农业装备与车辆工程2025，Vol.63Issue(5)：30-36,7.DOI:10.3969/j.issn.1673-3142.2025.05.005
基于改进YOLOv10的烟叶病害程度检测
Tobacco disease level detection based on improved YOLOv10
王泓博1
作者信息
- 1. 上海烟草机械有限责任公司,上海 201206
- 折叠
摘要
Abstract
The precise detection and severity assessment of tobacco diseases are the crucial steps for ensuring the yield and quality of tobacco leaves.In response to the problems such as the low efficiency of traditional manual visual inspection,the high missed-detection rate of small disease spots in low-resolution images and the insufficient quantification of disease severity in existing computer vision methods,an improved YOLOv10-3S model was proposed.The ability to extract detailed features of small targets was enhanced by introducing the Spatial Depth-Preserving Convolution(SPD-Conv).The positioning accuracy of disease regions was improved by integrating the Separable Enhanced Attention Module(SEAM).The performance of bounding box regression was optimized by using the SIoU loss function.The experimental results showed that in the task of detecting the severity of tobacco diseases,the precision of the improved model reached 99.1%,the recall rate reached 97.8%,and the mAP@0.5 and mAP@0.5∶0.95 were respectively increased to 98.9%and 87.9%.Especially,the performance improvement was significant in the detection of mild diseases.An efficient and reliable technical reference was provided for the intelligent grading,prevention and control of tobacco diseases.关键词
YOLOv10/烟叶/病害程度/目标检测/计算机视觉Key words
YOLOv10/tobacco leaf/disease degree/object detection/computer vision分类
农业科技引用本文复制引用
王泓博..基于改进YOLOv10的烟叶病害程度检测[J].农业装备与车辆工程,2025,63(5):30-36,7.
