A variational approach for ego-motion estimation and segmentation based on 3D TOF camera

A variational approach for ego-motion estimation and segmentation based on 3D camera is proposed in this paper. The employed sensor is a 3D TOF camera, SR-3000 from Mesa, which can provide intensity and range data at high frame rates. Unlike most other motion-based segmentation algorithm, this approach is performed by joining both segmentation and estimation into an energy functional. The problem is formulated by energy minimization based on level set method and solved by iterating between the estimation and segmentation steps derived from the energy functional. The efficacy of this method is validated by experiments in various indoor environments.