3DMOD

3D modeling from image information has its roots in photogrammetry. Automation is still not the primary goal. The focus in photogrammetry is set on accuracy issues. In the last few years progress in automation and robustness in the area of 3D modeling was made in the computer vision community. Nowadays, 3D scanning devices like laser scanners are becoming popular for the 3D reconstruction of existing objects. But, there is still a lack of algorithms to fully exploit redundant information in image sequences or to fuse the information from complementary sensor systems. Thus, in this project we investigate the combination of knowledge from different fields into one consistent framework using the available input information at the best possible rate. Important aspects will be:

• Direct 3D Reconstruction from Multiple Images
• Consistent 3D Reconstruction from Range Data
• Voxel Coloring and Space Carving on 3D Hardware
• Fusion of 3D Reconstruction and Recognition Techniques
• Fusion of Active and Passive Sensors (i.e. photogrammetry and laser scanning)
• High Quality 3D Reconstruction (High Geometric and Radiometric Reconstruction)