A Robust Object Detecting and Tracking Method

No feature-based vision system can work unless good features can be identified and tracked from frame to frame. This paper addresses robust feature tracking. We extend the well-known Shi-Tomasi-Kanade tracker by introducing a simple scheme for rejecting spurious features. Interest points extracted with the Harris-SIFT detector can be adapted to affine transformations and give repeatable results. In this paper, an efficient method of object tracking with motion prediction and object recognizing is presented. Then object recognizing are implemented by matching local invariant features which are learned online. The experimental results illustrate that the proposed method is capable of tracking objects under partial or severe occlusions.