Computers and Electronics in Agriculture最新文献

{"title":"Grain-YOLO: An improved lightweight YOLO v8 and its android deployment for rice grains detection","authors":"Changjiang Liu ,&nbsp;Lei Zhong ,&nbsp;Jie Wang ,&nbsp;Junming Huang ,&nbsp;Yuwei Wang ,&nbsp;Maoxue Guan ,&nbsp;Xuantian Li ,&nbsp;Huiwen Zheng ,&nbsp;Xihong Hu ,&nbsp;Xu Ma ,&nbsp;Suiyan Tan","doi":"10.1016/j.compag.2025.110757","DOIUrl":"10.1016/j.compag.2025.110757","url":null,"abstract":"<div><div>Grains quantity per panicle is a significant indicator for rice cultivation. A rapid and precise detection of grains quantity is important for crop yield prediction, and is thereafter essential for field management, food security and policy-making. However, traditional characterization of rice grains quantity is time consuming, high cost and labor intensive. In this research, the newest version of YOLO series, the YOLOv8n is selected as the base network and investigated lightweight methods to improve both the precision and speed in rice grains detection. First, Simplified Spatial Pyramid Pooling − Fast (SimSPPF) module is a substitute for the Spatial Pyramid Pooling − Fast (SPPF) module in the backbone of YOLOv8n. Second, VoVGSCSP module is constructed and introduced into the neck network to improve the ability to extract small target features while decreasing model’s parameters and computational load. Afterward, C2f_DShC2D module is constructed by integrating of DShC2D and C2f modules, and the C2f modules in the neck network is replaced with C2f_DShC2D module. The construction of the C2f_DShC2D module further cuts down unnecessary computations and memory usage and maintaining accuracy. Next, the commonly used Convolution modules in the Neck network is replaced with Depthwise Separable Convolution (DSConv) module to further reduce the volume of models and accelerate the inference speed of models. In addition, MPDIoU is introduced to replace CIoU, which can effectively deal with the cases that the prediction box and the ground-truth box share the same aspect ratio. Finally, the constructed Grain-YOLO is then deployed on android-based smartphone. With comparatively performance evaluation, Grain-YOLO demonstrates a significant improvement in rice grain recognition accuracy and reduction in the computational load, with mAP value of 92.7 %, Params of 2.62 M and GFLOP of 7.0/s. Compared with the baseline YOLOv8n, Grain-YOLO increases mAP by 1.1 %, reduces Params by 12.96 % and lowers GFLOP by 13.58 %. Visualization analysis additionally indicates that the Grain-YOLO enhances feature extraction capability, thus focusing on identification of the occluded and small-scale grains. The corresponding android-based APP exhibits excellent performance in complex field scenarios, including different light levels, shooting distances and grain density. Overall, the results suggests that Grain-YOLO is a promising tool for rice grain recognition, and show significant applicability in rice-field.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"237 ","pages":"Article 110757"},"PeriodicalIF":7.7,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144614692","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Mechanism analysis and performance optimization of corn stalk cutting and throwing based on CFD-DEM coupling approach","authors":"Wenhang Liu ,&nbsp;Zhihong Yu ,&nbsp;Aorigele ,&nbsp;Qiang Su ,&nbsp;Xuejie Ma","doi":"10.1016/j.compag.2025.110731","DOIUrl":"10.1016/j.compag.2025.110731","url":null,"abstract":"<div><div>The efficient conversion of corn stalk resources into yellow-storage feed is an important approach to achieving sustainable development in animal husbandry. However, mainstream cutting-throwing devices commonly face technical bottlenecks such as high energy consumption and low operational efficiency. To address this issue, this study systematically conducts optimization design and performance verification of corn stalk chopping-throwing devices based on multi-scale mechanical analysis and coupled simulation techniques. By establishing a layered breakable flexible stalk model and a structural coupling model of the chopping-throwing device, the kinematic and dynamic characteristics of the stalk crushing process are revealed. A CFD-DEM gas–solid coupling numerical method is employed to analyze the mechanical behavior of crushed stalk particles in the fluid–solid coupling field, and single-factor experiments clarify the interaction mechanisms between key parameters (moving blade arc radius, blade length, spindle speed), internal flow field, and material transport. The arc radius of the moving blade can reduce cutting torque by altering the sliding-shearing and pushing composite cutting mode. Adjusting the blade length can control the velocity distribution of secondary vortex-dominated airflow and mitigate material retention phenomena. Although increasing spindle speed enhances airflow intensity and material throwing efficiency, it also leads to a nonlinear rise in energy consumption and turbulence dissipation rate. A three-factor, three-level simulation experiment is designed using the response surface methodology to construct a specific energy consumption prediction model and conduct significance testing of parameters, yielding the optimal parameter combination: moving blade arc radius of 500 mm, blade length of 519 mm, and spindle speed of 605 r/min. Bench validation experiments show that the optimized chopping-throwing device achieves a specific energy consumption of 21.55 kWh/kg, representing a significant 3.13 % reduction compared to the pre-optimization state, thus achieving a synergistic improvement in energy efficiency and operational performance. This study provides a reference for the optimization design of chopping-throwing devices.