Improved object tracking technique using hybrid approach

Object tracking is need of activity log applications used in security surveillance videos. The object tracking using CAMShift method uses histogram estimation of seed point pixels of ROI object in each frame. The object is considered on the basis of its histogram estimation and then score is estimated using accuracy ratio based tracking window locations and ground truth locations. The occlusion occurrences in tracking is main challenge where, after occlusion updated window possesses the features that are not exact as needed and hence tracking window deviates from required position. To overcome this issue, we have proposed method with inclusion of speeded up robust features (SURF) based feature matching approach to estimate tracking window correctly even after occlusion. The results are estimated using standard dataset videos for tracking and compared with original tracking by estimating precision and tracking window location error.