基于简单彩色纹理特征的高分图像SVM分类方法研究

Method Research of SVM Classifying High-resolution Remote Sensing Images based on Simple Color Texture Features

彩色纹理特征结合了传统意义上的灰度纹理特征和光谱特征,将其引入高分图像的地物分类算法中,可以丰富特征提取手段、提高分类正确率。首先详细介绍了彩色纹理特征的提取方法和SVM算法的数学原理,然后分析了基于SVM的高分辨率遥感图像分类方法的模型建立以及分类器设计,最后以作者所在高校为例,验证了该方法在高分辨率遥感图像地物分类上的有效性,并通过混淆矩阵和Kappa系数评价了分类性能。

Color texture feature is a combination of the traditional gray texture characteristics and spectral characteristics,and it is introduced into classification algorithm of high resolution remote sensing image,which can enrich feature extraction methods and improve the classification accuracy.Firstly color texture feature extraction method and the mathematical principle of SVM algorithm are successively introduced,and then model design and classifier design are given for the high resolution remote sensing image classification method based on SVM,finally taking the universities author worked for example,the validity of method is proven on the high resolution remote sensing image feature classification,and through the confusion matrix and kappa coefficient to evaluate the classification performance.