Spatio-temporal motion segmentation and tracking under realistic condition

Tracking technique is one of the most active research topics in the domain of computer vision currently, and motion analysis presents the essence of moving object tracking in video. This paper proposed a novel method for tracking based on spatio-temporal motion segmentation. First, successively layered moving object silhouettes encode system time termed the timed motion history image for motion trajectory representation. Then, a spatio-temporal segmentation procedure is used to labeling motion regions by estimating density gradient in spatial-temporal domain. Several examples are shown to illustrate that this solution has high effectiveness and robust nature, providing practical tools for moving object tracking in real time surveillance system