基于模糊C均值聚类改进的最大似然分类法

基于参数密度分布模型的最大似然分类法(MLC)是遥感影像经典分类方法之一,它具有清晰的参数解释能力、易于与先验知识融合和算法简单而易于实施等优点,但是由于遥感数据具有高度的模糊性和随机性,使得贝叶斯(Bayes)判别函数中的均值向量和协方差矩阵很难准确确定。因此首先利用模糊C均值聚类得到模糊划分矩阵,然后基于模糊划分矩阵计算出每一个聚类类别模糊均值和模糊协方差矩阵,并利用模糊均值和模糊协方差矩阵来代替贝叶斯判别函数中的均值向量和协方差矩阵从而建立一个新的判别函数,最后与传统的最大似然分类结果进行比较,结果表明改进后的最大似然分类法在总体精度、Kappa系数均优于传统的最大似然分类方法。

Improvement for Maximum Likelihood Classification Based on Fuzzy C-means

Based on parametric density distribution model, Maximum likelihood classification (MLC) might be one of the classic methods for remote sensing image classification. It has several distinct advantages, such as clear parametric interpretability, feasible integration with prior knowledge based on Bayesian theory, and relative simple realization, etc. However, due to some fuzziness and randomness for remote sensing data, it is difficult to determine the Bayesian discriminant function of the mean vector and covariance matrix. So in this paper, Firstly fuzzy partition matrix are calculated by Fuzzy C-means clustering, Then based on the fuzzy partition matrix calculated fuzzy mean and fuzzy covariance matrix for each cluster type, and establishing a new discriminant function by using fuzzy mean and fuzzy covariance matrix instead of the Bayesian discriminant function of the mean vector and covariance matrix. Finally Comparison with the results of traditional maximum likelihood classification, showing the improvement of Maximum likelihood classification method is better than the traditional maximum likelihood classification method in the Overall accuracy, Kappa coefficients.