智慧农业（中英文）2025，Vol.7Issue(1)：146-155,10.DOI:10.12133/j.smartag.SA202410023
基于改进YOLOv10n的轻量化番茄叶片病虫害检测方法
Lightweight Tomato Leaf Disease and Pest Detection Method Based on Improved YOLOv10n
摘要
Abstract
[Objective]To address the challenges in detecting tomato leaf diseases and pests,such as complex environments,small goals,low pre-cision,redundant parameters,and high computational complexity,a novel lightweight,high-precision,real-time detection model was proposed called YOLOv10n-YS.This model aims to accurately identify diseases and pests,thereby providing a solid scientific basis for their prevention and management strategies. [Methods]The dataset was collected using mobile phones to capture images from multiple angles under natural conditions,ensuring complete and clear leaf images.It included various weather conditions and covered nine types:Early blight,leaf mold,mosaic virus,septoria,spider mites damage,yellow leaf curl virus,late blight,leaf miner disease,and healthy leaves,with all images having a reso-lution of 640×640 pixels.In the proposed YOLOv10n-YS model,firstly,the C2f in the backbone network was replaced with C2f_RepViTBlock,thereby reducing the computational load and parameter volume and achieving a lightweight design.Secondly,through the introduction of a sliced operation SimAM attention mechanism,the Conv_SWS module was formed,which enhanced the extraction of small target features.Additionally,the DySample lightweight dynamic up sampling module was used to replace the up sampling module in the neck network,concentrating sampling points on target areas and ignoring backgrounds,thereby effectively identifying defects.Finally,the efficient channel attention(ECA)was improved by performing average pooling and max pooling on the input layer to aggregate features and then adding them together,which further enhanced global perspective information and fea-tures of different scales.The improved module,known as efficient channel attention with cross-channel interaction(EMCA)attention,was introduced,and the pyramid spatial attention(PSA)in the backbone network was replaced with the EMCA attention mechanism,thereby enhancing the feature extraction capability of the backbone network. [Results and Discussions]After introducing the C2f_RepViTBlock,the model's parameter volume and computational load were reduced by 12.3%and 9.7%,respectively,with mAP@0.5 and F1-Score each increased by 0.2%and 0.3%.Following the addition of the Conv_SWS and the replacement of the original convolution,mAP@0.5 and F1-Score were increased by 1.2%and 2%,respectively,in-dicating that the Conv_SWS module significantly enhanced the model's ability to extract small target features.After the introduction of DySample,mAP@0.5 and F1-Score were increased by 1.8%and 2.6%,respectively,but with a slight increase in parameter volume and computational load.Finally,the addition of the EMCA attention mechanism further enhanced the feature extraction capability of the backbone network.Through these four improvements,the YOLOv10n-YS model was formed.Compared with the YOLOv10n al-gorithm,YOLOv10n-YS reduced parameter volume and computational load by 13.8%and 8.5%,respectively,with both mAP@0.5 and F1-Score increased.These improvements not only reduced algorithm complexity but also enhanced detection accuracy,making it more suitable for industrial real-time detection.The detection accuracy of tomato diseases and pests using the YOLOv10n-YS algo-rithm was significantly better than that of comparative algorithms,and it had the lowest model parameter volume and computational load.The visualization results of detection by different models showed that the YOLOv10n-YS network could provide technical sup-port for the detection and identification of tomato leaf diseases and pests.To verify the performance and robustness of the YOLOv10n-YS algorithm,comparative experiments were conducted on the public Plant-Village-9 dataset with different algorithms.The results showed that the average detection accuracy of YOLOv10n-YS on the Plant-Village dataset reached 91.1%,significantly higher than other algorithms. [Conclusions]The YOLOv10n-YS algorithm is not only characterized by occupying a small amount of space but also by possessing high recognition accuracy.On the tomato leaf dataset,excellent performance was demonstrated by this algorithm,thereby verifying its broad applicability and showcasing its potential to play an important role in large-scale crop pest and disease detection applications.Deploying the model on drone platforms and utilizing multispectral imaging technology can achieve real-time detection and precise lo-calization of pests and diseases in complex field environments.关键词
番茄叶片/病虫害检测/YOLOv10n/注意力机制/轻量化Key words
tomato leaves/pest detection/YOLOv10n/attention mechanism/lightweight分类
植物保护学引用本文复制引用
吴六爱,许雪珂..基于改进YOLOv10n的轻量化番茄叶片病虫害检测方法[J].智慧农业（中英文）,2025,7(1):146-155,10.基金项目
国家自然科学基金项目(51567014) （51567014）
甘肃省科技计划项目(22JR5RA797) National Natural Science Foundation of China(51567014) （22JR5RA797）
Gansu Science and Technology Plan Project(22JR5RA797) （22JR5RA797）
