沈阳农业大学学报2025，Vol.56Issue(3)：95-105,11.DOI:10.3969/j.issn.1000-1700.2025.03.010
基于CMS-YOLOv8n的葡萄叶片病害检测
Detection of Grape Leaf Diseases Based on CMS-YOLOv8n
摘要
Abstract
[Objective]In complex agricultural scenarios,manual observation of grape leaf diseases has the problems of low efficiency and a high misjudgment rate.To change this situation,improve the accuracy and efficiency of grape leaf disease detection,and meet the demand for early detection and prevention of diseases in agricultural production,a grape leaf disease detection algorithm named CMS-YOLOv8n based on the improved YOLOv8n model is proposed.[Methods]Firstly,CBAM(Convolutional Block Attention Module)is introduced into the backbone network and the neck network.By combining channel and spatial attention,CBAM enables the model to focus more effectively on the features of the diseased areas.When facing irregular disease targets on grape leaves,ordinary models may have difficulty in accurately capturing the features.However,the model introduced with CBAM can automatically learn the important features of the diseased areas in both the channel and spatial dimensions,thus significantly enhancing the representation ability for irregular disease targets.Secondly,a new C2f_MS-Block module is designed to replace the C2f module in the neck network.The multi-scale building block can extract disease target information from different scales,and can well capture the features of diseases with different sizes and shapes.While reducing the complexity of the model,it greatly improves the ability to process multi-scale information of disease targets,enabling the model to stably detect diseases in different environments.[Results]The improved model is verified through experiments,and the results show that the performance of the improved model has significantly improved compared with the original YOLOv8n.In terms of detection accuracy,the mAP50 has increased by 1.3%,and the mAP50-95 has increased by 0.3%.In terms of model complexity,the FLOPs have been reduced from 8.1 G to 7.8 G.[Conclusion]This means that the improved model not only has higher detection accuracy but also requires less computation during operation,which is more conducive to deployment and application in practical scenarios.In future agricultural production,it is expected to be further promoted and applied to help growers detect grape leaf diseases in a timely manner,reduce economic losses,and promote the healthy development of the grape industry.关键词
YOLOv8/病害检测/CBAM/MS-BlockKey words
YOLOv8n/disease detection/CBAM/MS-Block分类
计算机与自动化引用本文复制引用
冀常鹏,佐永吉,代巍..基于CMS-YOLOv8n的葡萄叶片病害检测[J].沈阳农业大学学报,2025,56(3):95-105,11.基金项目
辽宁省教育厅基本科研项目(LJKMZ20220677) （LJKMZ20220677）
