PROJECT TITLE :

Online Metric-Weighted Linear Representations for Robust Visual Tracking

ABSTRACT:

During this paper, we propose a visible tracker primarily based on a metric-weighted linear representation of appearance. In order to capture the interdependence of different feature dimensions, we tend to develop 2 on-line distance metric learning methods using proximity comparison info and structured output learning. The learned metric is then incorporated into a linear representation of appearance. We have a tendency to show that online distance metric learning significantly improves the robustness of the tracker, particularly on those sequences exhibiting drastic look changes. In order to bound growth in the amount of training samples, we have a tendency to design a time-weighted reservoir sampling technique. Moreover, we have a tendency to enable our tracker to automatically perform object identification during the method of object tracking, by introducing a assortment of static template samples belonging to several object classes of interest. Object identification results for a complete video sequence are achieved by systematically combining the tracking info and visual recognition at every frame. Experimental results on difficult video sequences demonstrate the effectiveness of the method for both inter-frame tracking and object identification.