Adaptive threshold based codebook background modeling algorithm

摘要

The background modeling algorithm of Codebook is efficient in memory, speed, and highly optimized for general use. So, it is widely used in moving object detection of video surveillance. To overcome the difficulty of setting threshold manually while constructing the codebook, a codebook background modeling algorithm of adaptive threshold is proposed. Codebook background modeling algorithm with self-adaptive threshold pays attention to the value of same pixel changing with different frames, and adopts Maximum Mahalanobis Distance to compute the difference which changes severest and the difference determined the threshold. This algorithm can solve the false detections which are caused by setting the threshold manually. Experiments show that the proposed algorithm can not only set the threshold adaptively but also get more accurate results.