Computers and Electronics in Agriculture最新文献

{"title":"BiLSTM-SAGCN: A hybrid model of BiLSTM with a semiadaptation graph convolutional network for agricultural machinery trajectory operation mode identification","authors":"Weixin Zhai , Yucan Wu , Jinming Liu , Jiawen Pan , Caicong Wu","doi":"10.1016/j.compag.2025.110193","DOIUrl":"10.1016/j.compag.2025.110193","url":null,"abstract":"<div><div>Agricultural machinery trajectory operation mode identification is an important task in the analysis of agricultural machinery trajectory data, and its main objective is to classify the massive amount of data generated by agricultural machinery into different categories according to their operation modes. However, factors such as regional topography, weather and operational tasks affect position changes in trajectories; therefore, the spatial features of trajectories are complicated, which poses a great challenge to identifying agricultural machinery trajectory operation modes. The existing methods fail to fully mine the relationships among different ranges in the trajectory data space and do not consider the identification bias problem caused by the imbalanced distribution of agricultural machinery trajectories. To overcome the above shortcomings, we propose a hybrid model of BiLSTM with a semiadaptation graph convolutional network (BiLSTM-SAGCN) for agricultural machinery trajectory operation mode identification. First, to enrich the representation of trajectories, we propose a statistical-based feature enhancement module to mine the spatiotemporal feature information embedded in trajectories, which further enhances the performance of the model. Second, we develop a tailored hybrid network, which contains two key computations: one is to provide a low-cost topology learning method for the graph of agricultural machinery trajectories; we propose a semiadaptation graph convolutional network (SAGCN), which autonomously learns the weights of the edge relationships between nodes by constructing a masked graph structure through a self-attention mechanism and a spatiotemporal graph of agricultural machinery trajectories; and the other is to combine SAGCN with BiLSTM to form a hybrid network, in which SAGCN can interact between trajectory points to capture the dependencies between points, while BiLSTM is used to extract feature correlations along feature dimensions within a single trajectory point. Finally, to eliminate the identification bias problem caused by the imbalanced distribution of agricultural machinery trajectories, we develop a lightweight data balancing module, which adopts the focal loss function to guide the model to pay more attention to points that are difficult to classify during the training process, thereby effectively improving training efficiency. To evaluate the performance of the proposed model, we conducted experiments on 120 real agricultural machinery trajectory samples provided by the Key Laboratory of Agricultural Machinery Monitoring and Big Data Application, Ministry of Agriculture and Rural Affairs, with a total of 2,493,154 trajectory points, and compared our results with those of existing advanced agricultural machinery trajectory operation mode identification methods. The results revealed that the F1 score of BiLSTM-SAGCN reached 89.35% and 89.24% on the paddy and wheat harvester trajectory datasets, respect","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"233 ","pages":"Article 110193"},"PeriodicalIF":7.7,"publicationDate":"2025-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143528647","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":"Creating synthetic data sets for training of neural networks for automatic catch analysis in fisheries","authors":"Jonatan Sjølund Dyrstad,&nbsp;Elling Ruud Øye","doi":"10.1016/j.compag.2025.110160","DOIUrl":"10.1016/j.compag.2025.110160","url":null,"abstract":"<div><div>Accurate estimation of catch is essential for sustainable fisheries. It ensures precise catch reporting, provides a better basis for stock assessment, and helps prevent overfishing. With recent advances in deep learning, this could be solved using computer vision, however, collecting and annotating data for different fisheries, all with diverse catch distributions and different imaging equipment, is expensive and time-consuming and is currently limiting the adoption of the technology. To address this issue, we propose the use of synthetic data sets, created in simulation, for training of neural networks for the task of automatic catch analysis. Although the domain is