The Research On Image Recognition Based On RBF Neural Network

Radial Basis Function Neural Network (RBFNN) is widely applied in image processing, because it can process a large amount of information, and has the simple structure, strong fault tolerance, fast convergence speed and global nonlinear approximation ability. The RBFNN has big potential in image recognition, it is considered to be a powerful tool for solving image recognition.The main contents are as follows:(1) The RBFNN trained by its traditional learning algorithm holds the low classification accuracy, the traditional k-means algorithm has sensitivity to the initial clustering center, and the clustering results are easily changeable as different initial inputs. In order to solve these problems, an improved learning algorithm based on improved k-means algorithm is proposed for the RBFNN. In the new algorithm, k-means algorithm is optimized with the Subtractive Clustering Algorithm to reduce the clustering sensitivity, and the structure of the RBFNN is determined by mean of the optimized k-means algorithm.The simulation results demonstrate the practicability and the effectiveness of the new algorithm. The improved learning algorithm for the RBFNN is employed to design the network classifier, is further applied in vehicle identification, the experimental results show that the classification precision is improved.(2)A learning algorithm based on combined clustering algorithm is proposed for the RBFNN. The new algorithm optimizes initial selection of improved fuzzy c-means clustering algorithm with the Subtractive Clustering Algorithm to reduce the clustering sensitivity to initial selection, then structure of the RBFNN is designed with the improved fuzzy c-means clustering algorithm and is simplified by pruning techniques.The simulation results demonstrate the effectiveness of the new algorithm.The improved algorithm for the RBFNN is used to design the network classifier,and is further applied in vehicle identification.The experimental results show that the classification precision is improved to some extent.