利用核密度估计改进遥感图像贝叶斯分类法  

Improved Bayes Classification Method for Remote Sensing Images by Kernel Density Estimation

作　　者：刘琳[1] 杨为民[1] 赖巧玲[1] 

机构地区：[1]西南林学院,昆明650224

出　　处：《东北林业大学学报》2008年第10期53-55,共3页Journal of Northeast Forestry University

摘　　要：把非参数核密度估计方法引入到遥感图像贝叶斯(Bayes)分类问题中,对各类的分布密度函数进行非参数核密度估计,从而改进了Bayes分类方法。通过对遥感图像实例分类,与传统Bayes分类方法和其它统计分类方法比较,分类精度得到了提高。该方法解决了其他分类方法单中心的局限,既保留了核密度估计法理论上的优点和平滑性,又适合云南由于地形和光照影响而产生的同一地类在相空间中是多中心的特点,具有一定的应用推广价值。The method of nonparametric kernel density estimation is applied to the Bayes classification of remote sensing images. The Bayes classification method is improved by performing the nonparametrie kernel density estimation to various types of distribution functions. The accuracy of the improved classification method increased by comparing the classification results of remote sensing image samples by the improved method, the traditional Bayes classification method, and other statistical classification methods. The method overcomes the single-center limitation of other methods while preserves the smoothness and other theoretical advantages of the kernel density estimation. It works well in Yunnan for the same type of earth surface which may generate multiple centers in the phase space due to the impact of landform and sunshine. Thus it has the potential of being widely applied.

关 键 词：遥感图像分类 BAYES分类 非参数估计 核密度估计 

分 类 号：S718.43[农业科学—林学]