加权FCM原型空间特征提取的高光谱图像分类

在原型空间特征提取方法的基础上提出一种基于加权原型空间特征提取的方法用于高光谱图像数据分类。通过加权模糊C]均值算法对每个特征施加不同的权重，从而保证提取后的特征含有较高的信息量。实验结果表明，与PSFE相比，w-PSFE对数据集大小的稳定性更高，同时在提取少量的特征用于高光谱图像数据分类时分类精度更高。

Hyperspectral image classification based on weighted FCM prototype space feature extraction

A method called weighted Prototype Space Feature Extraction（w-PSFE） is proposed for feature extraction of hyperspectral data in this paper. The approach is an extension of previous approach-Prototype Space Feature Extraction（PSFE）. Each feature with different weights in terms of weighted Fuzzy C-Means（FCM） algorithm to ensure that the features contain more information after extracted. Experimental results show that compared with results obtained from approach PSFE, w-PSFE has a stability to data set and higher classification accuracy when extracts a small number of features used to hyperspectral image data classification.