The Integration Adjacent Frame Difference of Improved ViBe for Foreground Object Detection
Yongqiang Li, Wanzhong Chen, Rui Jiang · 2011
When initializing the background model with ViBe algorithm, the prospects for moving target in background area will also be treated as a background model, and Video stream in the next area will appear the phenomenon of ghost. Although ViBe algorithm provide the method of updated background model, slowed to eliminate ghost region. When the foreground object passed these areas, the detection accuracy will be dropped; at the same time, there is ghosting region, the false detection algorithms are high. This paper presents a improved ViBe algorithm based on adjacent frame difference algorithm. The algorithm considers the time domain correlation between a few frames before the video, based on this correlation can quickly eliminate shadows. Experimental results show that the proposed algorithm while preserving the advantages of ViBe algorithm can quickly remove the ghost region and improve the detection accuracy which the prospects object through the Ghost zone and reduce the false detection rate.