摘要
用基于启发式搜索的结构学习算法,学习得到多光谱影像的贝叶斯网络结构,分析了TM的波段(特征)间条件独立性假设的合理性,给出了贝叶斯推理中后验概率计算的公式,并通过和最大似然法对比试验分析了简单贝叶斯网络应用于多光谱影像分类的优势。
From the structure learning methods of search heuristics, we constructed a NaY ve Bayesian Network for the multi-spectral image. We proved that the hypothesis of conditional independence between TM bands was successful, and analyzed the feasibility of applying NaI ve Bayesian Networks to the classification of multi-spectral image through experiment.
出处
《地理空间信息》
2009年第2期15-18,共4页
Geospatial Information
基金
国家重点基础研究发展计划资助项目(2006CB701303)
关键词
遥感
贝叶斯网络
结构学习
条件独立性假设
多光谱影像
分类
remote sensing
Bayesian networks
structure learning
hypothesis of conditional independence
multi-spectral image
classification