首页|期刊导航|安庆师范大学学报(自然科学版)|基于优化YOLOv5s的苹果叶片病害检测

基于优化YOLOv5s的苹果叶片病害检测OA

Apple Leaf Disease Detection Based on Optimized YOLOv5s

中文摘要英文摘要

植物叶片病虫害检测能够及早发现病虫害发生情况,减少作物损失,保障农业生产稳定进行.人工肉眼观察的方法存在着检测效率不高、耗时长等问题,文章采用了一种改进的YOLOv5s(you only look once v5 small)算法对苹果叶片的病虫害进行智能检测.首先引用Focal-EIOU(focal efficient intersection over union)损失函数代替原网络模型中的CIOU(complete intersection over union)损失函数,从而优化边界框收敛效果,减少漏检,提升模型召回率.同时,对激活函数进行改进,使用Mish激活函数来替换原始网络模型中的SiLU激活函数,缓解深层网络梯度消失.最后,在公共数据集plantviallage上开展苹果叶片病虫害检测实验以验证所提算法的功能性.实验结果表明,改进后的YOLOv5s模型召回率和平均精度分别达89.3%和93.6%,相比原始的YOLOv5s模型分别提高了1.2%和1.5%,在确保实时检测的同时提高了苹果叶片病害目标的检测精度,对叶片病虫害的预防具有重要的意义.

Plant leaf pest and disease detection can identify the occurrence of pests and diseases at an early stage,reduce crop losses,and ensure the stable progress agricultural production.The manual visual observation method has problems such as low detection efficiency and longtime consumption.This paper proposes an improved YOLOv5s algorithm for the intelli-gent detection of pests and diseases in apple leaves.First,the Focal-EIOU(focal efficient intersection over union,Focal-EIOU)loss function is introduced to replace the CIOU(complete intersection over union,CIOU)loss function in the original network model,thereby optimizing the bounding box convergence effect,reducing missed detections,and improving the mod-el recall rate.At the same time,the activation function is improved by using the Mish activation function to replace the SiLU activation function in the original network model to alleviate the gradient vanishing problem in deep networks.Finally,an ap-ple leaf pest and disease detection experiment is conducted on the public dataset Plant village to verify the functionality of the proposed algorithm.The experimental results show that the recall rate and average precision of the improved YOLOv5s(you only look once v5 small,YOLOv5s)model reached 89.3%and 93.6 respectively.Compared with the original YOLOv5s mod-el,the recall rate and average precision were increased by 1.2%and 1.5%,respectively.The improved v5s model improves the detection accuracy of apple leaf disease targets while ensuring real-time detection,which is of great significance for the pre-vention of leaf pests and diseases.

石冰;周涵;王远志

安庆师范大学 计算机与信息学院,安徽 安庆 246133安庆师范大学 计算机与信息学院,安徽 安庆 246133安庆师范大学 计算机与信息学院,安徽 安庆 246133

信息技术与安全科学

深度学习目标检测YOLOv5sEIOUMish激活函数

deep learningobject detectionYOLOv5sEIOUMish activation function

《安庆师范大学学报(自然科学版)》 2026 (2)

57-62,6

国家重点研发计划项目(SQ2020YFF0402315)

10.13757/j.cnki.cn34-1328/n.2026.02.008

评论