Automatic surface registration in computer vision systems

Registration of range data (surfaces obtained by scene object scanning) is a fundamental problem of three-dimensional computer vision. Efficiency of computer vision system mainly rely on successful decision of this issue. The principle of registration procedure is automatic detection of surfaces parts which correspond to the same area of the scene. The most successful approach to this problem is based on local surface descriptors. In the paper 3D spin images and oriented spin images are proposed and its higher efficiency in comparison with found to be traditional simple spin images is shown.