四川轻化工大学学报（自然科学版）2025，Vol.38Issue(5)：61-71,11.DOI:10.11863/j.suse.2025.05.07
一种改进YOLOv8的无人机航拍图像小目标识别算法
A Small Target Recognition Algorithm for UAV Aerial Images by Improved YOLOv8
摘要
Abstract
Aiming at the problems of missed detection and false detection of small targets in Unmanned Aerial Vehicle(UAV)aerial images,which have the characteristics of low pixel density,dense distribution,and susceptibility to background interference,the paper proposes an object detection algorithm based on improved YOLOv8.Firstly,to enhance small target feature extraction,an efficient multi-scale attention mechanism is integrated into the backbone network.Secondly,considering the small target's low pixel proportion,a P2 detection layer with a smaller receptive field is added to improve adaptability.Thirdly,the original Panet network is upgraded to a Bidirectional Feature Pyramid Network(BiFPN)to enrich semantic information.Finally,combining the normalized Wasserstein distance loss and CIoU loss further optimizes performance.Experimental results on the VisDrone2019 dataset show that the mAP50 reaches 41.4%,a 9.2 percentage point improvement over YOLOv8n.Compared with other advanced algorithms,this method outperforms in both accuracy and speed.关键词
小目标检测/YOLOv8/BiFPN/高效多尺度注意力机制/无人机图像Key words
small target detection/YOLOv8/BiFPN/efficient multi-scale attention mechanism/UAV images分类
计算机与自动化引用本文复制引用
刘诗童,刘小芳..一种改进YOLOv8的无人机航拍图像小目标识别算法[J].四川轻化工大学学报（自然科学版）,2025,38(5):61-71,11.基金项目
四川省科技计划资助项目(2017GZ0303) （2017GZ0303）
