基于改进PP-YOLO的农业病虫害识别算法的研究  

Research on the Identification Algorithm of Agricultural Pests and Diseases Based on Improved PP-YOLO

作　　者：田斌 顾斌 费晨 王晓拓 孙彦 倪成功 肖俊淳 TIAN Bin;GU Bin;FEI Chen;WANG Xiaotuo;SUN Yan;NI Chengong;XIAO Junchun(Suzhou Polytechnic Institute of Agriculture,Suzhou 215200,China;Suzhou Wanlian Magnetic Induction Communication Technology Co.,Ltd,Suzhou 215200,China)

机构地区：[1]苏州农业职业技术学院,江苏苏州215200 [2]苏州万联磁感应通讯科技有限公司,江苏苏州215200

出　　处：《农机使用与维修》2025年第3期74-77,共4页Agricultural Machinery Using & Maintenance

基　　金：苏州农业职业技术学院青年教师科研能力提升计划(QN[2022]10);苏州农业职业技术学院教师社会实践项目(无编号);2023年江苏省创新创业培育计划项目(GX019)。

摘　　要：在农业生产活动当中,病虫害对粮食的产量产生了很大的影响。在无人机中利用计算机视觉精准识别病虫害为农业生产提供了保障,但是同时也带来了目标太小、精准度不高的问题。针对此问题,通过在PP-YOLO网络的基础上引入注意力机制SENet(Squeeze-and-Excitation Networks),增强其主干网络的特征提取能力,提高小目标番茄叶斑病的识别准确率。经过试验表明,该文所提出的方法在PlantVillage中番茄叶斑病数据集上的检测性能优于PP-YOLO,平均精度达到86.44%,具有一定的应用价值。In agricultural production activities,pests and diseases have a great impact on food production.The use of computer vision in drones to accurately identify pests and diseases provides a guarantee for agricultural production,but it also brings the problem that the target is too small and the accuracy is not high.In order to solve this problem,the attention mechanism SENet(Squeeze-and-Excitation Networks)was introduced on the basis of PP-YOLO network to enhance the feature extraction ability of the backbone network and improve the identification accuracy of small target tomato spot disease.Experiments show that the detection performance of the proposed method on the tomato leaf spot dataset in PlantVillage is better than that of PP-YOLO,with an average accuracy of 86.44%,which has certain application value.

关 键 词：农业病虫害识别 改进PP-YOLO 深度学习 计算机视觉 目标检测 

分 类 号：S436.412[农业科学—农业昆虫与害虫防治] TP391.41[农业科学—植物保护]