Moving object detection using background subtraction and motion depth detection in depth image sequences

摘要

We propose a novel method that detects moving objects in depth image sequences using background images and motion depth (MD) values. The background image represents the camera view with no moving objects and the MDs are the depth values of moving objects. Foreground regions can be easily detected by background subtraction; however, the foreground regions have some noise and do not contain a regional diffusion of depth values. Therefore, we apply both background subtraction and motion depth detection to detect moving objects. Experimental results show the proposed method robustly detects moving objects, even if the moving region exists in close proximity to the background surfaces. Therefore, the method can be widely used to detect moving objects in depth image sequences.