基于聚类遗传算法的神经网络规则抽取及应用  

Rule Extraction from Neural Networks Based on Clustering Genetic Algorithm and its Application

作　　者：范艳峰[1,2] 徐朝辉[1] 

机构地区：[1]河南工业大学信息科学与工程学院 [2]西北工业大学计算机学院,西安710072

出　　处：《计算机工程与应用》2006年第23期225-228,共4页Computer Engineering and Applications

基　　金：国家"粮食丰产科技工程"项目(编号:2004BA520A)

摘　　要：提出了一种基于Gabor滤波器和神经网络规则抽取的烘焙面包品质分类方法。滤波器对烘焙面包切片区域灰度图像直接进行小波变换,用能量均值"和均方差!来表示灰度图像的纹理特征,并基于对隐层神经元输出值聚类的遗传算法实现了对面包品质分类的规则抽取。实验结果表明了该方法的实用性和可行性。This paper presents a method in classifying the baking quality-test bread based on Gabor filters and rule extraction from neural networks.Gray images of the field of view are transfonned with different orientations and scales by Filters.The texture features are denoted by the energy standard value and standard variance.Clustering Genetic Algorithm based on the clustering of the hidden units activation values is used to implement the role extraction for baking bread quality classification.The results indicate the practicability and feasibility of the method.

关 键 词：GABOR滤波器 神经网络 规则抽取 聚类遗传算法 烘焙面包 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]