沈阳农业大学学报2025，Vol.56Issue(4)：93-102,10.DOI:10.3969/j.issn.1000-1700.2025.04.010
基于改进YOLOv10n的石榴病害检测
Pomegranate Disease Detection Based on Improved YOLOv10n
摘要
Abstract
[Objective]Aiming at the problems of low detection accuracy,insufficient generalization ability,and complex feature extraction in pomegranate disease detection under complex backgrounds,variable environments,and multi-scale target recognition,a pomegranate disease detection model named MBC-YOLOv10n is proposed.[Methods]Firstly,a Mixed Local Channel Attention mechanism(MLCA)is integrated into all C2f modules of the YOLOv10n model to enhance the model's sensitivity to pomegranate disease features and improve detection precision.Secondly,considering the variability and multi-scale characteristics of pomegranate diseases,a Bi-directional Feature Pyramid Network(BiFPN)is incorporated to fuse features at different scales,thereby improving both precision and recall without significantly increasing the number of parameters.Finally,a Convolutional Block Attention Module(CBAM)is introduced to strengthen the model's anti-interference ability against complex backgrounds through spatial and channel attention mechanisms.[Results]The improved MBC-YOLOv10n model shows significant performance gains compared to the original YOLOv10n.The mean Average Precision at 0.5 IoU(MAP50)increases by 1.3%to 90.1%,Precision improves by 2.4%to 90.5%,Recall increases by 2.0%to 88.4%,and the mean Average Precision across IoU thresholds from 0.5 to 0.95(MAP50-95)improves by 5.0%to 57.9%.The parameter count remains low at only 2.9 M,significantly enhancing the model's capability for pomegranate disease detection.The MBC-YOLOv10n achieves excellent detection accuracy while maintaining a lightweight structure,ensuring both high precision and real-time performance.[Conclusion]The proposed MBC-YOLOv10n model can effectively detect pomegranate diseases under natural environmental conditions,balancing high accuracy and lightweight design.It provides a reliable technical foundation for early warning and precise prevention of pomegranate diseases in smart agriculture and lays a theoretical basis for the optimization and future application of disease detection models.关键词
石榴病害/MLCA/BiFPN/CBAMKey words
pomegranate disease/MLCA/BiFPN/CBAM分类
农业科技引用本文复制引用
乔世成,潘春宇,白明宇,党珊珊,赵晨雨,王国忱..基于改进YOLOv10n的石榴病害检测[J].沈阳农业大学学报,2025,56(4):93-102,10.基金项目
国家自然科学基金项目(62162049) （62162049）
内蒙古民族大学博士科研启动基金项目(BS658) （BS658）
人兽共患病自治区高等学校重点实验室开放基金项目(MDK2022019) （MDK2022019）
内蒙古自治区牧草智能装备创新中心开放基金项目(MDK2025050) （MDK2025050）
内蒙古自治区自然科学基金项目(2025LHMS06012) （2025LHMS06012）
