Introduction

One fundamental problem in the design of autonomous mobile cognitive systems is the perception of the environment. A basic issue of mobile robotics is automatic map building of environments. Digital 3D models of the environment are needed in rescue and inspection robotics, facility management and architecture. Autonomous mobile robots equipped with 3D laser scanners are well suited for the gaging task [#!RAAS2003!#]. To create realistic virtual realities from geometric models, textures, i.e., photos of the environments, have to be acquired and must be precisely mapped onto the scene. This mapping has to be computed automatically from the 3D point cloud of the scanned scene and the acquired photographs.

To compute the correct texture for a scanned scene, four steps are necessary: First, the cameras are calibrated; second, a meshing method generates a triangle mesh of the 3D data, and third, the texture for every triangle is chosen and mapped. Drawing to the screen is done by OPENGL. Finally and fourth, global color corrections are made to remove the systematic color and illumination differences between the individual texture maps. After discussing the state of the art in 3D reconstruction and presenting the robot Kurt3D these four steps are described in detail.