Image Recognition Based On Quantum Evolution RBF Network

It's a hot problem in computer vision research area that how to organize the image data, combine image character reasonably and effectively and import artificial intelligent and knowledge discovery into image classification. Quantum-inspired Evolutionary Algorithm get much more attention for its many advantages, such as easier in the balance between exploration and development, smaller population size, faster convergence, global optimization capability and so on. Quantum Evolutionary is introduced into the optimization of RBF network (Radial Basis Function Network) in this paper, simulation results show the effectiveness of the algorithm. The main work of this paper is as follows:Quantum Evolutionary Algorithm is introduced into the optimization of parameters and structure of RBF network. These parameters include the RBF function center, weights, and variance, structural optimization, it means the network hidden layer and output layer of the part of the link, rather than the entire link, it can reduce time complexity when carrying out image recognition after been optimized. Experiments on Brodatz texture image and SAR image data show that the image recognition accuracy of the RBF network based on quantum evolution is significantly higher than the RBF network based on genetic algorithms.This paper combines the energy characteristics of wavelet transform and the characteristics of gray-scale co-occurrence matrix in the texture image recognition; for SAR images, the Contourlet energy transform domain features and Hu invariant moment features is combined. Simulations of image recognition for both types of images were carried out using the appropriate features based on the quantum evolution RBF network.