Bayesian Network Classification for Aster Data Based on Wavelet Transformation

摘要

In this study, Bayesian networks are considered to be a classifier for the remote sensing image named Aster data, which involves 15 bands. Six bands, which have different spatial resolutions, are selected to be the attributes in Bayesian network classifier. The sample data from Aster image that is fused by wavelet transform is used to train Bayesian network classifier. Before the above-mentioned processing, the attributes from the transformed image should be normalized by some equal width schemes. Then the learning scheme process is used to acquire the structure of Bayesian networks from the training data set. The relationship of the attributes among all the constituents of the imagery data is mined through the Bayesian networks. To evaluate this classifier, a comprehensive study of the performance is investigated based on the training data set and the independent test data sets. The result shows that Bayesian network performs well on remote sensing imagery data.