Planar structure based registration of multiple range images

In this paper, we describe the method for aligning multiple range images given by a range finder. Especially we will use range images of inside and outside of buildings which contain many planar structures. In our method for registration of range images, we consider not only 3D positions of range data but normal vectors of planes in the scene in order to refine the ICP (Iterative Closest Point) algorithm. First, we extract planes of range images in order to calculate normal vectors of planes in the scene. Then, we estimate the motion parameters that are composed of a rotation matrix and a translation vector using our refined ICP algorithm. We present results that demonstrate our approach's ability to align range images more accurately and faster than the standard ICP algorithm.