基于概率神经网络的遥感图像分类MATLAB实现  

Matlab Realization of Sensing Image Classification Based on Probabilistic Neural Network

作　　者：李海洋[1] 范文义[1] 

机构地区：[1]东北林业大学,哈尔滨150040

出　　处：《东北林业大学学报》2008年第6期55-56,62,共3页Journal of Northeast Forestry University

基　　金：国家高技术研究发展计划项目(2006AA12Z104);黑龙江省攻关课题(GC04B713)

摘　　要：在遥感图像分类研究方面人工神经网络是一种有效途径,与传统的分类方法相比概率神经网络具有许多优良的性能,因此利用神经网络工具箱构建了概率神经网络,经对比分类精度选取最优SPREAD=0.009,并对一幅TM假彩色遥感图像通过训练后,仿真输出能真实地反映原始图像的特征,其分类总精度为82.62%,Kap-pa系数为0.7821,结果表明:分类精度能够满足遥感图像分类的需要。The artificial neural network is a kind of effective way in remote sensing image classification. Compared with traditional classification methods, probabilisfic neural network has many superior performances. Therefore, the probabilistic neural network was constructed by utilizing the neural network toolbox and choosing the optimal SPREAD （0. 009） compared with the classified precision. A TM false color composition image was trained and a satisfactory simulation result was got by applying the probabilistic neural network. The total classification precision is 82.62% , and Kappa coefficient is 0. 7821. The result shows that the classification precision can meet the demand of remote sensing image classification.

关 键 词：MATLAB 概率神经网络 分类 精度 Kappa系数 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构] S771.8[自动化与计算机技术—计算机科学与技术]