基于YOLOv8多分类模型的番茄病叶识别应用  

Tomato disease leaf recognition based on the YOLOv8 multi-classify model

作　　者：伍涛 WU Tao(China Railway Construction Heavy Industry Corporation Limited,Changsha,Hunan 410100,China)

机构地区：[1]中国铁建重工集团股份有限公司,湖南长沙410100

出　　处：《农业工程与装备》2024年第1期12-16,共5页AGRICULTURAL ENGINEERING AND EQUIPMENT

摘　　要：针对七种常见的番茄病害,基于深度学习技术YOLOv8的多分类模型,探索了其在农业领域中的应用场景。通过对YOLOv8n-cls和YOLOv8x-cls两个预训练模型在番茄病叶数据集上的训练,对番茄病害叶片图像进行了识别与分类。研究结果表明,对晚疫病、黄化曲叶病、花叶病毒、健康叶片的预测精度高,误检率低;斑枯病、早疫病、叶霉病由于症状相似,识别准确率较低,容易混淆。未来,数据集数量的增加和多样性的提升将成为优化模型参数的主要研究方向,可为提高所有类别病害的召回率和识别精度,实现YOLO深度学习技术的高效应用提供有效参考。A multi-classification model using the deep learning technology YOLOv8 was developed to identify seven common tomato diseases,focusing on its application in agriculture.Two pre-trained models,YOLOv8n-cls and YOLOv8x-cls,were trained on a dataset of tomato disease leaf images.The results show high prediction accuracy for late blight,yellowing leaf curl disease,mosaic virus,and healthy leaves,with low false detection rates.However,spot blight,early blight,and leaf mold had lower recognition accuracy due to similar symptoms,leading to confusion.Future work will increase the number and diversity of datasets to optimize model parameters,aiming to improve recall and accuracy for all disease categories and enhance the practical application of YOLO technology in agriculture.

关 键 词：番茄病害 YOLOv8多分类模型 预测精度 召回率 

分 类 号：S225.92[农业科学—农业机械化工程]