Fusing color and contour in visual tracking

Abstract

A tracking system with color and contour information is more efficient and robust than one with color or contour only. However, it is difficult to use both color and contour information. In this paper, we present an approach using the particle filter to fuse color and contour cues in tracking. First, we combine color and contour information in a Kalman filter to generate the proposal distribution which is one of the key points to improve the performance of particle filter. Subsequently, the particle filter will be applied to give the final tracking result. Our algorithm framework is flexible and it allows us integrate more measurements. Experimental result shows that this approach is an efficient and robust method. Copyright © 2005 by MVA Conference Committee.