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"237 ","pages":"Article 110731"},"PeriodicalIF":7.7,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144614693","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Visual to near-infrared image translation for precision agriculture operations using GANs and aerial images","authors":"Marios Krestenitis,&nbsp;Konstantinos Ioannidis,&nbsp;Stefanos Vrochidis,&nbsp;Ioannis Kompatsiaris","doi":"10.1016/j.compag.2025.110720","DOIUrl":"10.1016/j.compag.2025.110720","url":null,"abstract":"<div><div>The rapid growth of computer vision and artificial intelligence (AI) techniques, along with advancements in sensory systems and unmanned aerial vehicles (UAVs), have profoundly impacted various fields such as Precision Agriculture (PA). A core operation in PA for crop monitoring and yield improvement is the combination of visual and near-infrared (NIR) wavelengths using pixel-wise operations known as Vegetation Indices (VIs). However, deploying costly multi-spectral sensory systems limits the scalability of existing PA solutions. Towards this direction, Generative Adversarial Networks (GANs) can be employed for transforming visual images to near-infrared representations, enabling the utilization of affordable off-the-shelf visual sensors and reducing system cost and complexity. Nevertheless, existing GAN-based methods for spectral domain translation often are limited to colorization models that produce pseudo-realistic images, neglecting the crucial spectral characteristics of the target domain. These synthetic images are commonly evaluated based on their visual plausibility rather than the spectral characteristic’s consistency. In the context of precision agriculture, such translations from visual to NIR domain may lead to inaccurate VI calculations and unreliable vegetation health estimation, making the usability of the synthesized data questionable. To overcome these limitations, we propose a model-agnostic modification for GANs that leverages the semantic information of VIs in the translation process. Our approach introduces an additional branch to the GAN architecture, calculating the Normalized Difference Vegetation Index (NDVI) from the input RGB and the generated NIR image. By backpropagating the additional branch loss, our method enforces the network to produce meaningful NIR representations that accurately preserve the domain’s spectral characteristics. We deploy our approach on two widely used GAN architectures, Pix2Pix and CycleGAN, and evaluate the synthesized results on relevant datasets. Experimental results demonstrate that the proposed method provides accurate and meaningful translations of visual to NIR images. The synthesized images maintain their semantic context under the near-infrared spectral attributes, making them suitable for precise VI calculations, vegetation health estimation, and efficient utilization in relative precision agriculture applications.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"237 ","pages":"Article 110720"},"PeriodicalIF":7.7,"publicationDate":"2025-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144605556","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development of a six-axis force sensor system for paddy field walking wheel to construct load spectrum model for fatigue life analysis","authors":"Miao Su ,&nbsp;Zhihao Zeng ,&nbsp;Jianfei He ,&nbsp;Yongliang Qiao ,&nbsp;Weiqing Jia ,&nbsp;Ziyou Guo ,&nbsp;Wenneng Zhong ,&nbsp;Zuncheng Fan ,&nbsp;Yizheng Dai ,&nbsp;York Huang ,&nbsp;Zaiman Wang","doi":"10.1016/j.compag.2025.110740","DOIUrl":"10.1016/j.compag.2025.110740","url":null,"abstract":"<div><div>Rice yield is crucial for global food security, and mechanized rice transplanting plays a vital role in boosting production. The complexity of the soil environment is increasing as the global rice cultivation area is shrinking year by year. Mechanized rice seedling transplanting heavily relies on the paddy field chassis. However, challenging paddy fields can significantly affect chassis performance, such as factors like different soil types, mud depth, and the roughness of hard strata, which ultimately impact the quality and yield of rice planting. The walking wheel of paddy field chassis is subjected to complex mechanical interactions with challenging paddy fields that are difficult to quantify and control. Quantifying these interactions enables precise optimization of chassis mechanical dynamics, thereby improving chassis performance in paddy field operations. However, existing research has not achieved precise, multi-dimensional, and dynamic measurements of the mechanical data, and lacks analytical methods for optimizing chassis performance. To address this, this study developed a sensor system for real-time, dynamic, and multi-dimensional measurement of six-axis forces on the walking wheel. Using this system, a load spectrum model for three typical operating conditions of the chassis was constructed, enabling fatigue life analysis and structural optimization. Fatigue life simulations validated the system’s reliability and provided a systematic method for structure optimization. Calibration and experimental results demonstrated measurement accuracy (RMSE &lt; 0.2 % of full scale) and excellent waterproof performance. This work can provide technical support for studying paddy chassis mechanics and contributes to the optimization and advancement of paddy chassis and similar agricultural machinery.