It is the ability to identify and classify various objects that worth interest from mass high resolution remote sensing images that is a kind of technical requirement appealing for broad applicational prospects. The experiment takes Matlab as the experimental platform, and applies three image texture feature extraction algorithms, respectively, Gabor filter, Gauss-Markov random field model (GMRF) and the gray level co-occurrence matrix (GLCM) to extract features from Brodatz optical database images which is a popularly used sample set for texture image classification. Then following the method of constructing multi-kinds classifier based on binary classification SVM, the paper has completed the comparative experiments of two groups by classifying optical texture images with SVM classifiers. At last, by the proposed fusing multi-feature texture classification to generate texture look-up table method, in both experiment groups, the method brought forward in the article is verified to be able to obtain better classification effects on Brodatz optical texture set.%能够从大量高分辨率遥感图像中识别出各种感兴趣的目标并进行归类,是一种具有广泛应用前景的技术需求.实验以MATLAB为平台,应用Gabor滤波器、高斯马尔柯夫随机场(GMRF)和灰度共生矩阵(GLCM)三种纹理图像特征提取算法对当前广泛应用于纹理图像分类的样本集brodatz光学数据库图像进行特征提取;然后在二分类支持向量机的基础上构造多类分类器的方法,完成了利用支持向量机SVM分类器对光学纹理图像进行分类的两组对比实验;最后通过提出的融合多特征纹理分类生成纹理查找表的方法,在两组对比实验中验证了该文提出的方法能够在Brodatz光学纹理集上得到较好的分类效果.
展开▼