Fast 3D model generation in urban environments

Abstract

Describes an approach to fast generation of three dimensional (3D) models in urban environments. Rather than using a single, expensive 3D scanner, the 3D scene is captured by a combination of two fast, inexpensive 2D laser scanners. Additionally, the data acquisition system consists of a digital camera, a velocimeter and a heading sensor, all mounted on top of a truck which is driven in city streets under normal traffic conditions. One of the laser scanners is used for relative position estimation via a scan matching algorithm, while the other measures the shape of building facades for point cloud generation. The accuracy of the position estimates obtained from laser scans is further improved by taking into account the data from other sensors. Our approach is tested in a real city environment, and the resulting path estimates and 3D models are shown.