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"237 ","pages":"Article 110740"},"PeriodicalIF":7.7,"publicationDate":"2025-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144596481","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"KDI-Transformer: A method for identifying kiwifruit leaf disease severity in complex environments","authors":"Xiaopeng Li ,&nbsp;Shuqin Li","doi":"10.1016/j.compag.2025.110745","DOIUrl":"10.1016/j.compag.2025.110745","url":null,"abstract":"<div><div>The accurate identification of the severity of kiwifruit leaf diseases faces significant challenges due to the high morphological similarity between different disease states and interference from complex environmental factors. To address this issue, we propose a Vision Transformer-based severity grading model for kiwifruit leaf diseases, called KDI-Transformer. This model deeply integrates the global modeling capability of Transformer with the local feature extraction advantages of Convolutional Neural Networks (CNNs). It incorporates three innovative modules: the Multi-Scale Perception Module (MSP), which extracts multi-granularity lesion features using parallel multi-scale convolutional kernels and integrates contextual information at different scales; the Adaptive Feature Transmission Module (AFT), which uses dynamic gating weights to adaptively adjust the inter-layer feature transmission ratio, effectively alleviating the feature attenuation problem in deep networks; and the Local-Global Interaction Module (LGI), which employs an attention mechanism for dynamic calibration of local features under global semantic guidance, significantly enhancing the model’s sensitivity to subtle disease differences. Experimental results demonstrate that KDI-Transformer achieves an accuracy of 89.57 %, significantly outperforming various baseline models, and provides a new solution for precise crop management in smart agriculture.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"237 ","pages":"Article 110745"},"PeriodicalIF":7.7,"publicationDate":"2025-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144597178","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Scalable machine learning framework for adaptive irrigation management of maize and soybean in the U.S. Midwest","authors":"Precious N. Amori ,&nbsp;Derek M. Heeren ,&nbsp;Yeyin Shi ,&nbsp;Eric Wilkening ,&nbsp;Ivo Z. Goncalves ,&nbsp;Guillermo R. Balboa ,&nbsp;Daran Rudnick ,&nbsp;Abia Katimbo ,&nbsp;Randall S. Ritzema","doi":"10.1016/j.compag.2025.110710","DOIUrl":"10.1016/j.compag.2025.110710","url":null,"abstract":"<div><div>Conventional soil water balance (SWB) irrigation scheduling tools, such as FAO-56-based Spreadsheets and the Spatial Evapotranspiration Modeling Interface (SETMI), rely heavily on manual inputs and periodic field measurements, leading to delayed recommendations and missed opportunities to prevent crop stress. More critically, these tools lack the computational scalability and adaptability to leverage the high-frequency, high-volume datasets now available through modern sensing technologies. As precision irrigation increasingly depends on integrating spatially and temporally nuanced field information, there is a pressing need for decision-support systems that can process Big Data efficiently and respond in real-time. To overcome these limitations, we developed and validated a machine learning (ML) framework for near real-time prediction of soil water depletion (SWD) and site-specific irrigation recommendations in maize and soybean production systems. Unlike prior models, our approach integrates multi-source, multi-year (2020 and 2023) datasets, including remote sensing data, weather variables, soil properties, management practices, yield records, time-related features, and geospatial information, to train Decision Tree, Random Forest, Gradient Boosting, and Extreme Gradient Boosting (XGB) models. Feature selection combined agronomic domain knowledge, correlation analysis, and Random Forest feature importance to retain relevant predictors while minimizing model complexity. The SWD was typically between 0 mm (field capacity) and 110 mm (management allowable depletion with a dynamic root zone increasing up to 1000 mm for the second half of the season) for irrigated plots, and up to 180 mm in rainfed conditions. Among the models, XGB performed best in the 2024 independent validation, predicting SWD with high accuracy (maize: R² = 0.72, RMSE = 22 mm; soybean: R² = 0.78, RMSE = 24 mm). The average predicted SWD values were 53 mm (maize) and 54 mm (soybean), closely matching SWB Spreadsheet estimates (47 mm and 54 mm, respectively). Field deployment in 2024 demonstrated the model’s operational potential, with ML-generated irrigation recommendations (62–70 mm for maize; 73–83 mm for soybean) closely aligning with FAO-56 Spreadsheet (61 mm maize; 79 mm soybean) and SETMI (64–96 mm soybean) benchmarks. However, testing on independent 2021 data revealed reduced generalization performance, highlighting the need for more diverse training datasets. Overall, this study advances a practical, scalable, ML-driven decision support framework for real-time precision irrigation in commercial cropping systems.