Efficient 3D mapping with RGB-D camera based on distance dependent update

Recently, RGB-D camera has been used widely in indoor SLAM (Simultaneous Localization and Mapping) field. In this paper, we propose an efficient 3D mapping algorithm that considers appearance of dynamic objects. A full 3D mapping system estimates 6-DoF sensor pose and builds 3D voxel grid map simultaneously. In order to build a 3D map, effective method is proposed in terms of computation time. Depending on distance to the object each scan update frequency is varied. By controlling update frequency, we can build a 3D map efficiently. Experiments are carried out with RGB-D sensor and hand-held SLAM results are presented.