航空科学技术2025，Vol.36Issue(12)：44-50,7.DOI:10.19452/j.issn1007-5453.2025.12.006
基于DETR改进模型的模糊航拍图像目标检测
Object Detection in Blurry Aerial Images Based on Improved DETR Model
摘要
Abstract
Aerial images often suffer from blurring caused by environmental factors such as haze and jitter,along with difficulties in detecting small targets,which limits the robustness and generalization ability of existing object detection algorithms in aerial scenarios.To address this problem,this paper proposes a blurred-image object detection method that integrates an image adaptive processing module with an improved Transformer-based end-to-end detection model(DETR).The image adaptive processing module dynamically adjusts filtering parameters to remove blurring and enhance image details,effectively generating clear and high-resolution aerial images.The improved DETR model alleviates the issue of positive negative sample imbalance in aerial images through category counting,counting-guided feature enhancement,and a dynamic query selection mechanism,thereby improving the accuracy of small target detection.Experimental results demonstrate that the proposed method achieves clear advantages on blurred images and can accurately detect small objects in aerial scenes.This paper provides an efficient solution for aerial image object detection and enhances the visual perception capability of aerial sensing systems.关键词
航拍图像/图像模糊/目标检测/深度学习/去模糊Key words
aerial images/image blur/object detection/deep learning/deblurring分类
信息技术与安全科学引用本文复制引用
伍麒麟,吴少荃,孙维超..基于DETR改进模型的模糊航拍图像目标检测[J].航空科学技术,2025,36(12):44-50,7.基金项目
航空科学基金(2023Z071077008) Aeronautical Science Foundation of China(2023Z071077008) （2023Z071077008）
