Self-organizing background subtraction using color and depth data

Background subtraction from color and depth data is a fundamental task for video surveillance applications that use data acquired by RGBD sensors. We present a method that adopts a self-organizing neural background model previously adopted for RGB videos to model the color and depth background separately. The resulting color and depth detection masks are combined to guide the selective model update procedure and to achieve the final result. Extensive experimental results and comparisons with several state-of-the-art methods on a publicly available dataset show that the exploitation of depth information allows achieving much higher performance than just using color, accurately handling color and depth background maintenance challenges.