基于深度学习的目标跟踪技术及其多波段应用研究

Research on target tracking technology based on deep learning and its multi-band application

目标跟踪是计算机视觉领域中最具挑战性的领域之一。首先,回顾了目标跟踪技术发展现状并总结出目标跟踪的一般流程;其次,分析了卡尔曼滤波、粒子滤波等传统目标跟踪方法的优缺点;再次,依次重点介绍了固定窗跟踪等相关滤波目标跟踪方法,以及全网络自适应目标跟踪等高性能目标跟踪方法;最后,总结了目标跟踪在红外、毫米波、太赫兹等波段的应用,并对多波段融合目标跟踪进行了展望。

Target tracking is one of the most challenging topics in computer vision field.Firstly,the development of target tracking technology is reviewed and the general process of target tracking is summarized.Secondly,the advantages and disadvantages of traditional target tracking methods based on Kalman filter and particle filter are analyzed.Thirdly,the fixed window correlation filter target tracking methods and high-performance adaptive full network target tracking methods are successively highlighted.Finally,the applications of target tracking in infrared,millimeter wave and terahertz radiation band are summarized,and multi-band fusion target tracking is forecasted.