Multi-Part SIFT feature based particle filter for rotating object tracking

This paper presents an effective rotating object tracking algorithm based on Multi-Part SIFT (MPS) with particle filter. SIFT describes the characteristics of feature points that are invariant to changes in illumination, image noise, rotation, scaling and view direction. In order to use proper effect of SIFT feature for target tracking, single region based object representation is not enough. Therefore, the reference and target object are divided into some sub-regions in order to extract the potential SIFT features for measurement the similarity. This solution introduces spatial information in the representation, without compromising the benefits of SIFT features. In this propose algorithm, for smooth tracking we include the rotation information into the MPS based particle filter since the object rotation directly affects the tracking performance. Additionally, for the illumination changes as well as occlusions, it is required that reference model is adaptively updated. Within the particle filtering framework, we propose to address this adaptation problem using rapidly updated the reference model. We obtained a robust result with the proposed system in particle filter framework. Experimental results show that our system is robust against false target tracking, partial occlusion problem, rotation and scaling without extra computational cost.