Abstract
This paper presents a novel algorithm "Twig Match" for feature based shape retrieval systems. The algorithm exploits recent advances in computational methods for subgraph isomorphism, in order to enable databases containing many thousands of components to be searched in less than a second. A face adjacency graph representation is created from a B-Rep model, allowing model comparison to be treated as a labelled subgraph isomorphism problem. This paper describes an experimental implementation which allows interactive specification of a target "feature". By selectively including geometric filters, on faces and relations between neighbouring faces, the algorithm can ensure that matching topology is not incorrectly identified as matching geometry, while also offering users the ability to improve the precision of both query and results. Experimental results show that Twig Match accurately retrieves matching and similar sub-parts from collections at speeds suitable for interactive applications .
Original language | English |
---|---|
Publisher | American Society of Mechanical Engineers(ASME) |
DOIs | |
Publication status | Published - 21 Aug 2016 |
Keywords / Materials (for Non-textual outputs)
- Twig Match
- shape retrieval systems
- subgraph isomorphism
- face adjacency graph
- B-Rep models
- geometric filters
- matching topology