本文所提出的新算法 BANDC 是 BAyes Normal Dynamic Clustering 的缩写。这个算法的目的是从遥感图象识别地理环境。它的优点是:能由机器自动按照图象估算出分类器的参数;作为一种特殊的 Bayes 分类器,它能够依靠计算机估算出{wi}的先验概率,于是能剔除野点,提高分类器的鲁棒性。这个算法的关键是应用拟合法从灰度等级直方图估算出参数。 The new algorithm proposed in this paper BANDC is the abbreviation of BAyes Normal Dynamic Clustering. The purpose of this algorithm is to identify the geographical environment from remote sensing images. The advantage of this method is that the parameters of the classifier can be estimated automatically by the machine according to the image. As a special Bayes classifier, which can estimate the prior probability of {wi} by computer, it can eliminate the outliers and improve the classifier Robustness. The key to this algorithm is to apply fitting to estimate the parameters from the grayscale histogram.