目标跟踪研究综述

在监控系统中获取感兴趣的目标是目前极具挑战性的研究热点之一,在视频监控、人机交互和军事领域都具有巨大的应用前景。目标追踪问题主要的技术难点是实时、准确和稳定地将感兴趣的目标表现出来,但是由于目标运动方式、运动场景和目标外在特征的突然改变以及光照变化和拍摄时的抖动等问题都会导致监测追踪系统准确率和稳定性的下降。本文对国内外目标追踪问题的热点和难点进行了详细的分析论述,将目标跟踪归结为目标识别和追踪两个部分来详细讨论,同时考虑到场景对目标追踪的直接影响,将场景理解作为目标追踪的重要技术难点进行了探讨,指出了视觉跟踪模型的具体问题。

Survey of Visual Objects Tracking

Detecting and tracking objects are the most prevalent and challenging tasks,which will be wildly used in video monitoring,human and computer interaction as well as military fields.Real-time,accurately and steadily expressing the objects are the key problems,whereas abrupt object motion,changing appearance patterns of both the object and the scene may lead to low accuracy.In this paper,tracking problem is categorized into two parts:object detection and tracking.Besides,the scene understanding is considered in visual tracking.In the end,the paper points out the drawbacks of the existing visual tracking systems,and prospects the future researches.