A concept for feature based data registration by simultaneous consideration of laser scanner data and photogrammetric images

摘要

Surface measurement of complex geometric objects by means of optical techniques, such as laser scanning or photogrammetry, requires multiple views from various stations. The data sets of the views, e.g. point clouds and photogrammetric images, are given in individual coordinate systems. For a complete representation of the objects, the views have to be transformed into one common coordinate system. This task is denoted as registration and involves an estimation of the transformation parameters between different coordinate systems.rnIn this paper, a concept for a pairwise feature based registration of hybrid laser scanner data sets is proposed. The concept includes the feature extraction and description by simultaneous consideration of laser scanner data and photogrammetric images, as well as the recognition of consistent feature correspondences. The features are viewpoint invariant and are described with a rectified image patch on planar surface geometry in object space.rnTwo different data sets are used to demonstrate that the proposed concept for feature extraction, feature description and correspondence search is successful to register hybrid laser scanner data sets with a high level of automation.