Automatic Feature Extraction and Semantic Feature Matrix for VRML Building Database Retrieval
Hsuan Ting Chang, Kwang Chang · 2005
A semantic content based information retrieval system for a three-dimensional (3-D) database is proposed in this paper. Here the studied database is composed of 3-D building objects defined by virtual reality modeling language (VRML). First of all, the specific low-level features for building objects are defined and then searched and extracted from the content described in the VRML file. Then, a semantic feature matrix (SFM) is constructed with the middle-level features that determined from the low-level features of all the objects in the database. For a query object, a similar process is applied such that the low-level features and the corresponding semantic vector can be obtained. By multiplying the SFM with the query vector, the scores corresponding to the similarities between the query and all objects in the database can be calculated. Simulation results show that the desired 3-D objects can be successfully and efficiently retrieved with high recall and precision rates