Constructing 3D reality models of urban scenes from ground-based synchronized laser data and visual data

We discuss new methods for creating 3D reality models of urban scenes. Our algorithm automatically extracts 3D features from range scans, constructs feature unit, and defines virtual feature to complement some missing information due to occlusions. Our algorithm automatically computes local registrations between individual scans using feature units, builds a topological graph using MST and places all scans in the coordinate system. According to the features of TH-3DLCS-2001 scanner we could fuse laser data and visual data easily.