Particle Filter Object Tracking Based on Harris-SIFT Feature Matching

Abstract

The object often can't be tracked accurately in the case of illumination changes and occlusions with the traditional algorithm. In this paper, particle filter is used to establish the object motion model and Harris-SIFT is adopted to establish the object model for object tracking. The location of the object in the current frame is predicted with particle filter. Then the SIFT characterization vector of the Harris corner extracted from the possible object region is constructed. The characterization vector is updated according to the match of the object model and the feature points in the object candidate area during the tracking process. The experimental results show that the proposed method is robust and can improve the speed and accuracy of the object detection and tracking significantly.