基于改进YOLOv4的夏玉米主要害虫检测方法研究  

Detection Method of Major Corn Pests Based on Improved YOLOv4 Model

作　　者：段新涛[1] 王伸 赵晴 张杰 郑国清 李国强 Duan Xintao;Wang Shen;Zhao Qing;Zhang Jie;Zheng Guoqing;Li Guoqiang(College of Computer and Information Engineering,Henan Normal University,Xinxiang 453007,China;Institute of Agricultural Economy and Information,Henan Academy of Agricultural Sciences/Henan Engineering and Technology Research Center for Intelligent Agriculture,Zhengzhou 450002,China;Key Laboratory of Huang-Huai-Hai Intelligent Agricultural Technology,Ministry of Agriculture and Rural Affairs,Zhengzhou 450002,China)

机构地区：[1]河南师范大学计算机与信息工程学院,河南新乡453007 [2]河南省农业科学院农业经济与信息研究所/河南省智慧农业工程技术研究中心,河南郑州450002 [3]农业农村部黄淮海智慧农业技术重点实验室,河南郑州450002

出　　处：《山东农业科学》2023年第10期167-173,共7页Shandong Agricultural Sciences

基　　金：河南省农业科学院自主创新项目(2023ZC067);河南省农业科学院科技创新团队项目(2022TD14);河南省重点研发与推广专项(科技攻关)项目(212102110255,172102110092)。

摘　　要：为提高夏玉米主要害虫的检测精度,实现害虫的自动化测报,本研究基于性诱测报原理,设计玉米主要害虫诱集装置,并利用该装置自动采集黏虫、棉铃虫、玉米螟、甜菜夜蛾等玉米害虫图像,制作VOC数据集。以YOLOv4模型为基础,引入SENet模块和Soft-NMS算法,构建YOLOv4-Corn模型,解决玉米害虫体积小、虫体易重叠等造成的不易识别问题。结果表明,YOLOv4-Corn模型对黏虫、棉铃虫、玉米螟和甜菜夜蛾的平均检测精度分别为95.89%、96.59%、93.34%和99.07%;与Faster R-CNN、YOLOv3、YOLOv4等模型相比,YOLOv4-Corn的F1值、召回率、精确率、平均精度均最优。可见,YOLOv4-Corn对黏虫、棉铃虫、玉米螟和甜菜夜蛾具有较高的识别准确率,可用于田间夏玉米害虫种群监测预警。Corn pest detection is essential for field management.However,the adhesion and obscuration of pest bodies limit the detection accuracy.This study aimed to improve the detection precision of major corn pests,such as armyworms,cotton bollworms,corn bores and beet armyworms,and realize the automatical de-tection and forcast.A trap device was designed to automatically collect corn pest images.An improved YOLOv4-Corn algorithm was proposed based on the YOLOv4 model by adding the SENet module and Soft-NMS algorithm to improve the detection of small targets.The results showed that the average detection precision of the YOLOv4-Corn model was 95.89%for armyworms,96.59%for cotton bollworms,93.34%for corn borers,and 99.07%for beet armyworms.Compared with the Faster R-CNN,YOLOv3 and YOLOv4 models,the YOLOv4-Corn model had the highest F1 value,recall rate,accuracy and average precision.Therefore,YOLOv4-Corn could accurately detect armyworm,cotton bollworm,corn borer and beet armyworm,so it could be applied to monitor and predict summer corn pest populations.

关 键 词：玉米害虫 诱集装置 YOLOv4-Corn 目标检测 

分 类 号：S126[农业科学—农业基础科学]