一种新的高分辨率遥感影像模糊监督分类方法

针对高分辨率遥感影像分类中由于细节特征突出、同质区域光谱测度变异性增大所带来的像素类属的不确定性及模型的不确定性等造成的误分结果，提出一种基于模糊隶属函数的监督分类方法。对同质区域定义高斯隶属函数模型用来表征像素类属不确定性；模糊化该隶属函数参数建立影像模糊隶属函数，以建模同质区域光谱测度的不确定性；用训练样本在所有类别中的模糊隶属函数及原隶属函数（高斯隶属函数）中的隶属度为输入，建立模糊线性神经网络模型作为目标函数，实现分类决策。该算法和经典算法对World View-2全色合成影像及真实影像进行定性和定量分类实验，分类结果验证了文中方法具有更高的分类精度。

A New Method of Fuzzy Supervised Classification of High Resolution Remote Sensing Image

This paper presents a supervised image classification method based on fuzzy membership function to solve incorrect classification of high resolution remote sensing image, which caused by highlight detail information, the uncertainly of the pixels classification derived from the increase of the differences between pixels in the homogenous region, the uncertainly of classification decision and so on. First, Gaussian model is used to characterize the uncertainly of the membership of pixels; then the model is extended to build the image fuzzy membership function to define the uncertainly of the homogenous regions. To segment the image, the objective function is built by linear function of neural network, which the fuzzy membership functions and the membership degrees of the original fuzzy membership functions as input values. The proposed method is compared with the classification methods tested on the WorldView-2 panchromatic synthetic and real images. Through the qualitative and quantitative experiments, it can be found that the proposed method has better classification accuracy.