基于SSE-YOLO的红外小目标检测算法  

Improved Infrared Small Target Detection Algorithm Based on SSE-YOLO

作　　者：大妹 姜麟[1] 陶友凤 胡淼 DA Mei;JIANG Lin;TAO Youfeng;HU Miao(Faculty of Science,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学理学院,云南昆明650500

出　　处：《红外技术》2025年第4期475-483,共9页Infrared Technology

基　　金：国家自然科学基金项目(11761042)。

摘　　要：针对红外成像面积小、分辨率低、易被遮挡导致漏检、检测精度低等问题,本文提出了一种基于SSE-YOLO的红外小目标检测算法。首先在YOLOv8s的基础上引入深度非跨步卷积模块,避免检测过程中细粒度信息的丢失并提高特征学习的效率;其次在特征提取阶段增加专门针对小目标的检测层,以提升模型对红外小目标的提取能力;此外设计了一种高效的双注意力机制(efficient dual-attention mechanism,EDAM),自适应地学习每个通道和空间位置的重要性,从而更好地捕捉图像中的关键信息;然后使用Shape_IoU损失函数来聚焦边框自身形状与自身尺度,进一步提高边框回归的精确度;最后在FLIR数据集和艾睿光电公司拍摄的数据集上进行了一系列实验。结果表明:本文所提方法在两种数据集上的平均精度分别达到了89.8%与92.1%,相比于原始的模型分别提高了3.3%与2.9%。To address the problems of a small infrared imaging area,low resolution,and ease of occlusion—resulting in incorrect detection,missed detection,and low detection accuracy—this paper proposes an infrared small-target detection algorithm based on SSE-YOLO.Firstly,a depth non-stepwise convolution module is introduced on the basis of YOLOv8s to avoid the loss of fine-grained information during the detection process and to improve the efficiency of feature learning.Then,a detection layer specialized for small targets is added in the feature extraction stage to improve the model's ability to extract infrared small targets.In addition,an efficient dual attention mechanism(EDAM)is designed to adaptively learn the importance of each channel and spatial location to better capture key information in the image.Secondly,the Shape_IoU loss function is used to focus on the shape of the boundary itself and its scale,which further improves the accuracy of boundary regression.Finally,a series of experiments were conducted on the FLIR dataset and a dataset captured by IRay.The results show that the average accuracies of the proposed method on the two datasets reach 89.8%and 92.1%,which are 3.3%and 2.9%higher than those of the original model,respectively.

关 键 词：YOLOv8s 红外小目标检测 深度非跨步卷积 Shape_IoU损失函数 双注意力机制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]