subject to large amounts of variation in the image data, we hypothesize that much of this variation is due to clutter and variations in the appearance of the fish as captured by the camera, rather than inherent variations in the raw material itself. As such, the variation can be covered effectively in data sets generated in simulation, without the need for large data sets of 3D-models for each species, which are also costly to produce. This is demonstrated by training a neural network for instance segmentation, instance classification and key point detection, solely on synthetic data created with only five 3D-models of fish. The neural network is evaluated on real data, gathered with a variety of sensors onboard different fishing vessels, demonstrating that it generalizes across different domains. This evaluation concludes that synthetic data can be a valuable addition to real data for computer vision applications for catch analysis.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"233 ","pages":"Article 110160"},"PeriodicalIF":7.7,"publicationDate":"2025-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143528799","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development of a smart incubator for microalgae cultivation in food production: A case study of Spirulina","authors":"Albe Bing Zhe Chai ,&nbsp;Bee Theng Lau ,&nbsp;Irine Runnie Henry Ginjom ,&nbsp;Mark Kit Tsun Tee ,&nbsp;Pau Loke Show ,&nbsp;Enzo Palombo","doi":"10.1016/j.compag.2025.110163","DOIUrl":"10.1016/j.compag.2025.110163","url":null,"abstract":"<div><div>With the increasing awareness of nutritious food with environmentally friendly resources, microalgae cultivation is a promising sector to support the production of high-quality food. However, state-of-the-art cultivation solutions are mostly performed in large-scale settings at the<!--> <!-->industrial level. There is limited research that investigates the feasibility of developing small-scale solutions to support home-based microalgae cultivation. Hence, this study contributed to the Smart Microalgae Incubator system (SMIS), a novel and easy-to-manage IoT-based solution for small-scale home-based Spirulina cultivation. The SMIS is designed with functionalities such as growth monitoring and controlling, automated biomass harvesting, and medium recycling. A control center is included to control these operations based on the sensor readings of temperature, pH, water level, dissolved oxygen, and total dissolved solids in the main cultivation tank. Moreover, the turbidity center is designed to measure the turbidity level in the main tank so that the readiness for biomass harvesting is determined to trigger the automated harvesting. The proposed SMIS is utilized for a 125-day <em>Spirulina</em> cultivation and benchmarked with a control tank that cultivates <em>Spirulina</em> manually. Analysis of the<!--> <!-->growth rate and nutrient contents of <em>Spirulina</em> cultivated with both systems showed that the SMIS achieved comparable performance. Specifically, the harvested biomass at day 60 contains higher levels of protein (69.1 %), crude fat (10.3 %), and fiber (15.7 %). To conclude, the proposed SMIS is a significant and sustainable solution ideal for home-based <em>Spirulina</em> cultivation as a<!--> <!-->nutrient-rich food source. Further research is recommended to evaluate its effectiveness for cultivating other microalgae species. System refinement is also suggested to investigate its applicability for large-scale implementation.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"233 ","pages":"Article 110163"},"PeriodicalIF":7.7,"publicationDate":"2025-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143528797","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Precise Tomato Ripeness Estimation and Yield Prediction using Transformer Based Segmentation-SegLoRA","authors":"Sidharth N Pisharody ,&nbsp;Palmani Duraisamy ,&nbsp;Aravind Krishnaswamy Rangarajan ,&nbsp;Rebecca L. Whetton ,&nbsp;Ana Herrero-Langreo","doi":"10.1016/j.compag.2025.110172","DOIUrl":"10.1016/j.compag.2025.110172","url":null,"abstract":"<div><div>Accurate assessment of tomato (<em>Solanum lycopersicum</em>) ripeness is essential for the preservation of quality, meeting market demands and ensuring customer satisfaction. However, one of the key problems is accurately assessing the maturity levels of fruit under varying field conditions. Conventional computer vision models such as convolutional neural networks (CNN) demonstrate uneven performance under varying illumination conditions, particularly