Online RPCA Background Modeling Based on Color and Depth Data

This paper applies online RPCA framework on color and depth data for background modeling and foreground segmentation. First, it could model a better background scene for video sequence in real-time. Second, it is a refinement for foreground segmentation. In this paper, depth data and color data are processed separately under online RPCA method, the structured sparse matrix are combined together using data fusion method for better foreground segmentation. Experiments show that our combined foreground results have a better and robust performance than results with one way alone.