通信与信息技术Issue(3)：129-133,5.
基于自适应YOLOv8的遥感图像目标检测算法
An aerial image detection method based on improved YOLOv5
黄振宇 1赵亦戈 1刘宜欣1
作者信息
- 1. 中国兵器工业集团214研究所,江苏 苏州 215163
- 折叠
摘要
Abstract
In UAV remote sensing images,the captured target objects usually appear overlapping,occlusion and small targets,which leads to the problem that the existing algorithms are difficult to be applied stably.This paper proposes an adaptive YOLOv8-based remote sensing image target detection algorithm.Firstly,the multi-residual module is used to improve the Backbone of the original YOLOv8 and enhance its feature extraction ability.Then,the spatial attention mechanism is proposed to replace the original C2f module in Neck to im-prove the sensitivity of the network to small target information.Finally,the regression loss function is added,and the original loss of YO-LOv8 is combined to improve the robustness of the network to small target detection.The experimental results on Dior data set show that the accuracy and stability of the proposed algorithm are improved by 10.3%and 6.1%respectively compared with the original YOLOv8 al-gorithm,and the ablation experiment further proves the superiority of each module.关键词
无人机遥感图像/小目标检测/空间注意力机制/YOLOv8Key words
UAV remote sensing image/Small target detection/Spatial attention mechanism/YOLOv8分类
信息技术与安全科学引用本文复制引用
黄振宇,赵亦戈,刘宜欣..基于自适应YOLOv8的遥感图像目标检测算法[J].通信与信息技术,2025,(3):129-133,5.
