Automatic registration of partial overlap three-dimensional surfaces

The iterative closest point (ICP) algorithm is a widely used method for registering three-dimensional surface data. The quality of registration obtained by this algorithm depends heavily on a good prior initial estimation and choosing good pairs of corresponding points in the input datasets. If too many points are chosen from featureless regions of the data, the algorithm converges slowly, inds the wrong pose, or even diverges, especially without good initial alignment in the input point datasets. This paper proposed a robust point selection strategy for registration procedure which is based on selecting a small set of salient feature points whose neighborhoods are highly distinguishable from each other. Then a fast matching algorithm is developed based on distance matrix comparisons to select the optimal correspondence point set and bring the two point sets into a good alignment. Experiments confirm that the proposed salient feature-based algorithm is resulting in a great reduction of the complexity of the matching process and the computing time is only a few seconds.