Title: A novel hybrid model for image classification

Authors: Yi-Ming Liu, Min Yao, Rong Zhu

Addresses: School of Computer Science and Technology, Zhejiang University, 38 Zheda Road, Hangzhou, Zhejiang 310027, China. ' School of Computer Science and Technology, Zhejiang University, 38 Zheda Road, Hangzhou, Zhejiang 310027, China. ' School of Computer Science and Technology, Zhejiang University, 38 Zheda Road, Hangzhou, Zhejiang 310027, China; School of Information Engineering, Jiaxing University, 118 Jiahang Road, Jiaxing, Zhejiang 314001, China; State Key Laboratory for Novel Software Technology, Nanjing University, 22 Hankou Road, Nanjing, Jiangsu 210093, China

Abstract: Recently, biological intelligent computing gains more and more attention in analysing large-scale real world datasets. Because the performance of the support vector machine (SVM) classifier is always degraded by poor feature subsets and inappropriate parameters for training, an improved quantum-behaved particle swarm optimisation (IQPSO) is introduced to optimise the features and parameters synchronically, aiming to improve the generalisation of the SVM classifier. That is, a novel hybrid image classification model by combing SVM and IQPSO, called as IQPSO_SVM is presented in this paper. Experimental results show that the proposed IQPSO_SVM improves the classification accuracy greatly compared to the traditional SVM with grid search, and outperforms such SVM based on genetic algorithm (GA_SVM) without accuracy loss.

Keywords: image classification; feature selection; parameter estimation; support vector machines; SVM; quantum-behaved particle swarm optimisation; QPSO; PSO; genetic algorithms.

DOI: 10.1504/IJCSE.2011.041217

International Journal of Computational Science and Engineering, 2011 Vol.6 No.1/2, pp.96 - 104

Published online: 18 Mar 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article