each sensors readings at a fixed timestamp. It integrates environmental variables (ambient temperature, ambient RH, CO<inline-formula><tex-math>$_{2}$</tex-math></inline-formula>, and inner RH) with localized sensor and cable readings, enabling detailed analysis of thermal distribution, anomaly detection and multivariate prediction across interconnected sensing nodes. Experimental validation using controlled storage trials demonstrated a prediction accuracy of 95.54% for CO<inline-formula><tex-math>$_{2}$</tex-math></inline-formula>, 95.38% for RH, and 98.89% for temperature, enabling early warning of spoilage up to 30-day horizon before the onset of physical degradation. The frameworks also incorporate adaptive thresholds for automated environmental control, allowing timely action such as ventilation or drying. The proposed system’s results are categorized into stages-CO<inline-formula><tex-math>$_{2}$</tex-math></inline-formula> stage, humidity (RH) stage, temperature stage and a combined final stage to provide a clear and understandable assessment of safe and alert conditions. The system as a whole provides a robust, scalable and data-driven foundation for intelligent grain storage management. This represents an important step toward making postharvest monitoring more accurate, predictive and effective.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"4 2","pages":"609-626"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149133821","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}

{"title":"IEEE Circuits and Systems Society Publication Information","authors":"","doi":"10.1109/TAFE.2026.3739537","DOIUrl":"https://doi.org/10.1109/TAFE.2026.3739537","url":null,"abstract":"","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"4 2","pages":"C2-C2"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11731532","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149133641","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}

{"title":"Leveraging Label-Free and Gradient-Based Pruning for Efficient Neural Networks in Maize Disease Detection","authors":"Carlos Victorino Padeiro;Tse-Wei Chen;Takahiro Komamizu;Ichiro Ide","doi":"10.1109/TAFE.2026.3711909","DOIUrl":"https://doi.org/10.1109/TAFE.2026.3711909","url":null,"abstract":"Traditional methods for diagnosing crop diseases rely heavily on expert visual assessments, which are often costly, time-consuming, and prone to human error. Deep neural networks, particularly convolutional neural networks (CNNs), have demonstrated strong potential for automating disease detection. However, their deployment remains limited among smallholder farmers due to substantial computational demands and reliance on stable Internet connectivity. To bridge this accessibility gap, lightweight models that operate efficiently on resource-constrained offline devices are essential. Although model pruning is a promising approach for reducing computational overhead and enabling edge deployment, it often results in performance degradation, thereby limiting its practical applicability for high-accuracy crop disease detection. However, to determine the channels contributing little to the model’s discriminative capability but are incorrectly preserved is challenging. This article introduces a novel pruning method based on the gradient analysis of the output difference between pretrained and pruned models concerning noise-perturbed images and the weight parameters. We identify and eliminate less significant weight channels without compromising accuracy by measuring changes in the output of intermediate layers or blocks when small-weight perturbations are introduced. The proposed block weight and image perturbation-induced gradients method is applied to maize disease detection using the faster region-based CNN detector with different backbones. Notably, removing 30.00% of floating point operations and parameters of ConvNeXt results in only 0.62% drop of the Top-1 mean average precision at an intersection-over-union threshold of 0.50 (mAP@50) compared to the <italic>vanilla</i> model, particularly when the target model is tiny. It also shows the state-of-the-art performance across various model architectures such as MobileNetV2, ResNet-50, ConvNeXt, and Swin Transformer on standard Canadian Institute for Advanced Research 10-Class Dataset (CIFAR-10) and ImageNet classification benchmark datasets, even when 57.75% of the parameters are pruned. These gains hold consistently across the evaluated one- and two-stage detectors with different backbone architectures, such as CNNs and Transformers.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"4 2","pages":"733-747"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149131615","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}

{"title":"Underground-to-Aboveground Wireless Communication for IoT-Enabled Smart Agriculture: Multilayer Path Loss Characterization and Field Performance Evaluation","authors":"Mingqiang Han;Naiqian Zhang","doi":"10.1109/TAFE.2026.3729024","DOIUrl":"https://doi.org/10.1109/TAFE.2026.3729024","url":null,"abstract":"Reliable underground sensing is critical for Internet of Things (IoT)-enabled smart agriculture, yet underground wireless communication suffers from severe signal attenuation governed by the dielectric properties of stratified soil profiles. Existing path loss models typically treat the soil column as a homogeneous medium, which fails to capture the vertical dielectric variability of natural agricultural soils and degrades prediction accuracy as burial depth increases. This study investigates underground-to-aboveground communication using low-power, long-range IoT technologies and presents a multilayer signal propagation model (MLSPM) that, unlike prior models with fixed or material-defined layers, adaptively determines the soil layer structure from in-situ measurements of soil texture and volumetric water content. Extensive field experiments were conducted across five locations at burial depths of 10–50 cm, using three representative IoT transceivers: a Sigfox-compatible S2-LP (868 MHz) and two LoRa-based devices (SX1262 at 915 MHz and SX1280 at 2.4 GHz). The MLSPM achieves a root-mean-square error of 13.63 dB and a mean absolute error of 9.48 dB, substantially outperforming the modified Friis model and wireless underground sensor network path loss model under identical experimental conditions. Among the evaluated platforms, the SX1262 demonstrated the lowest path loss and highest communication reliability across diverse soil conditions and burial depths. These results provide practical guidelines for transceiver selection, burial depth optimization, gateway placement, and energy-efficient IoT network design for precision agriculture and agrifood sensing applications.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"4 2","pages":"780-789"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149133490","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}

