全极化SAR数据的最大后验概率分类

Classification of Full-polarimetric Synthetic Aperture Radar Data with Maximum a Posteriori

结合后验概率对分类的影响和全极化SAR数据特点,提出了一种全极化SAR数据分类方法。首先将全极化SAR数据的协方差矩阵转换为9个服从正态分布的强度量;然后通过迭代分类计算类别出现的概率,对9个强度量进行基于最大后验概率的分类。以黑龙江省逊克县境内的一景ALOS PALSAR全极化数据为例,用该方法进行分类,总体精度和Kappa系数分别达到81.34%和0.84,优于传统的最大似然分类方法。

Considering the influence of the posterior and the statistic distributions of full-po- larimetric SAR data, we proposed a new classification method of full polarimetric SAR data. First, the covariance matrix of polarization SAR data was converted to nine intensity quanti- ties with normal distribution. Then, the probability of occurance for each class was calculat- ed with iterative initial classification. Finally, the nine intensity images were classified with max- imum likelihood classification method taking the probabilities of occurance for the classes into ac- count. We applied the developed method to the ALOS PALSAR full-polarimetric data of Xunke County, Heilongjiang Province. The overall accuracy is 81.34% and the Kappa coefficient 0. 84. The developed method showed higher accuracy than that from the traditional maximum likelihood classifier. This indicates that our method can improve the accuracy of classification.