Computers and Electronics in Agriculture最新文献

{"title":"Analysis of energy transfer characteristics in multi-level branches of Lycium barbarum L. under vibration excitation","authors":"Bo Deng ,&nbsp;Xiaoyan Wang ,&nbsp;Kunyu Wang ,&nbsp;Hao Song ,&nbsp;Shijie Gao ,&nbsp;Yuhang Tu ,&nbsp;Xuemin Zhang","doi":"10.1016/j.compag.2025.110992","DOIUrl":"10.1016/j.compag.2025.110992","url":null,"abstract":"<div><div>Aiming at the problems of unknown vibration energy transfer mechanism and high risk of branch damage in the process of mechanised harvesting of <em>Lycium barbarum L. (L. barbarum)</em>, this study took the typical secondary bifurcation structure of <em>L. barbarum</em> branches as the object, and analysed the influence law of the branch length ratio, diameter ratio and clip angle on the efficiency of energy transfer through the establishment of a mathematical model and the combination of MATLAB simulation. A self-made vibration excitation loading device, combined with flexible acceleration sensors and high-speed camera device, was used to measure the three-dimensional vibration response accurately. The results show that: the experimental results are basically consistent with the simulation results, and the square of the primary-secondary branch diameter ratio and the length ratio are positively correlated with the energy transfer efficiency; based on the modified mathematical model of the inter-branch angle and acceleration ratio, it is known that the inter-branch angle in the range of 0 ∼ 90° has a significant nonlinear inhibitory effect on the energy transfer efficiency; and the acceleration value of the tertiary branches in the range of 765.36 ∼ 1224.58 m/s² were achieved for effective harvesting; in vibration harvesting operation, vibration of tertiary branches compared to vibration of the primary trunk, the excitation force can be reduced by 48.16 % ∼ 79.75 %, effectively avoiding the risk of damage to the branches of <em>L. barbarum</em>. The results of the study provide a theoretical basis for the structural design of low-damage vibration harvesting equipment for <em>L. barbarum</em>.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"239 ","pages":"Article 110992"},"PeriodicalIF":8.9,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158229","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":"Inversion of nitrogen concentration in crop leaves based on improved radiative transfer model","authors":"Shuang Xiang ,&nbsp;Shikuan Wang ,&nbsp;Zhonghui Guo ,&nbsp;Nan Wang ,&nbsp;Zhongyu Jin ,&nbsp;Fenghua Yu ,&nbsp;Tongyu Xu","doi":"10.1016/j.compag.2025.111017","DOIUrl":"10.1016/j.compag.2025.111017","url":null,"abstract":"<div><div>The timely and accurate prediction of nitrogen status within crops can provide certain data support for precision fertilization. However, few studies have considered using radiative transfer models to estimate the nitrogen concentration (<span><math><mrow><mi>C</mi><mi>n</mi></mrow></math></span>) of crops. This study is based on the PIOSL-5 model to generate a large number of simulation datasets, namely the PIOSLSD dataset. The successive projections algorithm (SPA) is used to select nitrogen-related features. A crop <span><math><mrow><mi>C</mi><mi>n</mi></mrow></math></span> inversion model based on the PIOSL-5 model is constructed using five models: Extreme Learning Machine, Genetic Algorithm Optimized Extreme Learning Machine, Particle Swarm Optimization Optimized Extreme Learning Machine, Third Generation Non Dominated Genetic Algorithm Optimized Extreme Learning Machine (NSGA-III-ELM), and Bat Algorithm Optimized Extreme Learning Machine. The model is compared with traditional data-driven methods and the accuracy of the model is verified using three datasets: RICE23, LOPEX93, and CALIFORNIA. The results showed that the nitrogen characteristic bands of the PIOSLSD dataset filtered by SPA were 1070, 1150, 1405, 1535, and 1725 nm. The Cn prediction based on the NSGA-III-ELM model, which uses these 5 feature bands as inputs, has the best performance. The determination coefficients of the validation set are 0.814, 0.785, and 0.792, respectively. The <span><math><mrow><mi>C</mi><mi>n</mi></mrow></math></span> inversion model based on the PIOSL-5 model constructed in this article achieves remote sensing prediction of crop nitrogen mechanism model, which has certain mechanistic significance for nitrogen nutrition management of crops and improving nitrogen utilization efficiency.