Humans tracking in the complicated background by multi-cue integration

摘要

Human tracking, as a basic research in computer vision, is a key technology in many applications, such as robotics for home using, auto-vehicles in city road and automatic surveillance systems. Although it is always a hot topic in the world of computer vision, it still contains many opening areas for researchers, mainly because there is no single framework could solve all the four problems, occlusion, illumination variation, shadow, and zooming. This paper starts by a new series algorithms for illumination reconstruction, and then introduces a new framework which is called part tracking by multi-cue integration. Aiming at an on-line tracking system in complicated background, this framework is suitable for our application.