基于改进YOLOv8的库尔勒香梨病虫害监测研究  

Research on insect pests and diseases monitoring of Korla fragrant pear based on improved YOLOv8

作　　者：曹冰玉 周鹏 阚明琪 周美艳 Cao Bingyu;Zhou Peng;Kan Mingqi;Zhou Meiyan(College of Information Science and Engineering,Xinjiang University of Science and Technology,Korla,841000,Xinjiang,China)

机构地区：[1]新疆科技学院信息科学与工程学院,新疆库尔勒841000

出　　处：《新疆农机化》2025年第1期14-17,共4页Xinjiang Agricultural Mechanization

基　　金：新疆科技学院2024年产教融合与新商科发展研究中心第二批招标项目(2024-KYJD05);国家级大学生创新创业项目(202413561002)。

摘　　要：库尔勒香梨的病虫害,如黑星病、黑斑病、腐烂病和梨小食心虫等,对香梨的产量和品质构成了严重威胁。为了提高香梨病虫害检测的准确性,有效防治上述病虫害,本研究开发了一种基于改进YOLOv8的轻量化库尔勒香梨病虫害检测模型,并构建了包含多种环境条件下库尔勒香梨病虫害的数据集。本文将YOLOv8的卷积模块替换为GSConv,并引入VoV-GSCSP模块来优化网络结构,形成了高效的Slim-neck架构,在保持高识别率的同时降低了模型的计算负担。该架构还集成了EMA多尺度注意力机制,以增强模型对库尔勒香梨病虫害特征的识别能力。试验结果表明,改进后的YOLOv8模型在库尔勒香梨病虫害检测中达到了87.9%的精度均值(mAP),相较于原始模型有所提升,模型体积缩减至5.5MB。该模型的精准识别能力可为库尔勒香梨病虫害检测提供了技术支持。Insect pests and diseases of Korla fragrant pear,such as black star disease,black blotch disease,rot disease,and Grapholita molesta,pose serious threats to the yield and quality of the fruit.To enhance the accuracy of insect pests and diseases detection and effectively control these issues,this study developed a lightweight insect pests and diseases detection model for Korla fragrant pear based on an improved YOLOv8.A dataset encompassing Korla fragrant pear insect pests and diseases under various environmental conditions was also constructed.In this research,the convolution module of YOLOv8 was replaced with GSConv,and the VoV-GSCSP module was introduced to optimize the network structure,forming an efficient Slim-neck architecture.This optimization maintained a high recognition rate while reducing the computational burden of model.Additionally,the architecture integrated the EMA multi-scale attention mechanism to enhance the ability of model to recognize features of Korla fragrant pear insect pests and diseases.Experimental results demonstrate that the improved YOLOv8 model achieved an average precision(mAP)of 87.9%in detecting insect pests and diseases in Korla fragrant pear,which was improved compared with the original model.The model size was reduced to 5.5 MB.The precise identification capability of this model can provide robust technical support for the detection of pests and diseases in Korla fragrant pear.

关 键 词：病虫害检测 YOLOv8 深度学习 库尔勒香梨 轻量化 

分 类 号：S24[农业科学—农业电气化与自动化] TP2[农业科学—农业工程]