无线电工程2025，Vol.55Issue(8)：1571-1579,9.DOI:10.3969/j.issn.1003-3106.2025.08.003
基于超分辨率和YOLO的红外小目标检测
Infrared Small Target Detection Based on Super-resolution and YOLO
摘要
Abstract
To solve the problems of low recognition rate and high false alarm rate in long-range,weak energy,and low-resolution infrared image target detection,a deep learning network SR-YOLOv8 based on YOLOv8n is designed by combining super-resolution and deep learning.An SRNet network has been established in the backbone network,and the super-resolution preprocessing is performed on the input image to increase the number of pixels of small targets,a Global Context Aware Module(GCAM)is added to the neck part to enhance the distinction between the target and the background,suppress the influence of complex backgrounds.The Complete Intersection over Union(CIoU)loss function is improved,and β-CIoU is proposed as a new bounding box loss function,which is able to flexibly adjust the loss value based on the size of the detected target to improve the model detection performance.On the infrared remote sensing dataset SIRST,the mAP0.5 and mAP0.5:0.95 reach 0.844 and 0.351 respectively,which are increased by 4.7%and 11.4%compared with the original model YOLOv8n.It effectively solves the problems of missed detection and false detection.关键词
红外遥感/超分辨率/完全交并比/小目标检测/YOLOv8Key words
infrared remote sensing/super-resolution/CIoU/small target detection/YOLOv8分类
信息技术与安全科学引用本文复制引用
李璐,陈清江..基于超分辨率和YOLO的红外小目标检测[J].无线电工程,2025,55(8):1571-1579,9.基金项目
国家自然科学基金(12101482) National Natural Science Foundation of China(12101482) （12101482）
