Improving 3D indoor mapping with motion data

Abstract

Using both RGB and depth information obtained from low-cost RGB-D cameras, 3D models of indoor environment can be reconstructed, which provide extensive knowledge for mobile robots to accomplish tasks such as localization, mapping, interaction with human, etc. Due to the limited views of RGB-D cameras, additional information about the camera pose is needed. In this paper, an enhanced 3D mapping algorithm is proposed to overcome the limitations. The motion of the RGB-D camera is estimated by a motion capture system after a calibration process. Based on the estimated pose, a multi-level ICP (Iterative Closest Point) algorithm is used to improve the alignment. The result shows that the 3D map can be generated in real-time. We compare our results with other approaches to show the robustness of our algorithm.