3D Part Identification Based on Local Shape Descriptors

Published

August 19, 2008

Author(s)

Xiaolan Li, Afzal A. Godil, Asim Wagan

Abstract

This paper explores 3D object recognition based on local shape descriptor. 3D object recognition is becoming an increasingly important task in modern applications such as computer vision, CAD/CAM, multimedia, molecular biology, robotics, and so on. Compared with general objects, CAD models contain more complicated structures and subtle local features. It is especially challenging to recognize the CAD model from the point clouds which only contain partial data of the model. We adopt the Bag of Words frame to do the partial-to-global 3D CAD retrieval. In this paper the visual words dictionary is constructed based on a classical local feature descriptor: spin image. The method is tested on Purdue Engineering Shape Benchmark. Furthermore, several experiments are performed to show how the size of query data and the dissimilarity measurement affect the retrieval results.

Conference Dates

August 19-21, 2008

Conference Location

Gaithersburg, MD

Conference Title

Performance Metrics for Intelligent Systems Workshop Proceedings (PerMIS)

Pub Type

Conferences

Download Paper

Local Download

Keywords

bag of words, CAD model retrieval, spin image

Citation

Li, X. , Godil, A. and Wagan, A. (2008), 3D Part Identification Based on Local Shape Descriptors, Performance Metrics for Intelligent Systems Workshop Proceedings (PerMIS), Gaithersburg, MD, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=890064 (Accessed October 31, 2024)