mehrerer 2,5D Aufnahmen eines Objektes

This thesis presents a complete system of algorithms to obtain a globally consistent surface representation, in the form of a triangle mesh, from a sequence of registered frames taken by a time-ofight (TOF) camera. Surface reconstruction from range images becomes increasingly important in wide elds of applications. The newly developed TOF imaging technology allows acquisition of 3D data in real-time. However, the range measurement from the TOF camera is relatively noisy, which brings in more di culties for the 3D surface modeling. This thesis proposes a solution to deal with the problem. The algorithm makes use of the amplitude images that are also delivered by the camera and imply the con dence of the range measurements. The proposed system comprises a data structure to handle triangle meshes, an algorithm to build up a mesh from a single frame, an integration framework to fuse triangle meshes, and an error controlled data reduction technique. The system leads to a globally consistent triangulation of the entire surface by iteratively integrating two triangle meshes. Thus it can process any number of frames. The integration framework has three major phases. First a mutual approximation deals with overlapping areas of two triangle meshes. Then redundant triangles are removed such that the meshes just meet at their border. Finally the boundaries are clipped applying a developed triangle evaluation function and a topology check. Several tests both on synthetic and real data sets were performed for evaluation of the integration framework, of the reduction technique, and of the entire system.