Online 3D world reconstruction with independently moving point detection using stereo vision

This paper deals with a problem of 3D world reconstruction and independently moving point detection using stereo vision. Our algorithm reconstructs 3D world through dense stereo matching, triangulation at local frame, and registration in world frame. In addition, independently moving points are detected by geometric constraints. We have tested our method with real outdoor datasets. The proposed algorithm runs online. Hence, it can apply to mobile platform applications such as intelligent vehicles and mobile robots.