Robust Object Tracking Based on ORB Features and Particle Filter

Abstract

In this paper, we present a new approach for robust object tracking in the particle filter framework, it can address well with illumination, occlusion, blur, perspective distortion and some other challenging problems. Current methods rely on global representation, such as color and texture, is sensitive to occlusion. Recently, local description has been the focus of intense research interest. Here, we use a very fast binary local descriptor based on BRIEF, called ORB, which is rotation invariant and resistant to noise. For this reason, we employ novel ORB features instead of traditional global feature and combine with particle filter algorithm for object tracking. The proposed approach is validated on some indoor and outdoor sequences. We made experiments to confirm effectiveness of this method.