Terrestrial Lidar Image Data Registration

The 3D reconstruction technique based on LiDAR range image has developed rapidly in recent years, demonstrated broad application prospects, and thus attracted lots of attention. Range image registration, one of the key parts of 3D reconstruction technique, has become a research focus.Large-scale buildings with obvious planar characteristics are taken as research objects in this paper. Their range images are acquired by our middle-long terrestrial LiDAR equipment. The registration algorithm of range images is designed, and tested. The main work of the paper is as follows:(1) Firstly, the flow of LiDAR range image acquisition and the registration via reflectors are studied with the LMS-Z series LiDAR system and its software Riscan Pro in our laboratory. The disadvantages and limitations of reflector-based registration are analyzed for objects we focus on, and the importance of registration algorithm for LiDAR range images is emphasized.(2) Secondly, Initial registration is achieved by feature extraction and data reduction based on normal direction error, planar segmentation for our objects based on Stamos' segmentation method, and planar point sets calculation.(3) Finally, result of step (2) is refined and an accurate registration is achieved by improving ICP algorithm after an analysis on the advantages and disadvantages of the basic ICP algorithm for objects with many planar features, and combining planar segmentation algorithm with spherical triangle constraint nearest neighbor (STCNN).The above algorithms are implemented with VC++ and Visual Data Structures Class Library VTK, tested by real data collected on the campus, and proved effective by experimental results.