基于YOLOv9葡萄病害识别检测算法研究  

Research on Grape Disease Identification and Detection Algorithm Based on YOLOv9

作　　者：萧峥嵘 梁烨锋 李菲 王义宗 田纪亚 XIAO Zhengrong;LIANG Yefeng;LI Fei;WANG Yizong;TIAN Jiya(School of Information Engineering,Xinjiang Institute of Technology,Aksu 843100,China)

机构地区：[1]新疆理工学院信息工程学院,新疆阿克苏843100

出　　处：《现代信息科技》2025年第4期64-68,73,共6页Modern Information Technology

基　　金：国家级大学生创新创业训练计划项目(202413558003);2021年度校级项目(ZY202105);2023年度校级重点项目(ZZ202303)。

摘　　要：YOLOv9作为YOLO系列模型中的最新版本之一,其平台移植方便与检测步骤简易,相比于传统的图像识别技术,基于深度学习的物体检测模型具有更强的特征提取和泛化能力,能够更好地识别复杂的物体和场景。基于YOLOv9c葡萄病害识别检测算法研究,针对传统的病害识别方法存在着识别准确率低、耗时长等问题,对我国七种葡萄病害进行识别,进行训练之后,平均检测度mAP50为92.7%,实验结果表明,该方法可以实现葡萄病害实时检测,大大提高了农业生产效率,满足葡萄病害检测应用场景的精度要求和实时性。As one of the latest versions in the YOLO series of models,YOLOv9 features convenient platform transplantation and simple detection procedures.Compared with traditional image recognition technologies,object detection models based on Deep Learning possess stronger feature extraction and generalization capabilities,and can better recognize complex objects and scenes.Based on the research on YOLOv9c grape disease identification and detection algorithm,aiming at the issues such as low recognition accuracy and long processing time existing in traditional disease recognition methods,this paper conducts recognition of seven types of grape diseases in China,and the average detection metric mAP50 reaches 92.7% after training.Experimental results demonstrate that this method can achieve real-time detection of grape diseases,significantly improving agricultural production efficiency and meeting the precision and real-time requirements of grape disease detection application scenarios.

关 键 词：YOLOv9 葡萄病害 实时检测 损失函数 高性能 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]