TNRAC: a system for tracking multiple moving non-rigid objects using an active camera

摘要

This paper presents an enhanced method for detecting and tracking moving objects from an active camera image sequence using active contour models. Such enhancements are: (1) Using best block matching instead of the conventional block matching technique to locate the promising feature points in the current frame. (2) Developing a new algorithm for rejecting bad feature points that affect negatively the background compensation process. (3) Applying a modified region growing and labeling algorithm to extract moving objects then applying active contour models around these moving objects in order to take into account the object inside this contour. Various factors are studied to measure the system performance and limitations. A comparison of TNRAC and the conventional approach is conducted using both rigid and non-rigid objects in both indoor and outdoor scenes.