Curvature feature based shape analysis

Yufeng Chen*, Fengxia Li, Tianyu Huang

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

2 Citations (Scopus)

Abstract

A novel method is presented to evaluate the similarity of shapes based on the curvature features and their distribution. Firstly the curvature information is used to define the curvature features, which are learned and searched by the proposed statistic method. Secondly the structural features of each pair are measured, so that the distribution of the curvature feature can be further measured. Taking both advantages of the local shape context analysis and the global feature distances optimization, our method can endure large nonrigid distortion and occlusion. The experiments, which have been implemented on the MPEG-7 shape database, show that this method is efficient and robust under certain shape distortion. Another experiment on the abnormal behavior detection shows its potential in shape detection, motion tracking, image retrieving and the related areas.

Original languageEnglish
Title of host publicationAdvanced Intelligent Computing Theories and Applications
Subtitle of host publicationWith Aspects of Theoretical and Methodological Issues - 4th International Conference on Intelligent Computing, ICIC 2008, Proceedings
Pages414-421
Number of pages8
DOIs
Publication statusPublished - 2008
Event4th International Conference on Intelligent Computing, ICIC 2008 - Shanghai, China
Duration: 15 Sept 200818 Sept 2008

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume5226 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference4th International Conference on Intelligent Computing, ICIC 2008
Country/TerritoryChina
CityShanghai
Period15/09/0818/09/08

Fingerprint

Dive into the research topics of 'Curvature feature based shape analysis'. Together they form a unique fingerprint.

Cite this