Moving Object Detection in Dynamic Scenes Based on Optical Flow and Superpixels

Moving object detection under a dynamic background has been a serious challenge in real-time computer vision applications. Global motion compensation approaches, a popular existing technique, aims at compensating the moving background for moving target segmentation. However, it suffers from inaccurate global motion parameters estimation. The paper presents a moving object detection technique that combines TV-L1 optical flow with SLIC superpixel segmentation to characterize moving objects from a dynamic background. SLIC superpixel segmentation can adhere to boundaries of objects, and thus improve the segmentation performance. TV-L1 optical flow implemented on GPU reports competitive smooth flow field with real-time performance. Experimental results on various challenging sequences demonstrate that the proposed approach achieve impressive performance.