Semiautomatic Registration of Terrestrial Laser Scanning Data Using Perspective Intensity Images

Point cloud registration is an important procedure for terrestrial laser scanning data processing. Artificial targets are usually used in practice to guarantee a robust registration. However, installing and locating targets are labor-intensive for large surveying and mapping project. A methodology for semiautomatic registration of terrestrial point clouds using perspective intensity images is presented in this letter. To register two point clouds, perspective intensity image series of the scanning area are first generated from the point clouds. Then, corner points are extracted from the generated perspective intensity images and are interactively selected as tie points. The 3-D coordinates of the selected tie points are directly obtained or estimated using least squares method. Registration parameters are solved using singular value decomposition. To improve robustness and accuracy of the registration, random sample consensus is used to remove outliers in the tie points. The robustness and effectiveness of the presented methodology is demonstrated by experimental results.