Contribution to knowledge

• We have introduced a new way of handling the data acquired by light striping in the form of a profile map, which allows us to process disparity maps and laser scanned data consistently as real valued images ${\bf u}_k$ on the parametric domains S_k, $k=1,\ldots ,L$.
• We have proposed a new and fast method to update the corresponding points on S_k for the simultaneous matching of multiple maps ${\bf u}_k$, $k=1,\ldots ,L$, acquired from different viewpoints without exactly known corresponding points.
• We have improved the performance of the registration algorithms as regards the hierarchical processing, the interpolation errors in smooth areas, and the weighting images for the direction of the surface normal, for the edge areas, and for the precision of the distance between the corresponding points.
• We have experimentally shown that for the registration of profile maps, the IPP algorithms with bicubic and bilinear interpolation give more accurate results and have larger pull-in ranges than the ICP ones for triangle sets, for bilinearly interpolated quadrilaterals, and a variation of the ICP one minimizing the distance from the points of a data set to tangent planes at the points of intersection between the surface normals of the data set and a bilinearly interpolated surface constructed from the points of the other data set.
• We have proposed a new and accurate method for the self-calibration of the light striping system based on matching multiple profile maps and using the IPP algorithm.
• We have developed a new technique for the initial estimation of the parameters of a cylinder from ${\bf u}_k$.
• We have presented techniques for the modeling of multiple maps ${\bf u}_k$ including the splitting and merging of compatible overlapping planar patches of different maps and the tracing of the border of a merged patch on multiple domains S_k, $k=1,\ldots ,L$.
• We have carefully analyzed the propagation of measurement and calibration errors to the registration parameters and to the modeled maps and thereby derived the precision estimates for two combined registration and modeling strategies.
• We have shown in a test case that our statistical analysis yields higher precision for the strategy where the maps are first registered and the model is built afterwards than for the strategy where the maps are registered against the model updated whenever a new map has been registered.
• We have demonstrated how most of the developed algorithms can be formulated using image algebra and implemented efficiently using MATLAB software.
• We have proved the performance of the algorithms developed in various experiments with synthetic and real data and developed the techniques needed for generating synthetic profile maps.