Integration of Successive Range Images for Robot Vision

Most conventional approaches to integrate range images from plural view points utilize range images obtained under some artificial restrictions set on movement of the target or measurement system by predetermined scheme, so as to skip determining process of mutual position relationship between range images. This paper presents a method for integration of range images obtained from plural view points under non-restricted condition. In this regards, we try to extract linearly independent four corresponding points automatically from 3D shape data of the same part of the target existing in successive two range images obtained under free condition by our high speed and continuous 3D measurement system [1], then, from those, derive 4x4 affine transformation matrix which gives mutual position relationship between the two range images and finally integrate them automatically in quest of thus determined affine transformation matrix.