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"237 ","pages":"Article 110710"},"PeriodicalIF":7.7,"publicationDate":"2025-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144605555","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CrYP: An open-source Google earth engine tool for spatially explicit crop yield predictions","authors":"Lorenzo Crecco,&nbsp;Sofia Bajocco,&nbsp;Simone Bregaglio","doi":"10.1016/j.compag.2025.110696","DOIUrl":"10.1016/j.compag.2025.110696","url":null,"abstract":"<div><div>Accurate yield forecasts are essential for supporting stakeholders in making informed strategic and tactical decisions. Process-based models are widely used for predicting crop yields, as they reproduce plant phenology and physiology in response to environmental conditions and agricultural practices. However, most crop models are point-based and must be integrated into spatially explicit simulation environments to produce yield predictions over larger areas at the desired spatial resolution. Remote sensing can provide critical data to inform crop models, offering consistent and high-quality observations of actual vegetation dynamics with time and space continuity. This study presents the Crop Yield Prediction (CrYP) app, an open-source tool designed for pixel-level crop yield forecasting over large regions. CrYP runs on the Google Earth Engine platform, applying a simple crop model executed in real time across geographic areas. The app uses ERA5-Land weather data and MODIS-derived Normalized Difference Vegetation Index, and can be adapted to different crops by tuning a few physiologically meaningful parameters. As a proof of concept, CrYP was tested on maize in the U.S. Corn Belt and on wheat and barley in two Italian regions. Results show that CrYP accurately captured seasonal crop phenology and the effects of abiotic stresses in both case studies, producing yield predictions consistent with official statistics. CrYP introduces an innovative approach for yield forecasting by assimilating remotely sensed data and real-time observed phenology into a process-based crop model. It also holds significant potential for near-real-time crop monitoring at local and regional scales, facilitating the timely identification of food security hotspots where adaptation strategies can be timely implemented.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"237 ","pages":"Article 110696"},"PeriodicalIF":7.7,"publicationDate":"2025-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144596482","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Optimization of pneumatic adsorption port for grape harvesting using discrete adjoint method and deflection analysis","authors":"Junlong Ma,&nbsp;Haochao Tan,&nbsp;Congcong Shen,&nbsp;Zhaoyang Guo,&nbsp;Huineng Zhou,&nbsp;Shuai Ma,&nbsp;Liming Xu","doi":"10.1016/j.compag.2025.110723","DOIUrl":"10.1016/j.compag.2025.110723","url":null,"abstract":"<div><div>To address the challenges of grasping grape bunches during robotic harvesting and reduce the pressure drop of the pneumatic adsorption ports, this study proposes a design and optimisation framework for the adsorption port based on the discrete adjoint method. A Computational Fluid Dynamics (CFD) model was established to simulate airflow within the adsorption port, using the airflow velocity at the port (<em>Vf</em>) to indicate adsorption capacity and the pressure drop between the port and the conveying pipe (<em>Pd</em>) to represent energy consumption. Model validation against bench tests demonstrated simulation accuracy, with relative errors of 9.65 % for <em>Vf</em> and 4.37 % for <em>Pd</em>. Single-factor experiments were conducted to investigate the effects of port length, height, and width on <em>Vf</em> and <em>Pd</em>. Central Composite Design (CCD) experiments were employed to optimise the port’s height and width, followed by further refinement of the wall geometry using the discrete adjoint method. The results indicated that an increase in port length led to a reduction in <em>Vf</em> and an increase in <em>Pd</em>, while increases in height and width resulted in simultaneous decreases in both parameters. The optimal dimensions obtained from the CCD experiments were a height of 70 mm and a width of 155.58 mm. Following discrete adjoint optimisation of the port’s wall, <em>Pd</em> was further reduced to 44.02 Pa, and <em>Vf</em> increased to 21.64 m·s^-1. To enable the grasping of grape bunches, orthogonal experiments were conducted using the discrete adjoint-optimised adsorption port to evaluate the deflection behaviour of grape bunches under pneumatic force. The results indicated that the distance between the fruit stem and the port (<em>Ds</em>), bunch’s weight (<em>M</em>), and stem length (<em>lp</em>) significantly influenced both the deflection angle (<em>a</em>) and the effective cutting length (<em>d</em>). It is recommended that <em>Ds</em> be maintained at 20 mm during the harvesting process. This study applies the discrete adjoint method to the design of the fruit adsorption port for the first time and evaluates the potential of this method in reducing <em>Pd</em> and improving <em>Vf</em>. Moreover, it reveals the influence of the growth parameters of grape bunches and the spatial distance between the bunches and the adsorption port on the deflection behaviour of grape bunches under pneumatic forces.