Research On Three-dimensional CAD Models Matching And Retrieval

In recent years, with the development of3D modeling technology and theimprovement of computer hardware and software, the application of threedimensional CAD model is more and more extensive. Especially in the field ofengineering analysis and product manufacturing, three dimensional CAD modelsmatching and retrieval has become a new research direction. For a long time, themethods based on texts and keywords of3D model retrieval are used widely, butthese methods are limited widely. Therefore how to retrieval3D models basedcontent is worthy further study and explore.There are many different format files of three dimensional CAD models. Inthis paper, model files of STL format are considered. A new method of3D modelretrieval based on projected area at mesh vertex is proposed. The method uses theinformation of the STL format files adequately. First, sum the projected area onvertical plane of the normal vector at mesh vertex, then normalized the list of theprojected area distributions and transfer these data by Fourier transform method.Then the result is defined as3D model’s feature vector which can be used tocalculate the similarity of different models. Experiments were conducted toevaluate the proposed algorithm utilizing the ESB database.For the three dimensional models, the global conceptual similarity is veryimportant, but we also need to consider the local features. The SIFT algorithm isa method for the two-dimensional image feature matching and widely used incomputer vision and image processing. This paper presents a description of theSIFT algorithm based on multi-view characteristics of the three dimensionalmodels matching and retrieval methods. The features of CAD models areextracted from two-dimension range images of the model viewed from uniformlysampled locations on a view sphere. Then these features are normalized andcalculating the distance between the feature vectors by using the KL divergencealgorithm. Finally, the results of model matching are propelled.At present, the research for building three dimension CAD model databaseis still very less. A three dimensional CAD model retrieval prototype system hasbeen developed in this paper and some basic problems in the proceeding arediscussed.