Bisheng Yang, Yufu Zang, Zhen Dong, Ronggang Huang
Abstract：Laser scanning techniques have been widely used to capture three-dimensional (3D) point clouds of various scenes (e.g. urban scenes). In particular, airborne laser scanning (ALS) and mobile laser scanning (MLS), terrestrial laser scanning (TLS) are effective to capture point clouds from top or side view. Registering the complimentary point clouds captured by ALS and MLS/TLS provides an aligned data source for many purposes (e.g. 3D reconstruction). Among these MLS can be directly geo-referenced to ALS according to the equipped position systems. For small scanning areas or dense building areas, TLS is used instead of MLS. However, registering ALS and TLS datasets suffers from poor automation and robustness because of few overlapping areas and sparse corresponding geometric features. A robust method for the registration of TLS and ALS datasets is proposed, which has four key steps. (1) extracts building outlines from TLS and ALS data sets independently; (2) obtains the potential matching pairs of outlines according to the geometric constraints between building outlines; (3) constructs the Laplacian matrices of the extracted building outlines to model the topology between the geometric features; (4) calculates the correlation coefficients of the extracted geometric features by decomposing the Laplacian matrices into the spectral space, providing correspondences between the extracted features for coarse registration. Finally, the multi-line adjustment strategy is employed for the fine registration. The robustness and accuracy of the proposed method are verified using field data, demonstrating a reliable and stable solution to accurately register ALS and TLS datasets.
Keywords：
Airborne laser scanning
Terrestrial laser scanning
Registration
Building outline
Laplacian matrix
Spectral space
