Wissenschaftliche Publikation

Texture – Based Fusion Between Laser Scanner and Camera for Tunnel Surface Documentation

Publikation aus Digital

G. Paar, A.Bauer and H. Kontrus

7th ISPRS Conference on Optical 3-D Measurement Techniques, October 3-5, 2005, Vienna, Austria. , 1/2005


Close range photogrammetry and laser scanner are widely used for surveying and quality control in construction and mining. Tunnel documentation in the field of refurbishment planning, defects documentation and low-cost frequent data acquisition requires textural resolutions in the mm or sub-mm range on the surface, whereas the necessary structural resolution remains in the cm range. A fusion between laser scanner and camera data is therefore near at hand. We report on a system that is based on a state-of-art laser scanner combined with prosumer camera images, emphasizing the mutual orientation process. Key idea for the fusion is matching between laser scanner reflectivity and RGB camera texture on measurement data level. Both image sources are transformed geometrically and radiometrically to increase similarity. Matching takes place between a rendered pseudo ortho camera and laser scanner reflectivity image on a suitable plane, the tunnel design grid, or the polar laser scanner co-ordinate system. Reliable matching results are used as properly distributed interest points (2D camera image co-ordinates, 3D corresponding laser points) to be fed into a standard orientation procedure that calculates exterior and (optionally) interior camera parameters. RANSAC selects a subset to omit remaining outliers. For data fusion the camera orientation is transformed to the global co-ordinate system using the laser scanner orientation. Laser scanner and the globally oriented image texture independently fill the tunnel design surface grid texture image. These two ortho images are combined using simple syntactic and geometric rules. The proposed mutual orientation and fusion is operational under industrial conditions within the tunnel environment, even if the initial relative orientation between laser scanner and camera is only roughly known.