Textured mesh generation of extracted regions from urban range-scanned LIDAR data

LIDAR range scanners are a popular tool for data acquisition in the field of urban modeling, archaeological preservation, city planning, and more. However, range scanners output a series of discrete distance samples as disconnected points, providing a fundamentally incomplete representation of the underlying structure in a scene. Data from a single LIDAR scan of a region can be trivially triangulated, but when multiple scanners are involved, or when the acquisition platform is mobile, those inter-point relationships are lost. We propose a technique to triangulate such data by identifying logical surfaces within the data and triangulating those surfaces individually. The resulting triangulations are simplified dramatically by using information about the shape of the regions, and texture is applied from camera imagery on the scan vehicle. The result is a high fidelity representation of a scene which is more efficiently rendered on modern hardware than point sets, which occupies less space in memory, and which brings us closer to a solid representation of the true scanned scene.