THREE-DIMENSIONAL POINT-CLOUD REGISTRATION USING A GENETIC ALGORITHM AND THE ITERATIVE CLOSEST POINT ALGORITHM

We present a method for three-dimensional surface registration which utilizes a Genetic Algorithm (GA) to perform a coarse alignment of two scattered point clouds followed by a slight variation of the Iterative Closest Point (ICP) algorithm for a final fine-tuning. In this work, in order to improve the time of convergence, a sampling method consisting of three steps is used: 1) sample over the geometry of the clouds based on a gradient function to remove easily interpolating singularities; 2) a random sampling of the clouds and 3) a final sampling based on the overlapping areas between the clouds. The presented method requires no more than 25% of overlapping surface between the two scattered point clouds and no rotational or translational information is needed. The proposed algorithm has shown a good convergence ratio with few generations and usability through automated applications such as object digitalization and reverse engineering.