可见光与红外融合目标跟踪技术研究进展综述  

Review of Research Progress on Visible and Infrared Images Fusion Target Tracking

作　　者：方彦策 赵君灵 黄昭龙 李旗挺 杜立超 张宏江[3] 宋眉眉[2] FANG Yance;ZHAO Junling;HUANG Zhaolong;LI Qiting;DU Lichao;ZHANG Hongjiang;SONG Meimei(College of Information Science and Engineering,Ocean University of China,Qingdao 266100,China;College of Science,Tianjin University of Technology,Tianjin 300384,China;R&D Department,China Academy of Launch Vehicle Technology,Beijing 100076,China)

机构地区：[1]中国海洋大学信息科学与工程学部,山东青岛266100 [2]天津理工大学理学院,天津300384 [3]中国运载火箭技术研究院研究发展部,北京100076

出　　处：《计算机测量与控制》2022年第10期7-16,共10页Computer Measurement &Control

基　　金：中国航天科技集团公司钱学森青年创新基金资助项目。

摘　　要：目标跟踪是计算机视觉领域中的一个重要的问题,自从被提出以来逐渐发展出了不同类型的技术方法;由于可见光与红外光在目标跟踪方面存在互补性,两者融合的目标跟踪技术在性能和鲁棒性上比传统目标跟踪方法更具有优势;近年来,人工智能技术的发展推动了可见光与红外融合目标跟踪技术的快速进步;面向可见光与红外融合目标跟踪技术的发展历程,首先针对可见光与红外融合领域经典方法和近年来出现的技术进行梳理和总结,特别是对近两年的研究进展进行了归纳整理,具体包括基于Transformer、基于注意力机制、基于时间序列、自适应融合和基于多模态编解码器的可见光与红外融合目标跟踪方法;然后介绍了可见光与红外融合数据集及目标跟踪的评价指标;最后对未来的发展方向作了展望。Object tracking is an important problem in the field of computer vision,and different types of technical methods have been gradually developed since it is proposed.Due to the complementarity of visible and infrared light in target tracking,the fusion of the two methods has more advantages than traditional target tracking in performance and robustness.In recent years,the development of artificial intelligence has promoted the rapid progress of the visible and infrared light fusion target tracking.Facing the development history of the visible and infrared light fusion target tracking technology,firstly,the classic method of visible and infrared light fusion and the technologies emerging in recent years are sorted out and concluded,especially the research progress in the past two years is summarized,which are based on Transformer,attention mechanism,time series,self-adaptive fusion and multi-modal decoder;then the visible and infrared light fusion datasets and target tracking evaluation index are introduced;Finally,the future development of this field is prospected.

关 键 词：目标跟踪 可见光与红外融合(RGBT) 深度学习 TRANSFORMER 注意力机制 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]