Human Motion Tracking and Alarm System Based on DaVinci

Abstract

The paper proposes a improved Camshift algorithm which solve the problem of the original Camshift that have limitations when the tracking target have similar color with the background and is obstructed. The paper combines codebook model with the Camshift. The YUV space is used in foreground detection rather than the RGB. The results of experiments show that the algorithm works well in complex background, occlusion and the same color interference. At last we achieve a warning system.