基于灰度共生矩阵的板材纹理模糊分类器设计

Design of Fuzzy Classifier for Wood Board Texture based on GLCM

针对板材表面的纹理特征,设计了直纹、弯纹二维模糊逻辑分类器。首先对灰度共生矩阵提取的板材特征量进行散点分布分析,提取出用于纹理分类的能量和相关性基本特征;设计能量、相关性与纹理的模糊论域及隶属度函数,完成了特征量与纹理类别的模糊化;通过提取样本的模糊规则和Mamdani重心法,完成模糊推理,并得到纹理分类查询表。应用设计的纹理模糊分类器对样本库中100幅直纹和弯纹样本进行分类实验,分类总体正确率达到90.5%。

We designed a classifier based on fuzzy logic for straight and curved texture. We extracted energy and correlation from the features of gray level co-occurrence matrix （GLCM） by analyzing its scatter, and then designed the fuzzy domain and the membership function for the extracted features and texture patterns. We constructed fuzzy reasoning and the lookup ta- ble by the extracted fuzzy rules and Mamdani gravity method. The experiments were conducted with 100 straight and curved textures with the classification rate of 90.5%.