YOLOv5在农田害虫图像识别中的应用  

Image Recognition of Farmland Pests Based on YOLOv5

作　　者：许星宇 谢维奇 XU Xingyu;XIE Weiqi(School of Computer Engineering,Jinling Institute of Technology,Nanjing,China,211169)

机构地区：[1]金陵科技学院计算机工程学院,南京211169

出　　处：《福建电脑》2022年第10期7-11,共5页Journal of Fujian Computer

基　　金：教育部产学合作协同育人项目(No.202101110006);金陵科技学院高层次人才工作启动费资助项目(No.Jit-rcyj-201512)资助;金陵科技学院教育教改研究课题(No.JYJG202124)资助。

摘　　要：为了对农田病虫害进行有效的预防和控制,需要进行虫情分析。由于农田害虫的多样性和复杂性,通过人工观察统计的传统害虫监测方式已经难以满足现代大规模农业生产对虫害预防工作的需要。由计算机进行识别的目标检测算法恰好可以解决这一问题。随着目标检测的发展,以卷积神经网络为核心算法的深度学习由于对图片计算效率大幅提升,计算成本也大幅降低,被广泛运用在目标检测识别中。本文通过分析前人对基于深度学习的目标检测算法的研究,综合考虑、选取YOLO v5模型,对虫情测报灯所所采集到的对28种害虫进行识别,以便减少农药的滥用和误用,改善农田生态环境。最终识别准确度达到了99.3%。In order to effectively prevent and control farmland diseases and insect pests, it is necessary to conduct pest analysis. Due to the diversity and complexity of farmland pests, the traditional pest monitoring method through manual observation and statistics has been difficult to meet the needs of modern large-scale agricultural production for pest prevention work. Object detection algorithms that are identified by a computer do just that.With the development of target detection, the target detection algorithm based on deep learning with convolutional neural network as the core algorithm is widely used in target detection and recognition due to the significant improvement in the computational efficiency of images and the reduction in computational cost. In this paper, through the analysis of previous research on the target detection algorithm based on deep learning,and comprehensive consideration, the YOLOv5 model is selected to identify 28 kinds of pests collected by the insect forecast light, so as to reduce the abuse and misuse of pesticides. Improve farmland ecological environment. The final recognition accuracy reached 99.3%.

关 键 词：害虫识别 目标检测算法 YOLOv5 卷积神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]