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"239 ","pages":"Article 111017"},"PeriodicalIF":8.9,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158233","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":"Discrete element simulation and experimental verification: Effect of spray position in rotary tillage on the spatial redistribution of liquid soil amendments","authors":"Zhengyang Wu ,&nbsp;Hongwen Li ,&nbsp;Jin He ,&nbsp;Xu Zhang ,&nbsp;Caiyun Lu ,&nbsp;Chao Wang ,&nbsp;Dejian Zhang ,&nbsp;Shan Jiang ,&nbsp;Hongdao Shan ,&nbsp;Rongrong Li ,&nbsp;Zongfu Yang","doi":"10.1016/j.compag.2025.111021","DOIUrl":"10.1016/j.compag.2025.111021","url":null,"abstract":"<div><div>Rotary tillage is a common cultivation for mixing cultivated soils with various dopants, including liquid amendments. The mixing performance of rotary tillage should be understood quantitatively. This study aimed to further validate the added value of the discrete element method (DEM) for spray-location selection in rotary tillage with the application of liquid amendments. A normalized amendment mixing index (AMI) was defined to describe the mixing of liquid amendments with soils. The AMI was used to quantify the mixing situations of the horizontally sliced subspace (HSS) and the vertically sliced subspace (VSS). The effect of spray position on the AMI of the slices was statistically analyzed. A field experiment was conducted using a spray position configuration that yielded the highest AMI in the simulation. The experimental AMIs were captured by processing images of vertical soil profiles. Simulated results show that spray position significantly affects the AMI, and spraying in the front of the tillage obtained the highest AMI. The experimental average AMI of VSS had an error of 7.08 % related to the simulation. Statistical analysis showed no significant difference between the simulation and experimental results. These results indicate that the AMI can distinguish between soil spaces containing only impregnated components and those containing only unimpregnated components, and can quantitatively describe the mixing situation in experiments and simulations. DEM simulation can provide reliable insights on spray-location selection to apply amendments with a rotary tiller. These are expected to support the DEM simulation of solid–liquid mixing to investigate the mixing situation.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"239 ","pages":"Article 111021"},"PeriodicalIF":8.9,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158232","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":"Effectiveness of image-based classifiers for on-site identification of lichens in reserved forest","authors":"Karthikumar Sankar ,&nbsp;M. Chengathir Selvi ,&nbsp;R. Shyam Kumar ,&nbsp;Ponmurugan Karuppiah ,&nbsp;Govindasami Periyasami ,&nbsp;Perumal Karthikeyan ,&nbsp;E. Basil Tamil Selvan","doi":"10.1016/j.compag.2025.110994","DOIUrl":"10.1016/j.compag.2025.110994","url":null,"abstract":"<div><div>The collection and identification of lichen samples within the reserved forest pose significant challenges, leading researchers to seek alternative methodologies. Manual identification of lichens requires microscopic and chemical analysis, which is time-consuming and requires domain expert knowledge. Deep learning models can automatically learn complex lichen features from images, thereby avoiding manual feature extraction. In this context, an in-depth study was conducted using two approaches to evaluate the efficacy of several network models for the identification of lichens in digital images. In the initial method, a total of 119 lichen images, comprising 15 images from experimental work and additional images from open sources, were employed for the traditional classification approach utilizing MATLAB. In which, 90 different features were extracted from various color models such as RGB, YCbCr, HSV, CMYK, LAB, YIQ and subjected for classification in ANN model, which resulted 76.6 % accuracy for the class of lichen family. With the limited original images, the transfer learning enables effective model training and makes the process faster and scalable. In this approach, each species of lichen images was augmented using generative augmentation tool and subjected for multiclass deep learning network models, precisely transfer learning model. In total, 10,410 images comprising 5 thallus type, 8 order and 77 distinct lichen species were used to train three different network models (VGG16, VGG19, Res.Net50). It was found that Res.Net50 model exhibited highest sensitivity (&gt;0.99) and precision. The class lichen thallus type achieved a maximum accuracy of 86.6 %, while the order exhibited 80 %. Overall, the lichen species under the order of Laccanorals with fruticose thallus type showed highest classification accuracy in all the three models attempted in this study. The study concluded that utilizing transfer learning with a multilayer perceptron model proved to be more effective and accurate for the lichen dataset, especially when dealing with greater complexity in picture information. The obtained results were comparable to those of other recent deep learning approaches.