A method characterized by its excellent invariability in translation

rotation

scale and angle of view for recognizing motion object in 3D space was presented. The contour curve of object was parameterized by an affine invariable and the parameterized curve of object waiting for recognition was projected orthogonally. So it could keep better correlation with the object of the same type in store. Finally

normalized related function was regarded as the criterion of recognition. The experimental results show that the correlation of objects is over 95% in the 3D's change of pose

the proposed method can recognize objects of multiform changes with superior discrimination rate.