Specifications

Language: English
ISBN-13: 9781615677214
Document type: Conference article (CA)
Conference name: 2007 International Conference on Artificial Intelligence and Pattern
Recognition, AIPR 2007
Conference date: July 9, 2007 - July 12, 2007
Conference location: Orlando, FL, United states
Conference code: 96737
Sponsor: Int. Soc. Res. Sci. Technol. (ISRST)
Publisher: ISRST, PO Box 2464 Tallahassee,, FL 32316-2464, United States
Abstract: In this paper, two image segmentation methods, namely Genetic Simulate based FCM
(Fuzzy C-Means, FCM) image segmentation and Rough Set based FCM image segmentation, are
proposed. In the first methods, the FCM Clustering algorithm, Simulated Annealing algorithm
(Simulated annealing, SA) and Genetic algorithm (Genetic algorithm, GA) are combined to
overcome the drawbacks of conventional FCM segmentation algorithm, namely slow
computation speed and over-dependence on initial value. In this method, the fuzzy cluster center
is coded as a variable length chromosome, genetic operators such as intercross and mutation are
introduced into a Simulated Annealing algorithm as an enhancement, which allows to recombine
solutions produced by individual simulate annealing processes at fixed time intervals. At the
same time Metropolis criterion is taken as a standard for a genetic operation to accept crossover
and mutated individuals, this improves the convergence of the algorithm. Owing to the
complementarities of FCM, SA and GA, this modified algorithm not only can escape from local
minima but also holds higher parallel clustering segmentation capability concurrently. In the
second method, Rough Set theory is used to optimal the performance of FCM in analyzing
vagueness and uncertainty inherent in building clustering set. By reduction technique (the core of
Rough Sets), those redundant initial cluster centers in the initial cluster set are eliminated this is
very useful for improving the convergence of the FCM algorithm. Experimental results
demonstrate the efficiency and the effectiveness of the proposed methods.
Number of references: 21
Main heading: Clustering algorithms
Controlled terms: Artificial intelligence - Genetic algorithms - Image segmentation -
Pattern recognition - Rough set theory - Simulated annealing - Uncertainty analysis
Uncontrolled terms: Initial cluster centers - Metropolis criterion - Reduction techniques
- Segmentation algorithms - Segmentation methods - Simulated annealing algorithms
- Vagueness and uncertainty - Variable length chromosome
Classification code: 716 Telecommunication; Radar, Radio and Television - 721 Computer
Circuits and Logic Elements - 723 Computer Software, Data Handling and Applications -
921 Mathematics - 921.4 Combinatorial Mathematics, Includes Graph Theory, Set Theory -
922.1 Probability Theory
Database: Compendex
Compilation and indexing terms, © 2013 Elsevier Inc.
3.
Accession number: 20131816300259