利用偏态二叉树最小二乘支持向量机进行高光谱遥感影像分类

Application of skew binary tree multi-class LS-SVM classifier in hyperspectral remote sensing image classification

本文采用偏态二叉树最小二乘支持向量机的方法来进行高光谱遥感影像的分类,分别采用交叉验证法、遗传算法、粒子群优化算法来优化高斯径向基核函数的2个参数。以北京昌平小汤山地区的高光谱影像为例,对这3种参数优化方法进行比较验证,其中基于交叉验证法优化参数所获得的分类精度最佳。实验也证明了本文采用的分类方法明显优于其他传统的分类方法,有效地提高了高光谱数据的分类精度。

Three different parameter optimization methods of skew binary tree multi-class Least Squares Support Vector Machines classifier（LS-SVM） including cross validation, genetic algorithms and particle swarm optimization were proposed for hyperspectral remote sensing image classification in this paper, in order to optimize the parameters of Gaussian radial basis kernel function respectively. They were tested on the hyperspectral data of Beijing Changping Xiaotangshan area. The experimental results showed that, the parameter optimization method based on cross validation had the best classification results, whose overall accuracy and kappa coefficient reached 96.76% and 0. 9627 respectively. The method proposed by this paper was proved significantly better than other traditional classification methods with its higher classification precision of hyperspectral data.