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"239 ","pages":"Article 110994"},"PeriodicalIF":8.9,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158069","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":"A variable-rate spraying system for vineyards based on RGB-D imaging and tensor acceleration","authors":"Qi Gao ,&nbsp;Alberto Carraro ,&nbsp;Qiang Huang ,&nbsp;Francesco Marinello ,&nbsp;Marco Sozzi","doi":"10.1016/j.compag.2025.110960","DOIUrl":"10.1016/j.compag.2025.110960","url":null,"abstract":"<div><div>Sustainable vineyard management requires precise and efficient application of plant protection products to minimise environmental impact while ensuring plant health. This study presents a variable-rate spraying system that integrates an RGB-D camera with a GPU-equipped edge computing platform to enable accurate, real-time adjustment of spray flow rates in vineyards. A tensor-based representation of RGB-D data is employed to accelerate the entire processing pipeline. Based on this structure, a fast approximate meshing method is applied to rapidly generate 3D meshes from point clouds. To incorporate semantic information from RGB images, an instance segmentation model is used to detect grapevine canopies and trellis posts. The resulting canopy masks are used to isolate the canopy meshes, while the trellis posts serve as reference planes for canopy volume estimation via mesh projection. Based on the computed volume, pulse-width modulation signals are generated to dynamically control spray flow rates. Field experiments were conducted to evaluate the system’s effectiveness and real-time performance. The results demonstrated that the estimated canopy volume is a reliable indicator for regulating application rates. Compared to uniform-rate spraying, the proposed system reduced plant protection product consumption by 57.4% while ensuring adequate droplet coverage. Additionally, the system demonstrated satisfactory real-time performance even on entry-level hardware. Overall, the proposed variable-rate spraying system offers an accurate, real-time, and cost-effective solution for precision viticulture, highlighting its potential for commercial deployment in sustainable vineyard management.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"239 ","pages":"Article 110960"},"PeriodicalIF":8.9,"publicationDate":"2025-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158126","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":"A self-learning based explainable AI framework for giant crab sex classification","authors":"Yanyu Chen ,&nbsp;Wenli Yang ,&nbsp;Scott Hadley ,&nbsp;Rafael León Leiva ,&nbsp;Quan Bai ,&nbsp;Byeong Ho Kang","doi":"10.1016/j.compag.2025.110989","DOIUrl":"10.1016/j.compag.2025.110989","url":null,"abstract":"<div><div>Sexual dimorphism is prevalent in crustaceans, and in crabs, it can be observed in body size, carapace shape, and larger and differently shaped claws in males, among other traits. With the advancement of artificial intelligence (AI), automatic sex classification is now possible by leveraging these differential traits between males and females. Traditionally, most research has relied on analysing the entire crab to classify its sex. Subsequently, studies have focused on specific features, such as the abdominal flap or carapace, which facilitate sex differentiation by humans. However, human intervention creates a significant labelling burden and constrains the exploration of other potentially valuable characteristics for research. In this study, a novel framework is introduced for automatic sex classification in crabs. Unlike traditional approaches that rely heavily on manual labelling and predefined features such as abdominal flaps, our framework minimizes human intervention and enables the model to autonomously highlight image regions most informative for classification. This draws attention to underutilized morphological regions that may be useful for sex classification without requiring prior biological knowledge. Furthermore, the framework provides visual explanations of the model’s decisions, enhancing interpretability. Using this approach, we achieved a classification accuracy of 95.4%, while also pinpointing specific regions contributing to the decision-making process.