An efficient RGB-D camera based point cloud registration algorithm

Abstract

This paper proposes an indoor scene three-dimensional (3D) reconstruction system using pan-tilt platform and RGB-D camera. The proposed system can automatically reconstruct 3D indoor scenes on a fixed position. An efficient point cloud registration algorithm is proposed to align point clouds based on extrinsic parameters of the RGB-D camera from every presetted pan-tilt control points. Then, a local registration method is performed to refine the alignment result. Experimental results verify the quality and efficiency of the proposed point cloud alignment algorithm by comparing with a state-of-the-art method.