View planning for 3D reconstruction using time-of-flight camera data as a-priori information

Solving the next best view (NBV) problem is an important task for automated 3D reconstruction. An NBV algorithm provides sensor positions, from which maximal information gain about the measurement object during the next scan can be expected. With no or limited information available during the first views, automatic data driven view planning performs suboptimal. In order to overcome these inefficiencies during startup phase, we examined the use of time-of-flight (TOF) camera data to improve view planning. The additional low resolution 3D information, gathered during sensor movement, allows to plan even the first scans customized to previously unknown objects. Measurement examples using a robot mounted fringe projection stereo 3D scanner with a TOF camera are presented.