可见光红外跨模态行人重识别方法综述  

Review of Visual-infrared Cross-modal Person Re-identification Methods

作　　者：范慧杰[1] 郁航 赵颖畅 唐延东[1] FAN Huijie;YU Hang;ZHAO Yingchang;TANG Yandong(State Key Laboratory of Robotics,Shenyang Institute of Automation,Chinese Academy of Sciences,Shenyang 110016,China;School of Information Engineering,Shenyang University of Chemical Technology,Shenyang 110142,China;School of Automation and Electrical Engineering,Shenyang Ligong University,Shenyang 110158,China)

机构地区：[1]中国科学院沈阳自动化研究所机器人学国家重点实验室,辽宁沈阳110016 [2]沈阳化工大学信息工程学院,辽宁沈阳110142 [3]沈阳理工大学自动化与电气工程学院,辽宁沈阳110158

出　　处：《信息与控制》2025年第1期50-65,共16页Information and Control

基　　金：国家自然科学基金项目(62273339,U20A20200)。

摘　　要：可见光红外跨模态行人重识别技术因不受夜间限制,可实现全天候监控而受到广泛关注。本文希望通过分析现有可见光红外跨模态行人重识别研究方法及其适用场景、算法优缺点来帮助研究人员根据研究需求找到合适的解决方案。同时,寻找该领域难点和困境,以此探讨可见光红外跨模态行人重识别未来方向。首先介绍行人重识别的概念,回顾发展历程,介绍可见光红外跨模态行人重识别的意义;其后将可见光红外跨模态行人重识别研究方法分为基本方法、辅助模型方法、无监督方法、基于视频的方法,并对每种方法的适用场景、优缺点和未来研究方法展开分析;再对当前可见光红外跨模态行人重识别的评价指标以及现有数据集进行介绍,并对每种数据集的优劣进行分析;最后讨论了可见光红外跨模态行人重识别未来发展。Visible-infrared cross-modality person re-identification technology has garnered wide attention for its ability to provide round-the-clock monitoring regardless of nighttime conditions.We aim to help researchers find suitable solutions by analyzing existing methods,their applicable scenarios,and the advantages and disadvantages of different algorithms.Additionally,We seek to identify current challenges and explores future directions for visible-infrared cross-modality person re-identification.We first introduce the concept of person re-identification,review its development,and highlight the significance of visible-infrared cross-modality person re-identification.Subsequently,it categorizes research methods into basic methods,auxiliary model,unsupervised methods,and video-based approaches,analyzing their applicable scenarios,advantages,disadvantages,and potential future research directions.We also discuss current evaluation metrics and existing datasets for visible-infrared cross-modality person re-identification,analyzing the strengths and weaknesses of each dataset.Finally,We discuss the future prospects in this area.

关 键 词：跨模态行人重识别 计算机视觉 无监督 辅助模型 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]