基于人工神经网络的织物疵点聚类分析  

Fabric defect clustering analysis based on artificial neural network

作　　者：杨晓波[1] 

机构地区：[1]浙江财经学院信息分院,浙江杭州310018

出　　处：《纺织学报》2011年第9期29-33,共5页Journal of Textile Research

基　　金：高等学校博士学科点专项科研基金资助项目(99025508)

摘　　要：为解决织物疵点分类过程中由于人为因素造成分类准确率不高的问题,提出一种基于人工神经网络的织物疵点分类方法。首先利用灰度共生矩阵提取织物疵点图像的纹理特征参数;然后阐述前馈BP神经网络的拓扑结构,并提出该网络的具体训练过程;最后利用人工神经网络对真实织物疵点样本进行分类。实验采用5类织物样本,网络训练完成后得到实际分类的疵点数据,并利用该数据进行织物疵点分类。结果表明,人工神经网络可以对常见类型的织物疵点进行分类,分类准确率较高,从而验证了该方法的可行性。For solving this problem that classification accuracy is not high due to human factors in the process of fabric defect classification,a method is proposed to classify fabric defects based on artificial neural network.Firstly,gray co-occurrence matrix is used to extract texture feature parameters from fabric defect image.Then,the topology structure of forward feedback BP neural network is narrated,and also indicated the training process in detail.Finally,the BP artificial neural network is applied to fabric defect classification,and five kinds of fabric samples are used in the experiment.The defect data for classification can be gotten through neural network training process.These data can be used to classify fabric defects.The results show that artificial neural network can be used to classify the common defects with higher accuracy,verifying the feasibility of this method.

关 键 词：人工神经网络 特征提取 模式识别 疵点分类 

分 类 号：TP311.131[自动化与计算机技术—计算机软件与理论] TS101.9[自动化与计算机技术—计算机科学与技术]