A new algorithm for target tracking using fuzzy-edge-based feature matching and robust statistic

摘要

We present a new algorithm for real-time tracking of moving targets in terrestrial scenes using a mobile camera. We used the fuzzy edge of the target and the modified LMedS statistic for robust tracking. In this method, we first select proper feature points from the edge of the target. These feature points are then matched with points in the region of interest in the next frame using fuzzy-edge-based feature matching. Then using the modified LMedS statistic and affine transformation, a motion model is calculated for the target. Using this model the location of the target is identified in the next frame. In addition to the robust statistic, the use of reflectance information for edge detection has made our tracking algorithm reliable against high illumination changes. The tracking system is also capable of target shape recovery and therefore it can successfully track targets with varying distance from camera or while the camera is zooming.