基于YOLOv11改进的无人机图像小目标检测模型  

Small-Object Detection Model for UAV Images Based on Improved YOLOv11

作　　者：卢彬 王俊 LU Bin;WANG Jun(Shanghai Electric Smart City Information Technology Co.,Ltd.,Shanghai 200082,China)

机构地区：[1]上海电气智慧城市信息科技有限公司,上海200082

出　　处：《信息与电脑》2024年第22期150-152,共3页Information & Computer

摘　　要：YOLO(You Only Look Once)算法在目标检测领域因其速度与精度优势,历来都是研究重点,然而YOLOv11直接应用于无人机图像检测的效果欠佳。因此,算法在YOLOv11算法的基础上加入了下采样倍数为4的高分辨特征图P2用于检测微小目标,加入通道-空间注意力机制模块来优化特征图提取,加入多尺度卷积注意力模块来提升不同尺寸目标的检测性能。实验结果表明,算法在无人机航拍图像的小目标检测方面表现优异,其m AP@0.5从0.408提高到0.445。YOLO(You Only Look Once)algorithm has always been the focus of research in the field of object detection because of its speed and accuracy.However,YOLOv11 does not perform well when directly applied to UAV image detection.Therefore,on the basis of YOLOv11 algorithm,this algorithm introduces a high-resolution feature map P2 with a downsampling factor of 4 for detecting small targets,incorporates channel-space attention mechanism module to optimize feature extraction,and multi-scale convolutional attention modules to enhances the detection performance for targets with different sizes.Experimental results show that the algorithm performs excellently in the detection of small targets in UAV images.The mAP@0.5 of models increased from 0.408 to 0.445.

关 键 词：注意力机制 小目标检测 YOLOv11 Vis Drone数据集 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]