The Research Of3D Model Retrieving Based On Feature Combination

The key to study content-based retrieval of3D models is feature extraction.Due to small specialized model library has characteristics of high similarities, low complexity,this paper use the view and NVM to extract the feature, and then compare with other methods.At last, using the feature combination combine the two retrieval method to enhance the effectand to validate the effectiveness of the method with a lot of experimental data.This paper introduces the framework of3D model retrieving based on content, takes thestandardization of the coordinate and preprocessing as the examples, gives several kinds ofapproximate matching methods, introduces the feature extraction and3D model query ofretrieval system, some main points are as follows:Firstly, on basis of analyzing the advantage and disadvantage of view-based retrievalmethods, this paper presents a new method: three views-based retrieving that extract the threeviews projection from3D model surrounded by the cube, experiments emphasis on improvingthe number of retrieved views, and raise the accuracy under the constant retrieved time.Secondly, according to the simple structure, small number of3D model in the modellibrary, this paper extracts the normal vector and mesh (Referred to as NVM) as the feature,and compares with the case model. This method can retrieve3D model rapidly.Finally, combined with the technology of feature combination which put forward byCCCC group in Germany, the paper combines the view-based with NVM-based to reinforcethe description ability. And the proportion and gross of the combined feature have beenproved by experiment. All of these can be applied to the drilling model library, and makefurther improvement of retrieval results.