An accurate hand tracking system for complex background based on modified KLT Tracker
Joyeeta Singha, Vijay Bhaskar Semwal, Rabul Hussain Laskar · 2016
In a continuous video sequences, various difficulties arise in hand detection and tracking due to various difficulties such as complex background, changes in illumination, occlusion and change in the shape of hand during gesticulation. In this paper, we propose a robust hand detector and tracker which is able to overcome these various challenges. The scheme comprises of two algorithms. Initially, the hand is segmented using an improved method of combination of information from skin color detection and three frame differencing. After detection, the tracking of hand is done using the modified KLT feature tracker where the feature selection is done a priori using the compactness and diversity criteria. Finally, color cues are used to locate hand in the neighborhood of the tracked region. Our experimental results shows that our proposed tracker provides better results in comparison to existing trackers on various challenging hand sequences.