机电工程技术2025，Vol.54Issue(6)：119-123,5.DOI:10.3969/j.issn.1009-9492.2024.00165
基于改进YOLO模型的红外图像微小目标检测方法
A Micro Target Detection Method in Infrared Image Based on Improved YOLO Model
摘要
Abstract
Infrared remote sensing imaging plays an important role in military observation,night safety monitoring,forest fire monitoring and other fields.However,under the condition of complex background and low contrast,the detection accuracy of weak small targets has always been faced with the problem of low accuracy.The existing model driven methods usually lack robustness when dealing with noise and small-scale targets,while the methods based on deep learning are highly dependent on data and have limitations in feature processing and fusion,resulting in missed detection and false detection.An improved deep learning method is proposed for small target detection in infrared images.In order to capture the long-distance dependence in infrared images,a deep learning network named YOLO-SR based on YOLO is designed.The bottleneck converter module is introduced after the spatial pyramid pool module in the backbone layer.C3 Neck module is designed in the neck layer to better extract and fuse spatial and channel information.Experimental results show that compared with the most advanced data-driven detection method,the proposed method achieves 95.2%mAP(IoU is 0.5)on public data sets.关键词
红外遥感/小目标检测/深度学习/YOLOKey words
infrared remote sensing/small target detection/deep learning/YOLO分类
通用工业技术引用本文复制引用
周虹,陈嘉,周国栋..基于改进YOLO模型的红外图像微小目标检测方法[J].机电工程技术,2025,54(6):119-123,5.基金项目
湖南省教育厅科学研究项目(24C0971) （24C0971）
长沙市社科联哲学社会科学规划课题(2024CSSKKT132) （2024CSSKKT132）
湖南开放大学科研课题(XDK-2024-JG-3,XDK-2024-C-2) （XDK-2024-JG-3,XDK-2024-C-2）
湖南省社科评审委员会课题(XSP2023JYC164) （XSP2023JYC164）
湖南省职业教育教学改革研究项目(ZJGB2022805) （ZJGB2022805）
