Fast similarity search in databases of 3D objects

Abstract

Given a database D of three dimensional (3D) objects and a target object Q, the similarity search problem (also known as good-match retrieval) is defined as finding the objects D in D that approximately match Q, possibly in the presence of rotation, translation, node insert, delete and relabeling in D or Q. This type of query arises in many AI applications. We study the similarity search problem and a class of related queries. We present a computer vision based technique called geometric hashing for processing these queries. Experimental results on a database of 3D molecules obtained from the National Cancer Institute indicate the good performance of the presented technique.