in arable farms. Further, it requires extensive training that involves fine-tuning entire model parameters and lags in global context learning. To address these issues, this work introduces a novel segmentation framework that integrates the SegFormer architecture with the Low-Rank Adaptation (SegLoRA) module. The proposed model attained significant performance improvement compared to state-of-the-art (SOTA) methods with a mean Intersection over Union (mIoU) of 83.25 %, an F1-score of 90.07 %, a test accuracy of 99.19 %, and a balanced accuracy of 93.88 %. Additionally, the computational cost was reduced by 26.98 % compared to existing SegFormer models. Further, the deployment on an edge computing device confirmed the proposed model’s feasibility in real time, with a minimal prediction delay of 0.065 s per frame. Moreover, its incorporation with an approximate yield estimation algorithm enables precise enumeration of harvestable tomatoes. These results demonstrate the scalability and efficiency of the SegLoRA, adding to the progress in automated ripeness detection and agricultural automation for selective harvesting operations.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"233 ","pages":"Article 110172"},"PeriodicalIF":7.7,"publicationDate":"2025-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143528798","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":"YOLOR-Stem: Gaussian rotating bounding boxes and probability similarity measure for enhanced tomato main stem detection","authors":"Guohua Gao,&nbsp;Lifa Fang,&nbsp;Zihua Zhang,&nbsp;Jiahao Li","doi":"10.1016/j.compag.2025.110192","DOIUrl":"10.1016/j.compag.2025.110192","url":null,"abstract":"<div><div>The tomato is a widely cultivated solanaceous vegetable worldwide and plays a crucial role in meeting human nutritional requirements. Non-invasive, time-dynamic automated representation and analysis of tomato main stems is critical for autonomous monitoring of canopy morphology throughout the entire tomato growth management cycle. Plant growth is influenced by genotype and environment, making naturally curved main stems and mutual shading of the branches and leaves, combined with the limited camera field of view and horizontal camera movement along crop rows, the sensing system observes only discontinuous and curved segments of the main stems. This study proposes an end-to-end YOLOR-Stem approach by optimizing the core components of YOLO v8. First, an innovative method for segmental labelling of main stems using multiple rotating bounding boxes is defined to ensure a precise description. Second, additional angular regression parameters are introduced to capture the orientation and scale information of main stem segments at any angle, overcoming the limitations of horizontal bounding boxes in unstructured field environments. Finally, the Hellinger distance measure is used to quantify the similarity between the predicted and ground truth distributions, integrated into the positive and negative sample matching strategy, loss function computation for rotated bounding boxes, and the prediction box screening during non-maximum suppression. The experimental results demonstrated that YOLOR-Stem (input size of 960 × 960 pixels) with the backbone of EfficientViT-M1 achieved 91.90 % mAP@50, 9.75 M parameters, 35.5GFLOPs, and 10.06 ms inference time. This study enables fast and accurate detection of visible segments of tomato plants, which lays the foundation for intelligent robot-plant interactions such as high-throughput phenotyping, branch and leaf pruning, growth detection, and autonomous harvesting.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"233 ","pages":"Article 110192"},"PeriodicalIF":7.7,"publicationDate":"2025-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143528648","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 design of a double planet carrier planetary gear train transplanting mechanism based on an MBD–DEM simulation of potted plant movement","authors":"Xingxiao Ma ,&nbsp;Chennan Yu ,&nbsp;Lihui Wang ,&nbsp;Xiaowei Zhang ,&nbsp;Jianneng Chen ,&nbsp;Xiong Zhao","doi":"10.1016/j.compag.2025.110141","DOIUrl":"10.1016/j.compag.2025.110141","url":null,"abstract":"<div><div>To design a simple and efficient flower transplanting mechanism, a thorough analysis of the potted plant cultivation process was conducted, and methods for designing the mechanism based on the posture constraints during the seedling retrieval and planting phases were investigated. A multidegree-of-freedom-driven virtual end-effector system was