遥感波段与样本组合及贝叶斯网络结构变化分析

贝叶斯网络表达了输入数据与分类结果之间的依赖关系,网络结构则表达了节点之间的条件概率状态。遥感数据的贝叶斯网络结构训练涉及先验知识和样本数量两个方面,是贝叶斯网络结构分类的重要环节。该文以应用目标和遥感数据波段的物理意义为先验知识指导,进行了贝叶斯网络结构建立中的遥感数据波段数和样本数的优化组合实验,为贝叶斯网络在遥感数据分类方面提供了基础性实验结果。

Analysis of Different Remote Sensing Bands and Training Samples Combinations in Bayesian Networks Structure Change

In recent years,the Bayesian network has been used in many study fields as a data-mining tool,but the study using it to process remote sensing data is very seldom seen.Bayesian network classification expresses the dependence of input data and classification result,the network structure can well express the state of conditional probability between different nodes.The remote sensing data Bayesian networks structure training involves the prior knowledge and amount of samples,which is important tache of Bayesian network classification.This paper depends on the application target and remote sensing bands physical meaning,does the experiment of different combinations of remote sensing bands and amount of samples,the experiment results can provide basic guidelines to the Bayesian network classification.