Three-dimensional Scene Understanding in Mobile Laser Scanning data

Balázs Nagy, Csaba Benedek
Three novel approaches are proposed in this thesis connected to 3D point cloud processing, namely semantic scene segmentation, point cloud registration, and camera-Lidar extrinsic parameter calibration. The proposed results are based on the combination of 3D geometry processing and deep learning based algorithms. The proposed methods have been evaluated in large point cloud databases containing various complex urban traffic scenarios and we have compared the proposed approaches to
more » ... es to state-of-the-art methods and the results show significant improvements against the existing novel approaches.
doi:10.15774/ppke.itk.2020.010 fatcat:yksnot25hvftjgd557dix3hlgu