视频目标跟踪综述
摘 要: 研究近年来主流的目标跟踪算法。通过文献阅读和归纳对比,分析了使用生成式模型和判别式模型的目标跟踪算法。结果显示,对于存在复杂干扰因素的场景,采用第二类模型的目标跟踪算法的跟踪效果更好。文章为视频跟踪领域的研究者们提供了一个关于目标跟踪算法的客观分析。
关键词: 目标跟踪; 深度学习; 相关滤波; 计算机视觉
中图分类号:TP399 文献标识码:A 文章编号:1006-8228(2022)01-32-04
Overview on video target tracking
Zhang Feng, Feng Ping
(Guizhou University of Finance and Economics, Guiyang, Guizhou 550025, China)
Abstract: This paper focuses on the mainstream target tracking algorithms in recent years, and analyzes the target tracking algorithms using generative and discriminative models through literature reading and inductive comparison. The results show that for the scenes with complex interference factors, the target tracking algorithms using the second kind of model have better tracking effect. This paper provides an objective analysis of target tracking algorithms for researchers in the field of video tracking.
Key words: target tracking; deep learning; correlation filtering; computer vision
0 引言
根据人们对感知觉的研究发现,人们通过视觉、听觉、味觉、嗅觉和肤觉来接受外部的刺激,其中视觉是人们获取外部刺激的最主要途径,约占80%。(剩余7303字)