Visual Object Tracking via Joint Learning Method

Abstract

A novel visual object tracking algorithm Using Spatio-Temporal Contextual reasoning via joint learning method is proposed. The schema extracts the rectangle and high-dimensional features at different scales of samples, then constructs a measurement matrix to map high-dimensional features to lower-dimensional image space using a prior knowledge of sparse video frame, and formulates the spatio-temporal relationships between the object of interest and its local context based on a joint method combining feature and deformation handling with classification model. Experimental results on some publicly available benchmark video sequences show that the proposed algorithm can handle occlusion efficiently, and be robust to pose and illumination variations over other approaches.