基于机器视觉的青花椒病虫害智能诊断研究  

Intelligent diagnosis of green pepper pests and diseases based on machine vision

作　　者：程晗 CHENG Han(Chongqing Vocational College of Transportation,Chongqing 402247,China)

机构地区：[1]重庆交通职业学院,重庆402247

出　　处：《自动化与仪器仪表》2025年第1期328-332,337,共6页Automation & Instrumentation

基　　金：重庆市教委科学技术研究计划项目《基于机器视觉的江津区青花椒病虫害智能诊断研究》(KJQN202305710)。

摘　　要：中国农业产业结构正处于转型发展关键时期,加快农业现代化进程意义重大。为了进一步提升农业发展的现代化程度,研究基于机器视觉技术,使用深度学习框架YOLOv5(You Only Look Once v5)与迁移学习技术设计了农作物病虫害的自动识别与诊断模型。实验结果表明,研究改进的YOLOv5s模型识别准确率高达96.04%,损失值为0.026。消融实验验证了不同改进策略的有效性,模型交并比取值0.94,检测速度达120.39帧/秒、F1值为0.92。相比于现有先进模型,研究设计的方法在3种不同误差指标上具有一定优势,平均精度均值达到了92.37%,高于其他模型的85.2%和79.3%。此次研究的设计成果可实现病虫害的自动诊断,有助于实现农业病虫灾害的早期预警,为更多农产品的病虫灾害防治措施提供了技术支持。China’s agricultural industrial structure is in a critical period of transformation and development,and it is of great significance to accelerate the process of agricultural modernization.In order to further enhance the modernization of agricultural development,the study designed an automatic identification and diagnosis model of crop pests and diseases based on machine vision technology using the deep learning framework YOLOv5(You Only Look Once v5)and transfer learning technology.The experimental results show that the research improved YOLOv5s model recognition accuracy is up to 96.04%with a loss value of 0.026.The ablation experiment verifies the effectiveness of the different improvement strategies,with the model intersection and merger ratio taking the value of 0.94,and the detection speed reaches 120.39 frames/sec with an F1 value of 0.92.Compared with the existing advanced models,the research designed method has some advantages,with an average accuracy mean value of 92.37%,which is higher than the other models of 85.2%and 79.3%.The designed results of this research can realize the automatic diagnosis of pests and diseases,which can help to realize the early warning of agricultural pest and disease disasters,and provide technical support for pest and disease disaster control measures for more agricultural products.

关 键 词：机器视觉 YOLOv5 迁移学习 病虫害 智能诊断 

分 类 号：S435[农业科学—农业昆虫与害虫防治]