摘要
提出了一种基于Hough变换优化的RBF神经网络模式识别新方法,该方法把Hough变换应用于RBF神经网络的参数确定中,实现了RBF神经网络的隐层节点数和数据中心值的自适应获取,提高了RBF神经网络的泛化能力。仿真结果表明:此改进的RBF网络用于模式识别中具有识别能力强,计算量小,识别速度快的优点,具有广阔的应用推广前景。
A new method of pattern recognition based on optimized RBF neural networks using Hough Transform was improved. Hough Transform was applied to the parameters selection and the adaption of the number and position of data centers of RBF neural networks that were realized in this method. Consequently, RBF neural networks designed with this method could generalize well. Experiments results show that the improved RBF neural networks applied in pattern recognition turn out to be a higher accuracy, faster and elegant way. The method possesses high value being generalized.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2006年第1期181-184,共4页
Journal of System Simulation
基金
国家自然科学基金(60274023)
关键词
HOUGH变换
RBF神经网络
函数逼近
模式识别
泛化能力
Hough Transform
RBF neural networks
function approach
pattern recognition
generalization ability
作者简介
李国友(1972-),男,河北玉田县人,博士生,研究方向为图像处理、计算机控制;
姚磊(1979-),男,天津人,硕士生,研究方向为模式识别;
李惠光(1947-),男,博导,研究方向为计算机视觉;
吴惕华(1937-,男,博导,研究方向为高级过程控制。