航空科学技术2025，Vol.36Issue(1)：11-21,11.DOI:10.19452/j.issn1007-5453.2025.01.002
脉冲与卷积深度融合的红外弱小目标检测
Deep Fusion of Spiking and Convolutional Networks for Infrared Small Target Detection
摘要
Abstract
With the rapid development of UAV and other aerial vehicle technology,aircraft-mounted infrared small target detection has become a crucial research field.However,traditional algorithms face significant challenges in detection due to factors such as long target distances,small target sizes or partial occlusion.Therefore,developing more precise target detection algorithms is of utmost importance.In response to these challenges,this paper proposed a novel aircraft-mounted infrared small target detection model which integrated spiking and convolutional technologies,namely spike-enhanced fusion feature network(SEFFN)model.Unlike previous infrared small target detection algorithms based on deep learning,this model enhanced feature representation related to small targets through a biomimetic spiking neural network structure,enabling more accurate extraction of small target regions.Specifically,SEFFN comprises four key modules:dilated pyramid convolution(DPC),dual attention fusion(DAF),multi-spike enhance(MSE),and supervised attention module(SAM).These modules work collectively to improve the focus on the small target regions while retaining most of the features of small target without noise interference.Experimental results on two datasets demonstrate that SEFFN outperforms existing model-driven and data-driven algorithms,especially with its F-measure and mIoU on Sirst Aug dataset achieving 85.74%and 75.04%,respectively.This breakthrough validates the effectiveness and superiority of SEFFN in infrared small target detection task.SEFFN is suitable for the aviation field and can be deployed on the edge devices of aircraft to improve the air platform's ability to detect long-distance targets,enabling it to carry out remote monitoring missions,counter enemy threats,and perform other related tasks.关键词
红外弱小目标/脉冲神经网络/目标检测/深度学习/注意力机制Key words
infrared small target/spiking neural network/object detection/deep learning/attention mechanism分类
信息技术与安全科学引用本文复制引用
李乐潇,纪禄平,陈圣嘉,朱思成,郭朝祥..脉冲与卷积深度融合的红外弱小目标检测[J].航空科学技术,2025,36(1):11-21,11.基金项目
航空科学基金(2022Z071080006) Aeronautical Science Foundation of China(2022Z071080006) （2022Z071080006）
