Algorithm for Finding Registration Sphere Center in Merging Multiple-View 3D Point Clouds

From the measured sphere data,the normal of each point is calculated according to its k-nearest neighbors,and every normal vector is adjusted to the same orientation as that of the scanning frame center,i.e.pointing towards the sphere center.Then,all points on the sphere surface are extracted with the criterion that their diametrically opposite points are located on the sphere,centered at the normal vector of scanning frame center with the radius taking into account the scanner’s measurement error.Finally ,the centroid of diametrically opposite points located on the lengthened normal of all extracted points determines the center of registration sphere.Experiments on real data show the high accuracy and good stability of this algorithm.