Collaborative Gaussian mixture model for background subtraction
Yongxin Jiang, Xing Jin, Jun Tang, Zhiyou Zhang · 2020
Gaussian mixture per-pixel model cannot handle complex background motion and needs different parameters setting for variant target motion speed scenario. In this paper, a Collaborative Gaussian mixture model for background subtraction is proposed. In which, each pixel was modeled by a background Gaussian mixture model or a foreground Gaussian mixture model. The foreground Gaussian mixture model is respond for pixel value statistics and prepare new background model for the background Gaussian mixture model. The background Gaussian mixture model implement the background Gaussian models update procedure. Furthermore, A periodic control parameter and new parameter update method are proposed to improve the robustness of the algorithm. Evaluation results based on the Cdnet 2012 database are presented in this paper. The results indicate that the proposed algorithm work well on various scenario.