Hybrid Scene Reconstruction by Integrating Scan Data and Stereo Image Pairs

3D geometric models obtained by using laser scanners are often incomplete due to the equipment limitations. In this paper, we propose a novel hybrid reconstruction algorithm, which integrates stereo image pairs to overcome these limitations. This work focuses on a category of scenes that can be approximated by planes (such as manmade scenes). The scan data are first segmented into planar patches, many of which may be incomplete. Stereo image pairs are then incorporated to retrieve the missing data. Specifically, we decide each plane´s extension in the image domain via segment-based stereo matching with disparity layers extracted from scan data. Then, points in the missing area are interpolated by calculating ray-plane intersections. All procedures are automatic except for the range to image registration. The validity of our approach is demonstrated by experiments on real data.