Multi-camera occupancy reasoning With a height probability map for efficient shape modeling

Abstract

This paper deals with the occupancy reasoning in a multi-camera environments. With the recent development of hardware technologies, the use of multiple cameras for vision algorithms has gained more attention because they are useful for conducting 3D reconstruction from the 2D scenes captured by camera sensors and solving occlusion in crowded environments. As cameras involved are increasing, shape modeling from multi-view sequences gains advantages in aspect of accuracy and coverage but not in performance. To build shape model in large spaces efficiently, we estimate three-dimensional occupancy of multiple moving objects for shape modeling by restricting the scanning space within the occupancy region.