Object Tracking Using Maximum Color Distance and Shape Density

Abstract

This paper presents an effective object tracking algorithm for simultaneous changes, such as deformation, translation, scale, and color variations. The proposed algorithm is based on the maximum color distance (MCD) and the shape density. The MCD provides invariance for illuminations changes, while the shape density is the density of shape distribution corresponding to the respective distance from the center position of an object. The generated probability density function (PDF) of shape density is invariant for object orientation and robust to deformable object and linear scale variation. The initialized model of a target for PDF of the MCD and shape density is compared with candidates by Least Mean Square Error in a next frame. Experimental results show that the tracking system using the proposed algorithm with the MCD and the shape density provides effective results for simultaneous changes in target such as translation, deformation, scale and color variations.