<FONT face=Verdana>Extracting shape feature is the key point of content-based 3d model retrieval. For this problem

a new approach is proposed for constructing 3d shape descriptor using spatial symmetry transformation. The main idea of this method is that symmetry is an important cue for all objects. At first

symmetry transformation of pair point is calculated

and the points come from spherical uniform sampled. Then symmetry space is built by pair symmetry relation. And using space clustering gets region symmetry description of 3d model. At last

range region symmetry descriptors in descend order to form shape descriptor of whole 3d model. This descriptor is robust to 3d model simplification

subdivision and other geometry operation. Retrieval response time is also satisfied. The effectiveness of the proposed approach is verified by three different sides that are feature similarity comparing

retrieval efficiency and results demonstration.