{"title":"Toward Intelligent GPR: Wavelet Scattering Networks for Soil Water Content Prediction","authors":"Vincent Filardi;Noushin Khosravi Largani;Seyed Zekavat;Majid Moradikia;Brian Wilson;Radwin Askari;Oren Mangoubi","doi":"10.1109/TAFE.2026.3680066","DOIUrl":"https://doi.org/10.1109/TAFE.2026.3680066","url":null,"abstract":"This work investigates the development of intelligent ground-penetrating radar (GPR) for mapping the received GPR signal features into soil water content (SWC) across multiple depths. To address this, this article introduces a flexible feature-extraction and machine learning (ML) framework for estimating subsurface SWC at multiple depths from GPR measurements. Specifically, the article assesses the role of wavelet scattering networks (WSNs) data preprocessing for high-performance SWC prediction. The proposed methods’ performance is evaluated on in situ field measurements conducted via an off-the-shelf GPR that transmits multicarrier signals over 4 096 frequency components within the 400 MHz—2 GHz frequency range. The measurements were conducted in the Soilx Lab fields at Worcester Polytechnic Institute and are based on loamy soil texture material that is consistent with the topsoil composition of many local farms. Using this data, the performance of different supervised learning models is evaluated, including transformers, deep convolutional neural networks, recurrent neural networks (RNNs), random forest, linear regression (LR), and least absolute shrinkage and selection operator (LASSO), with diverse signal features extracted from 4 096 frequency components. To properly train the ML models, the received GPR signals were labeled via a high-performance moisture probe capable of measuring SWC at different depths. The WSN was deployed as a preprocessing step to extract higher order derivatives of the attenuation curve, i.e., the received signal power versus frequency components, enabling high-performance moisture per depth estimation via downstream ML models. We assess the WSN’s performance in predicting multiple root zone SWC by adjusting the filter count and maximum scale of the transform. In addition, this work determines the most important and optimal number of carrier frequencies required to maintain high prediction accuracy for SWC, aiming to minimize power consumption and the GPR operation speed for scanning all frequency components. Compared to the best-performing benchmark ML model, the proposed WSN feature extraction, paired with a downstream random forest model, achieved 41% higher accuracy with as low as 2.79 SWC root-mean-squared error in SWC prediction over four root zone depths. The article marks an excellent progress toward the implementation of intelligent GPR for megafarm soil subsurface moisture characterization.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"4 2","pages":"587-600"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149131560","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}

{"title":"Machine Learning Methods for Leaf Wetness Prediction Using Agrometeorological Data","authors":"Giovanni Paolo Colucci;Elena Filipescu;Fabio Scatozza;Daniele Trinchero","doi":"10.1109/TAFE.2026.3711283","DOIUrl":"https://doi.org/10.1109/TAFE.2026.3711283","url":null,"abstract":"This article addresses the estimation of leaf wetness by means of machine learning methods applied to agrometeorological data. The novel contribution lies in the application of the models to the top and bottom sides of the leaf, separately. For this purpose, data have been acquired in the field using weather stations equipped with several sensors: double-sided capacitive leaf wetness, air temperature and relative humidity, soil temperature and volumetric water content, wind speed, and rainfall. Leaf wetness measurements have been used as ground truth, whereas the other sensor data as features. Results demonstrate the applicability of the method, which has also been tested on edge. The dataset has been made publicly available under a Creative Commons license.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"4 2","pages":"690-698"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11631632","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149131994","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}

{"title":"IEEE Circuits and Systems Society Information","authors":"","doi":"10.1109/TAFE.2026.3739541","DOIUrl":"https://doi.org/10.1109/TAFE.2026.3739541","url":null,"abstract":"","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"4 2","pages":"C3-C3"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11731472","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149131543","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}