constructed. On the basis of the MBD-DEM analysis of the planting process for potted plants, a comparative analysis was conducted on the planting effects of the ordinary shovel, V-shaped shovel blades, and the bionic shovel under the same motion parameters. The bionic shovel was chosen as the structural form of the end-effector. Through parameter simulation optimization analysis of four factors, namely, the attitude angle of the end effector at the entry point into the soil and at the deepest planting point, the length of the hole, and the lateral planting distance, a set of motion parameters for the end effector was subsequently determined. This set of motion parameters was then translated into kinematic parameters for mechanism design; specifically, the length of the hole was 40 mm, the planting depth was 55 mm, the attitude angles of the seedling needle fixed at the entry point into the soil and at the deepest planting point were <span><math><mrow><mn>130</mn><mo>°</mo></mrow></math></span> and <span><math><mrow><mn>82</mn><mo>°</mo></mrow></math></span>, respectively, and the lateral planting distance was 8.6 mm. These parameters serve as the basis for the mechanism design posture. On the basis of the characteristics of a single-row two-stage noncircular gear transmission set, a design method for the double planetary gear train transplanting mechanism was proposed to address mixed postures. This method involves variables such as the rotation angle of the sun gear, the rotation angle of the middle gear, the length of each rack and the initial installation angle of each rack. The objective is to minimize the deviation between the actual position and the target position of the end-effector while ensuring the transmission performance of the noncircular gearset. The motion parameters obtained from the simulation results were converted into kinematic parameters for mechanism design, completing the design of the flower transplanting mechanism. A potted plant cultivation test bench was constructed, and potted plant cultivation experiments were conducted. The average planting rate reached 85.94 %, validating the effectiveness of the planting motion analysis results based on virtual simulation technology. These results demonstrate the practicality of the designed flower transplanting mechanism.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"233 ","pages":"Article 110141"},"PeriodicalIF":7.7,"publicationDate":"2025-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143527559","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":"AI-driven time series analysis for predicting strawberry weekly yields integrating fruit monitoring and weather data for optimized harvest planning","authors":"Shiyu Liu ,&nbsp;Yiannis Ampatzidis ,&nbsp;Congliang Zhou ,&nbsp;Won Suk Lee","doi":"10.1016/j.compag.2025.110212","DOIUrl":"10.1016/j.compag.2025.110212","url":null,"abstract":"<div><div>Strawberries, as an indeterminate crop, produce fruit multiple times per season, making fruit monitoring and wave-specific yield prediction essential for optimizing harvest planning. This study developed an AI-driven approach to predict next week’s yield using real-time plant image data collected by a machine vision system and environmental data. YOLOv8n was employed to count flowers, immature fruit, and mature fruit per plant, with manual counts used to evaluate the system’s accuracy. The YOLOv8n-based data, combined with weather features, were used to train several AI models for yield prediction. These models included traditional time series machine learning approaches, such as Multiple Linear Regression (MLR) with time lag features, Vector Autoregression (VAR), Gradient Boosting Machines (GBM), Random Forest, and deep learning time-series models, including Long Short-Term Memory (LSTM) and Temporal Convolutional Networks (TCN). Recursive Feature Elimination (RFE) was employed to identify the most relevant features. The performance of these models was evaluated across three strawberry varieties: Sensation, Brilliance, and Medallion. Results showed that MLR outperformed other models for Sensation and Brilliance, with R² values of 0.633 and 0.908, respectively. For Medallion, GBM achieved the best performance with an R² score of 0.848. LSTM, which outperformed TCN, achieved R² scores of 0.522 (Sensation), 0.839 (Brilliance), and 0.740 (Medallion). This AI-driven system automates yield forecasting, reducing labor costs and enabling more efficient harvest planning. The study highlights the potential of combining machine vision and predictive analytics for precise, scalable yield prediction, offering valuable insights for proactive farm management and supply chain optimization.