Registration of terrestrial laser scanning data using planar patches and image data

The fully automatic registration of terrestrial scan data is still a major topic for many research groups. Existent methods used in commercial software often use artificial markers which are placed in the scene and measured from each scan position. This is a reliable method to get the transformation parameters, but it is not very efficient. These manual or semi-automated registration techniques should be substituted by new methods in order to make terrestrial laser scanning also profitable for larger projects. In this paper we present a registration method based on the extraction of planar patches from 3D laser scanning data. A search technique is used to find corresponding patches in two overlapping scan positions. Since laser scanning instruments are nowadays often equipped with an additional image sensor, we also use the image information to improve the registration process. Assuming that the calibration parameters of a hybrid sensor system are known, the extracted planar patches can be textured automatically. The correlation between corresponding textured patches can be calculated and the registration method is improved by shifting the patches until they fit best.