Registration of oblique photography point clouds with terrestrial laser scanning point clouds based on geometric features of irregular building

Abstract

Matching the terrestrial laser scan point clouds with the unmanned aerial oblique photography point clouds of high-rise irregular buildings can make them complementary. The challenge lies in the different sources and qualities, besides arbitrary angular deviations of the two types of point clouds. We propose a registration method based on point and line features, which automatically recognises the features of these point clouds using normal vectors and 3D-Harris, and solves the registration parameters through rigorous adjustment. The registration model, process and method are introduced. Finally, through the analysis of real case examples, it was found that the registration result of this method was better than the traditional Iterative Closest Point (ICP) in the analysed cases, the proposed registration method not only exhibits good automation but also has the ability to ignore various differences in heterogeneous point clouds, thereby improving the accuracy and efficiency of registration for such point clouds.