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"233 ","pages":"Article 110212"},"PeriodicalIF":7.7,"publicationDate":"2025-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143527560","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Discrete element flexible modeling and experimental verification of rice blanket seedling root blanket","authors":"Xuan Jia ,&nbsp;Xiaopei Zheng ,&nbsp;Licai Chen ,&nbsp;Cailing Liu ,&nbsp;Jiannong Song ,&nbsp;Chengtian Zhu ,&nbsp;Jitong Xu ,&nbsp;Shuaihua Hao","doi":"10.1016/j.compag.2025.110155","DOIUrl":"10.1016/j.compag.2025.110155","url":null,"abstract":"<div><div>Using the discrete element method (DEM) to simulate the rice machine transplanting operation is important for assessing the plant injury and optimizing the rice transplanter performance, while the DEM flexible model establishment that can accurately reflect the mechanical properties of the rice blanket seedling root blanket is an important foundation. Based on the root blanket’s stratification and the root system structure’s measurement and statistics, a new method for root blanket flexible modeling was proposed in this study. Firstly, the Hertz-Mindlin with bonding V2 contact model was used to establish substrate Ⅰ (SⅠ), substrate Ⅱ (SⅡ), substrate Ⅲ (SⅢ), stem-root combination (SRC), and netted layer (NL) flexible models, respectively, and the model parameters were calibrated and determined by angle of repose (AOR), direct shear, and mechanical tests. The calibration results showed that the deviations of AOR simulated values for SⅠ and SⅡ were both less than 1.5 %, and the deviations of shear strength simulated values were both less than 4 %. Secondly, the shear characteristics of SⅠ and SⅡ were determined by direct shear test. The results showed that the physical and simulated shear stress-displacement relationship curves of SⅠ and SⅡ were basically the same; the hair roots mainly relied on the cohesive between them and the substrate to improve the substrate strength; the fitted lines of simulated shear strength and normal stress of SⅠ and SⅡ were in high agreement with these of the measured values; the deviations of the simulated cohesion and internal friction angle were both less than 5 %. After that, the Hertz Mindlin with JKR V2 contact model was used between SRC and substrate. The interfacial surface energy of the root blanket and the bonding parameters of SⅢ were calibrated by stem, half-SRC, and SRC pulling-out tests layer by layer. The calibration results showed that the deviation of the maximum pulling-out force of SRC was 5.83 %, verifying that the model could accurately simulate the intertwining effect of the crown roots. Finally, the flexible model of the root blanket was verified by cutting, curling, and tensile tests. The simulated test results were consistent with the trends of the physical test results; the deviations of the maximum cutting resistance of front cutting and side cutting were both within 8 %, the error percentage range of the marked points height was 0.35 % to 17.16 %, and the deviation of the maximum tensile force was 9.22 %, indicating the good feasibility of the modeling method and accuracy of the flexible model. The results of this study lay a foundation for the DEM simulation of the rice machine transplanting operation. They can also provide a reference for the numerical simulation of other multi-plant root-soil complexes.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"233 ","pages":"Article 110155"},"PeriodicalIF":7.7,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143527558","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":"Improved lightweight YOLOv5n-based network for bruise detection and length classification of asparagus","authors":"Xia Chuang,&nbsp;Chen Qiang,&nbsp;Shi Yinyan,&nbsp;Wang Xiaochan,&nbsp;Zhang Xiaolei,&nbsp;Wu Yao,&nbsp;Wang Yiran","doi":"10.1016/j.compag.2025.110194","DOIUrl":"10.1016/j.compag.2025.110194","url":null,"abstract":"<div><div>Efficient and accurate online quality recognition is crucial for asparagus production. To address the slow manual detection speed and low recognition efficiency in asparagus grading, as well as the limitations of traditional single-label classification algorithms in identifying bruise locations and classifying lengths, this study proposes an asparagus length classification and bruise detection method using the You-Only-Look-Once (YOLOv5n) convolutional