智慧农业导刊2025，Vol.5Issue(10)：27-31,5.DOI:10.20028/j.zhnydk.2025.10.007
基于改进YOLOv11的农作物实时检测研究
摘要
Abstract
In the process of visual inspection,robotic harvesters or UAVs equipped with smart agricultural equipment often encounter complex background environments and a large number of samples.Furthermore,existing target detection methods do not adequately address low-quality samples,which makes it challenging for traditional target detection models to accurately identify objects.This paper proposes a lightweight real-time crop detection model based on YOLOv11-MW.A module based on the Mixed Local Channel Attention(MLCA)mechanism is integrated into the Cross Stage Partial with Spatial Attention(C2PSA)structure of the backbone network,enhancing the model's ability to extract features from numerous samples and overcoming interference from complex backgrounds to improve detection accuracy.The loss function is replaced with Wise-IOU,leading to the introduction of a new detection head,WIOUv3-Detect(WDetect),which reduces the excessive penalty on low-quality samples due to geometric factors,thereby decreasing misidentification and missed detections.Experimental results indicate that the proposed algorithm not only meets the requirements of edge computing power and increases computing speed but also improves the mAP of wheat samples by 1.2%and that of grape samples by 1.9%.The detection of image samples is more balanced and accurate,effectively achieving crop detection.关键词
机器人/无人机/目标检测/YOLO/通道注意力/Wise-IOUKey words
robot/UAV/target detection/YOLO/channel attention/Wise-IOU分类
信息技术与安全科学引用本文复制引用
冉书伟,张丽丽,王维政,宋昕蕊,王佳佳,倪雪松,马玉国..基于改进YOLOv11的农作物实时检测研究[J].智慧农业导刊,2025,5(10):27-31,5.基金项目
国家级大学生创新创业训练项目(202411104021) （202411104021）
中央高校基本科研业务费专项资金项目(3142024038) （3142024038）
