Pair-wise automatic registration of three-dimensional laser scanning data from historical building by created two-dimensional images

Abstract

Registration of a point cloud is a great challenge in the process of laser scanning data. So far, many registration methods have been introduced by range data, integrated camera image, and a combination of them. Moreover, the automatic registration of three-dimensional point clouds is an important research topic in both geomatics and computer sciences. In this study, keypoint-based registration of point clouds was introduced. Intensity images were created from the laser scanning data, and then a pair-wise automatic registration was performed with the keypoints extracted from the intensity images by a scale invariant feature transform (SIFT) and affine SIFT (ASIFT). The results were compared with the iterative closest point, which has high accuracy and is the extensively adopted method for the pair-wise registration. Consequently, SIFT and ASIFT keypoints which were extracted from intensity images can be exploited to pair-wise automatic registration of the point clouds.