空间控制技术与应用2025，Vol.51Issue(6)：50-62,13.DOI:10.3969/j.issn.1674-1579.2025.06.005
基于事件增强的RGB-Event多模态目标跟踪方法
Event-Enhanced RGB-Event Multimodal Tracking Method
摘要
Abstract
Target tracking is a fundamental task in the field of space exploration,and it can be applied to realize functions such as cooperative target navigation,non-cooperative target surveillance,and on-orbit spacecraft collision early warning.Traditional RGB-based tracking methods have limited performance because their performance is affected by factors like cosmic illumination conditions and target velocity.To address this problem,a multi-modal RGB-Event tracking method named EETrack is designed in this paper,and the expression of event information is enhanced by this method.Firstly,event information and RGB information are fused through an event information enhancement module.Then,a Transformer backbone network optimized by a Hopfield associative layer is used to extract features.Finally,the extracted features are fed into the tracking head,and the bounding box and center offset are predicted by the tracking head.Experiments are conducted on the long-term large-scale single-target tracking dataset FELT.The results demonstrate that compared with internationally advanced methods,EETrack achieves the optimal performance in terms of target tracking Success Rate(SR),Precision Rate(PR),and Normalized Precision Rate(NPR).Specifically,in scenarios with high-precision requirements,the SR,PR,and NPR indicators are increased by 0.6%,1.3%,and 1.1%respectively.This tracking algorithm can well accomplish task under extreme illumination and high-speed targets.Long-term stable tracking is ensured by this method,more accurate target tracking results are achieved,and a reference is provided for precise target tracking in the space field.关键词
Transformer网络架构/RGB-Event/多模态/长期/目标跟踪Key words
Transformer/RGB-Event/Multimodal/Long-term/Target Tracking分类
信息技术与安全科学引用本文复制引用
CHENG SiLu,MA XinMing,HE YingBo,XU JiangTao..基于事件增强的RGB-Event多模态目标跟踪方法[J].空间控制技术与应用,2025,51(6):50-62,13.基金项目
国家自然科学基金资助项目(62334008)、天津市自然科学基金资助项目(23JCQNJC01940)和空间光电测量与感知实验室开放基金资助(LabSOMP-2023-14) National Natural Science Foundation of China(Grant No.62334008),Natural Science Funds of Tianjin(Grant No.23JCQNJC01940),Open Fund of Laboratory of Space Optoelectronic Measurement and Perception(Grant No.LabSOMP-2023-14) （62334008）
