A People Tracking System Based on Binary Encoding of Continuum-Removed Color-Histogram

摘要

In this paper, we have proposed a real-time and robust system for tracking and recognition of multiple people, with the use of a binary encoding of continuum-removed color histogram based technique which is commonly applied to the image processing of hyper-spectral remote sensing. The color characteristic information of tracked people is simplified as a 32-bit binary number. However, the main color feature has been retained. A color space transformation and multiple color-vectors (10 groups) similarity comparison schemes are also devised to make the recognition of people robust, against change in background illumination and change in pose of tracked people, after an occlusion. The system has been tested on a variety of video data and the experimental results show that the simplified multiple color-vectors can track multiple people more effectively.