neural network. Two lightweight, improved models—YOLOv5n with a spatial grouping strategy (YOLOv5n-SGS) and YOLOv5n with a global enhancement strategy (YOLOv5n-GES)—were connected in series to perform length classification and bruise detection of asparagus. The YOLOv5n-SGS model incorporated the ShuffleNet backbone, ghost spatial convolution (GSConv), and VoVNet ghost spatial cross-stage partial (VoV-GSCSP) modules, reducing computational complexity while improving efficiency in length classification. The YOLOv5n-GES model integrated the GhostNet backbone, GhostConv, and C3Ghost modules to enhance detection speed. A simple parameter-free attention module (SimAM) was added to improve semantic feature extraction, while an efficient intersection over union (EIoU) loss function was employed to enhance convergence and recognition accuracy. Test results demonstrated that the YOLOv5n-SGS model achieved a mean average precision (mAP) of 96.5 %, reducing computational complexity to 10 % of that of YOLOv5n. The YOLOv5n-GES model attained an mAP of 97.3 %, with complexity reduced to 59 %. The combined performance of both models outperformed similar approaches. By connecting both models with a cropping layer, overall classification and detection accuracy exceeded 95 % with total number of parameters only 1.221 M. The proposed method significantly improves high-precision length classification and bruise detection in asparagus, advancing industrial automation and enhancing the economic value of asparagus production.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"233 ","pages":"Article 110194"},"PeriodicalIF":7.7,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143520478","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":"Characterizing key foliar functional traits of subtropical evergreen forests in South China using leaf and UAV-based spectroscopy","authors":"Zhihui Wang ,&nbsp;Zhongyu Sun ,&nbsp;Nanfeng Liu ,&nbsp;Shoubao Geng ,&nbsp;Meili Wen ,&nbsp;Hui Zhang ,&nbsp;Long Yang","doi":"10.1016/j.compag.2025.110178","DOIUrl":"10.1016/j.compag.2025.110178","url":null,"abstract":"<div><div>Spectroscopy provides an efficient way of predicting foliar functional traits at both leaf and canopy scales. Previous studies mainly used hyperspectral data from airborne and spaceborne platforms, and few studies assessed the feasibility of the more cost-effective and flexible unmanned airborne vehicle (UAV) −based imaging spectroscopy. Moreover, the research was mainly conducted in temperate and tropical biome, with less attention to subtropical evergreen forests. This study aims to characterize key foliar functional traits of subtropical evergreen forests in South China using leaf and UAV-based spectroscopy data. Dry and fresh leaf spectra as well as key foliar functional traits (leaf chlorophyll and carotenoids, water, dry matter, nitrogen, phosphorus, and potassium) were collected from 360 samples over 97 species across four field sites. For the site Heshan, we also collected UAV-based hyperspectral imageries at two dates. Partial least squares regression (PLSR) and random forest regression (RF) were used to link foliar traits with dry leaf spectra, fresh leaf spectra and canopy spectra, respectively. Cross-site and cross-season validations were performed to assess model transferability. Results showed that dry and fresh leaf spectra could estimate most foliar traits with comparable accuracies (normalized RMSE (NRMSE) &lt; 20 %). Leaf trait models can be transferred across sites and species. Canopy models could provide accurate estimations of pigments, N_mass, N_area, C/N and N/P (NRMSE = 12.92 – 21.02 %), but had low transferability across seasons due to phenological differences in both foliar traits and canopy spectra. PLSR outperformed RF in predicting foliar traits with leaf spectra, whereas RF achieved better estimations at the canopy scale. Trait maps from PLSR and RF were consistent, and both can capture the seasonal variation of foliar traits. We conclude that <em>in-situ</em> measurements covering the variability of foliar functional traits and spectra are needed to develop robust and transferable models. Our study provides a template for characterizing key foliar functional traits in subtropical evergreen forests and can promote our understanding of ecosystem functioning and processes.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"233 ","pages":"Article 110178"},"PeriodicalIF":7.7,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143520477","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}