对两种类型的纹理图像，各选取30幅，计算灰度共生矩阵，提取特征向量；构件BP神经网络分类器，用每种类型的前20幅对分类器进行训练，后10幅用来测试分类效果。-Two types of texture image for each select 30 calculated GLCM, extracting feature vectors　member BP neural network classifier, before use each type of 20 pairs classifier is trained to test after 10 classification results.