Abstract

The reusability of three-dimensional models has lead to a proliferation in the number and size of 3D object repositories. These typically store objects in the some polygon mesh file format and allow selection by means of categories and keywords. While keyword searches are useful in many circumstances, they suffer from the subjectivity, ambiguity and resolution difficulties inherit in natural language. These problems become more acute sa database seizes increase and may force users to sift through large numbers of returned thumbnail images.

On solution is to query the object database by example. The user constructs a rough low-resolutions query model with the shape characteristics of the target object. This is then compared against entries in the database and the closest matches returned in order of similarity. The core of this method is the matching algorithm.