{"title":"Enhancing Strawberry Yield Forecasting With Backcasted IoT Sensor Data and Machine Learning","authors":"Tewodros Alemu Ayall;Andy Li;Matthew Beddows;Milan Markovic;Georgios Leontidis","doi":"10.1109/TAFE.2026.3702855","DOIUrl":"https://doi.org/10.1109/TAFE.2026.3702855","url":null,"abstract":"Rapid global population growth underscores the need for digitally enabled agricultural systems that support sustainable food production and data-driven resource management for farmers and stakeholders. The adoption of Internet of Things (IoT) technologies—capable of capturing real-time environmental (e.g., temperature, humidity) and operational (e.g., irrigation) parameters—is a crucial step toward enabling advanced applications, such as AI-based yield forecasting. However, the effectiveness of such models is often constrained by limited data availability, particularly in dynamic farm environments where IoT observations must be accumulated over multiple growing seasons. In this study, we deployed IoT sensors in strawberry production polytunnels over two growing seasons to collect data on water usage, internal and external temperature and humidity, soil moisture, soil temperature, and photosynthetically active radiation. These observations were combined with manually recorded yield data spanning four seasons. To address gaps in IoT data for the two seasons without sensor coverage, we developed an AI-based backcasting approach that synthesizes missing sensor observations using historical weather data from a nearby station and existing polytunnel measurements. We then trained AI-based yield forecasting models using both real and synthetic datasets. In this retrospective evaluation, results show that incorporating synthetic data improved yield forecasting accuracy, with models trained on the combined dataset outperforming those using only real sensor, weather, and yield data.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"4 2","pages":"659-667"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149130759","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}

{"title":"Deep Learning-Based Image Analysis for Automated Fruit Sizing","authors":"Emmanuel ZE Butsana;Zahid Rashid Sheikh;Navid Yazdi","doi":"10.1109/TAFE.2026.3718736","DOIUrl":"https://doi.org/10.1109/TAFE.2026.3718736","url":null,"abstract":"Accurate fruit sizing is a critical task in the agricultural supply chain, but current methods are either labor-intensive or rely on expensive, calibration-sensitive hardware. This study develops and evaluates a novel, low-cost deep learning framework for automated tomato circumference estimation using multiview RGB imagery and an embedded physical scale reference. A dataset of tomato images from two orthogonal viewpoints was captured using a smartphone, and a custom-designed gridded marker was applied to provide an embedded reference for scale and local curvature. A dual-branch convolutional neural network (CNN) based on EfficientNet-B1 was trained to regress circumference by fusing image features with extracted geometric metadata, such as curvature, eccentricity, and scale factor. The proposed dual-view model, which integrates the reference marker and extracted geometric features, achieved high accuracy, yielding an mean absolute error (MAE) of 3.70 mm (1.46% MAPE). This result represents a 55% improvement in MAE over the single-view baseline (8.19 mm MAE), and establishes a highly competitive relative percentage error (1.46% MAPE) that outperforms recent state of the art depth-sensing and artificial intelligence pipelines. The fusion of image features with metadata was found to be critical, as dual-view models without these features failed to converge effectively. The study concludes that the fusion of multiview geometric constraints with explicit, marker-derived physical scale cues is essential for achieving high-accuracy, low-cost fruit sizing. This framework presents a viable, scalable, and low-cost alternative to complex depth-sensing or stereo-vision systems, enabling field-deployable, millimeter-scale fruit sizing using only standard RGB cameras. Using this method, we achieve a 36% increase in accuracy compared to approaches relying on 3-D vision systems and outperform existing automated RGB vision systems.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"4 2","pages":"720-732"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11667678","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149131003","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}

{"title":"Farm-Scale Autonomous Welfare Monitoring in Smart Livestock Farming: A Systematic Review of Robotics and Multimodal AI With an Emphasis on the Lab-to-Farm Deployment Gap","authors":"Francois Gonothi Toure;Abdoulaye Baniré Diallo;Mounir Boukadoum","doi":"10.1109/TAFE.2026.3687491","DOIUrl":"https://doi.org/10.1109/TAFE.2026.3687491","url":null,"abstract":"While breakthroughs in autonomous robotics and multimodal artificial intelligence (AI) promise continuous, real-time monitoring for precision livestock farming, their practical on-farm application faces significant limitations, revealing a critical “lab-to-farm” deployment gap driven by fundamental challenges to the embodied AI in robotics community, including poor model generalization, simulation-to-reality fragility, and the absence of standard validation benchmarks. This review highlights today’s state of the art in autonomous robotics and AI to help understand and bridge the gap. From an initial pool of over 900 articles, we selected 33 studies from 2021 to 2025 to propose recommendations for adopting farm-scale autonomous monitoring. Our review reveals that about 67% of the robotics and autonomy research relies on simulation, with no validation in dynamic farm environments. Based on this finding, we propose a technical roadmap that focuses on three pillars: first, the use of Generative AI for data standardization and simulation-to-reality adaptation; second, the adoption of a “leave-one-farm-out” or LOFO protocol for rigorous field validation; and third, the development of Edge native systems. Furthermore, we define a standardized welfare Insight schema to facilitate the creation of reproducible datasets, thus enabling the development of truly robust models for livestock welfare.","PeriodicalId":100637,"journal":{"name":"IEEE Transactions on AgriFood Electronics","volume":"4 2","pages":"530-539"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149133437","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}