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"237 ","pages":"Article 110723"},"PeriodicalIF":7.7,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144596483","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Classification of seed maize using deep learning and transfer learning based on times series spectral feature reconstruction of remote sensing","authors":"Wanqiu Chang ,&nbsp;Shuai Yang ,&nbsp;Xiaojie Xi ,&nbsp;Hengbin Wang ,&nbsp;Zhe Liu ,&nbsp;Xiaodong Zhang ,&nbsp;Shaoming Li ,&nbsp;Yuanyuan Zhao","doi":"10.1016/j.compag.2025.110738","DOIUrl":"10.1016/j.compag.2025.110738","url":null,"abstract":"<div><div>Accurately mapping the spatial distribution of seed maize fields is critical to securing seed supply, yet seed maize classification remains challenging due to similarities in interspecific crop cultivation. This study aims to construct a more suitable classification system for seed maize. A feature construction method based on time series spectral reconstruction was proposed, which explicitly enhanced the temporal-spectral correlation by reflecting the spectral reflectance information along with the temporal constraints onto the pixel-level grayscale image simultaneously. To clarify the benefits of classification strategies on the task of fine interspecific classification of maize, we compared two classification strategies, end-to-end direct classification and hierarchical classification, from the perspectives of both time cost and accuracy. The results showed that under the optimal time series range (March-early September) obtained by incorporating agricultural knowledge, ResNet-101 achieved an average accuracy of 91.34 %, which was better than other classification models. The input feature importance analysis revealed the classification mechanism of the model throughout the growth period. In order to improve the generalization ability of the model, we constructed two transfer learning frameworks for comparison. The accuracy of the method of constructing a joint dataset to train the model improved faster when the proportion of target domain samples introduced was small; the accuracy of the source domain pre-training-target domain fine-tuning method was higher when the number of samples introduced was larger. This study may provide a reference for the interspecific classification problem of other crops.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"237 ","pages":"Article 110738"},"PeriodicalIF":7.7,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144588727","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Agricultural processes simulation using discrete element method: a review","authors":"C. Maraveas,&nbsp;N. Tsigkas,&nbsp;T. Bartzanas","doi":"10.1016/j.compag.2025.110733","DOIUrl":"10.1016/j.compag.2025.110733","url":null,"abstract":"<div><div>The core aim of this review paper was to examine the latest state-of-the-art modeling of agricultural processes based on employing DEM and DEM-coupling methods using Multiphysics tools such as Computational Fluid Dynamics (CFD), Molecular Dynamics (MD), Finite Element Method (FEM), and Multi-Body Dynamics (MBD). DEM-coupling methods involve the integration of DEM with simulation methods such as CFD to model multi-phase systems while Multiphysics tools are simulation platforms which model physical phenomena, for example, fluid flow. The research motivation arose from the lack of review papers that examined how DEM was adopted to understand different agricultural processes and address the limitations of conducting expensive field experiments to develop such knowledge. The systematic review of 112 journal articles identified important insights regarding using DEM and coupling the method using Metaphysics tools. A key insight was the widely published DEM applications in areas such as modeling the interactions between soil and agricultural machinery such as mechanized tractors, plows, and shovels; simulating crop behavior such as bendable cotton stalks; modeling interactions between crops and agricultural machinery such as combine wheat harvesters and rice harvesters; and modeling the behavior between crops and soils such as seedling-emergence from the soil. The diverse applications indicate the robustness of the method in understanding agricultural processes. However, noteworthy gaps in the research include undertaking more work on DEM coupling applications, including MBD, MD, FEM, and CFD.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"237 ","pages":"Article 110733"},"PeriodicalIF":7.7,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144588728","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}