Abstract:Backgrounds of many scenes include complex moving objects, a comparatively better way to solve such kind of problem is to acquire time sequence model of one or a group of pixels, which can better handle the variation with time, but it costs computational complexity. In order to obtain effect close to adaptive filtering, this paper adopts codebook to model background, select RGB color space to learn codebook elements covering RGB channels, and be able to solve the problem with drastic variations of pixel values, then detect the foreground objective robustly. Via experiment shows, the proposed codebook-based background model has better accuracy and robustness than other traditional objective detection algorithms.