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"239 ","pages":"Article 110989"},"PeriodicalIF":8.9,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145118617","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":"3D skeletonization and phenotyping for soybean root system architecture using a bio-inspired algorithm","authors":"Xuehai Zhou ,&nbsp;Tianzi Yang ,&nbsp;Rui Xu ,&nbsp;Alexander Bucksch ,&nbsp;Pierre Dutilleul ,&nbsp;Davoud Torkamaneh ,&nbsp;Shangpeng Sun","doi":"10.1016/j.compag.2025.110890","DOIUrl":"10.1016/j.compag.2025.110890","url":null,"abstract":"<div><div>Characterizing root system architecture (RSA) is essential for understanding plant acclimatization and guiding breeding strategies to enhance stress tolerance and optimize resource uptake. Although 3D root analysis provides significantly more detailed and structurally informative insights than conventional 2D methods, the development of robust and quantitative tools for 3D root phenotyping has been hindered by challenges such as data complexity, noise, and root overlap. In this study, we present a biologically inspired skeletonization framework that segments root architectures by tracing root growth trajectories. The primary objective is to enable anatomically accurate extraction of RSA traits from 3D point clouds. Our method begins by segmenting the primary root through shortest-path extraction and tangent-plane-based clustering. Lateral root initiation points are then detected, and candidate paths are grown using a bionic pathfinding strategy with adaptive parameters; an optimal, non-overlapping skeleton is selected through clustering and combination sorting, and finally refined via an inward back-tracing procedure to improve junction connectivity. To support downstream phenotyping, we compute root length and angle from the segmented skeletons, and reconstruct anatomically faithful tubular meshes for each lateral root to analytically estimate surface area and volume. Our method achieved high accuracy across multiple traits, including an <span><math><msub><mrow><mi>F</mi></mrow><mrow><mn>1</mn></mrow></msub></math></span> score of 0.88 for lateral root numeration, <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> values of 0.992 and 0.987 for primary and lateral root length estimation, respectively, and strong agreement in surface area (<span><math><mrow><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup><mo>=</mo><mn>0</mn><mo>.</mo><mn>953</mn></mrow></math></span>) and volume (<span><math><mrow><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup><mo>=</mo><mn>0</mn><mo>.</mo><mn>912</mn></mrow></math></span>) validation against reference methods. Overall, our method offers a robust and biologically meaningful solution for 3D root phenotyping. The extracted traits provide plant breeders with critical insights for genotype selection and offer plant scientists a powerful tool to evaluate the effects of agronomic treatments and environmental interventions.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"239 ","pages":"Article 110890"},"PeriodicalIF":8.9,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145118688","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":"A novel path planning approach for plant protection UAV based on DDPG and ILA optimization algorithm","authors":"Pei Wang ,&nbsp;Peixin He ,&nbsp;Chenyuhao Ma ,&nbsp;Chenxi Niu ,&nbsp;Huayu Gao ,&nbsp;Hongmei Wang ,&nbsp;S.M. Muyeen ,&nbsp;Daming Zhou","doi":"10.1016/j.compag.2025.111006","DOIUrl":"10.1016/j.compag.2025.111006","url":null,"abstract":"<div><div>With the advancement of agricultural modernization, plant protection UAVs play an increasingly crucial role. However, traditional path planning methods struggle to meet the demands of irregular work areas and autonomous obstacle avoidance during transfers. This research introduces a novel dual-layer planning architecture, which innovatively proposes the synergistic combination of artificial intelligence and optimization algorithms in this field. Specifically, the DDPG algorithm is applied to the path planning between farmlands. By constructing a virtual environment replete with random obstacles based on actual geographical data, the UAV learns the optimal response strategy. Minimizing flight path length and turning amplitude is the objective, and multiple reward mechanisms are devised to accelerate convergence, enabling real-time and efficient obstacle avoidance. For the spraying operations in irregular farmlands, the ILA optimization algorithm is utilized. A trajectory planning model considering the UAV’s heading is established, and optimization criteria are formulated. Through this algorithm, the optimal operation route is determined. Simulation results reveal that in the inter-farmland path planning, compared with the Particle Swarm Optimization Algorithm (PSO) and the Zebra Optimization Algorithm (ZOA), the method based on Deep Deterministic Policy Gradient (DDPG) generates paths with shorter lengths and requires less flight time, and demonstrates excellent adaptability to unknown and dynamic obstacles. In the intra-farmland path planning, the ILA optimization algorithm improves the rebroadcast rate and reduces the turn times by 6.4% and 7.7% respectively compared to the particle swarm optimization algorithm. Overall, the integration of DDPG and ILA optimization algorithms successfully addresses the global path planning challenges of plant protection UAVs in complex agricultural scenarios.