Study on Multi-Views Point Clouds Registration

A new registration method for multi-views point clouds scanned by optical measurement instruments is proposed. Firstly, By means of close-range photogrammetry technology, a point set of artificial marks pasted on surface of the measured object was reconstructed, to form a global locating frame. Structured light scanner acquired partial marks and dense point cloud of the same local area, Distance map of global frame marks and local marks were created by calculating the distance between points in each point set respectively. Because same pair of marks has same distance in difference point set, 3 pairs of corresponding marks could be found by searching the smallest sub graph isomorphism. The largest sub graph isomorphic could be achieved by adding nodes to the smallest sub-graph, as a result a set of matched marks are found. The rigid transformation matrix can be figured out by solving minimum square distance. Secondly, an accurate registration method based on iterative closest point algorithm is proposed, the standard Iterative Closest Point is improved by means of establishing matched pair points with weighting function. The proposed method is verified through experiment.