利用云模型实现FSVM遥感影像分类的隶属度算法

USING CLOUD MODEL TO REALISE MEMBERSHIP ALGORITHM OF FSVM CLASSIFICATION OF REMOTE SENSING IMAGE

针对遥感影像分类面临的数据边界模糊性以及遥感信息解译过程不确定性的问题,结合模糊支持向量机在分类应用中可以有效避免噪声样本干扰的特点,提出一种基于云模型求解模糊支持向量机隶属度的方法。该方法通过无需隶属度的逆向云算法输入样本的定量位置得到样本类别的数字特征,再根据正向云算法计算得到每个样本对其定性类别的隶属度。实验结果表明,采用基于云模型隶属度的模糊支持向量机对遥感影像的分类方法是可行的,并能够有效提高对遥感影像的分类精度。

Remote sensing image classification encounters the problems of fuzziness in data boundary and uncertainty in interpretation process of remote sensing information.In light of this,and in combination with the characteristics of fuzzy support vector machine（FSVM） which can effectively avoid the interference of the noise samples in classification applications,we propose a method for solving the membership of fuzzy support vector machine which is based on cloud model.The method inputs the quantitative position of samples by the reverse cloud algorithm without membership requirement to obtain the numerical characteristics of sample categories,and then gets the membership of each sample to its qualitative category according to the positive cloud algorithm.Experimental results show that it is feasible to use membership of the cloud model-based fuzzy support vector machine on the classification of remote sensing image,and this can effectively improve the classification accuracy of the images.