Street view cross-sourced point cloud matching and registration

Abstract

Object registration has been widely discussed with the development of various range sensing technologies. In most work, however, the point clouds of reference and target are generated by the same technology, such as a Kinect range camera, LiDAR sensor, or Structure from Motion technique. Cases in which reference and target point clouds are generated by different technologies are rarely discussed. Due to the significant differences across various point cloud data in terms of point cloud density, sensing noise, scale, occlusion etc., object registration between such different point clouds becomes extremely difficult. In this study, we address for the first time an even more challenging case in which the differentlysourced point clouds are acquired from a real street view. One is generated on the basis of an image sequence through the SfM process, and the other is produced directly by the Li-DAR system. We propose a two-stage matching and registration algorithm to achieve object registration between these two different point clouds. The experiments are based on real building object point cloud data and demonstrate the effectiveness and efficiency of the proposed solution. The newly proposed solution can be further developed to contribute to several related applications, such as Location Based Service.

A Coarse-to-Fine Algorithm for Matching and Registration in 3D Cross-Source Point Clouds

“…In this experiment, starting from the VSFM point clouds, we conduct coarse matching to select 10 potential point clouds; then, we undertake fine registration and compare this with other methods. We use method in [47] to downsample the point cloud to 10% of the original points and run our GM-CSPC method with [8], ESF+CPD, ESF+GO-ICP, ESF+JR-MPC. The quantitative evaluation results are shown in Table III, indicating that our results achieve the highest accuracy in matching and registration.…”

“…We evaluate the coarse-to-fine pipeline and compare the performance of proposed fine registration algorithm against the well known registration methods. For coarse matching, we use the ESF method; for the fine registration, we compare our method with registration of [8], CPD [36], JR-MPC [38] and GO-ICP [17].…”

“…For example, similar to our work, Peng. et al [8] proposed a two-stage algorithm using ICP. Due to the limitations of ICP, the accuracy of the final registration results is low, and the results are not visually registered correctly.…”

A Coarse-to-Fine Algorithm for Registration in 3D Street-View Cross-Source Point Clouds

“…Different to previous generative GMM, we propose a coarse-to-fine, scale normalization and complexity reduction strategies to extend it to be suitable to the new cross-source problem. Different to previous ICPbased method [8], we utilize the statistic property of crosssource point cloud to deal with the large variations in local points. It successfully overcomes the point-to-point limitation in ICP-based methods.…”

“…The top 2 registration results of the 4 objects with the proposed method and[8]. Each row represents the results for one object.Figures in the 1th column and the 3th respectively represent the GMM registration results of retrieved rank 1 and rank 2 candidate regions in LiDAR and SFM point cloud.…”