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"239 ","pages":"Article 111006"},"PeriodicalIF":8.9,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145118619","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":"Pattern analysis of daily beehive weight variation for colony health assessment","authors":"Yih-Lin Liu,&nbsp;Ta-Te Lin","doi":"10.1016/j.compag.2025.111016","DOIUrl":"10.1016/j.compag.2025.111016","url":null,"abstract":"<div><div>Continuous monitoring of beehive conditions plays a crucial role in assessing colony health, improving productivity, and supporting effective beekeeping management. This study presents a data-driven framework for classifying daily hive weight variation patterns using a multi-sensor beehive monitoring system. Weight data were collected at 10-minute intervals and processed using piecewise linear regression to extract key features. Principal component analysis (PCA) and engineered features were used to reduce dimensionality and enhance interpretability. Six supervised learning models were evaluated using K-fold cross-validation. The Support Vector Classification (SVC) model achieved the highest performance in classifying daily weight variation patterns, with an F1-score of 0.905 using the nine most significant features, comprising seven weight points and two derived meaningful indicators − net daily weight change and harvested weight. The analysis revealed six common daily weight variation patterns, each reflecting distinct behavioral and environmental scenarios, such as active foraging, nectar scarcity, or post-feeding inactivity. Seasonal analyses further indicated clear relationships between these weight patterns and environmental conditions, confirming the method’s effectiveness in capturing colony behaviors and reflecting overall hive health. This study provides an effective and scalable approach for real-time hive health monitoring, contributing to the advancement of precision apiculture through the integration of smart sensing and machine learning.</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"239 ","pages":"Article 111016"},"PeriodicalIF":8.9,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145118618","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":"An image synthesis framework for enhanced salmon louse larvae (Lepeophtheirus Salmonis) detection in complex seawater conditions","authors":"Chao Zhang ,&nbsp;Lars Christian Gansel ,&nbsp;Marc Bracke ,&nbsp;Ricardo da Silva Torres","doi":"10.1016/j.compag.2025.110985","DOIUrl":"10.1016/j.compag.2025.110985","url":null,"abstract":"<div><div>Salmon lice (<em>Lepeophtheirus salmonis</em>) infections pose significant threats to farmed and wild salmon populations, making early detection during pre-infective and infective stages essential for sustainable management. Traditional methods, such as PCR, are costly and unsuitable for large-scale deployment, while existing machine-learning approaches are limited by the lack of annotated data from complex seawater environments and are largely restricted to simplified laboratory conditions. To address these challenges, this study presents an automated synthetic data generation method based on the Segment Anything Model (SAM), specifically designed to improve the detection of multi-stage salmon louse larvae in real-world and complex seawater environments. A total of 54 orthogonal experiments were conducted to optimize the factor configuration for synthetic data generation, resulting in a high-quality dataset comprising 120,864 images. To maximize the utility of synthetic data and enhance model performance, we evaluated YOLO series models and the transformer-based RT-DETR-L model. Our experiments revealed that pretraining on synthetic data before fine-tuning (or hybrid training) consistently improved model performance across multiple evaluation metrics. The YOLOv8n’s F1-score increased from 70.6% to 87.7%, a notable relative improvement of 24.22%. In real-world seawater tests, models trained with synthetic data improved recall by 11.0% to 87.0% compared to models trained exclusively with the original data and outperformed well-trained biologists, further validating the effectiveness and practicality of the synthetic approach. This study provides a scalable solution for monitoring salmon louse larvae in large-scale seawater environments, improving the welfare of farmed and wild salmon. The source code is publicly available at <span><span>https://github.com/jay-zc/synthetic-dataset</span><svg><path></path></svg></span> (as of July 2025).</div></div>","PeriodicalId":50627,"journal":{"name":"Computers and Electronics in Agriculture","volume":"239 ","pages":"Article 110985"},"PeriodicalIF":8.9,"publicationDate":"2025-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145118687","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}