边缘保持滤波的遥感影像多特征联合分类

针对高分辨率影像在分类时存在的“海量数据灾难”“椒盐”现象、地物边缘不可分性强的问题,提出边缘保持滤波的高分辨率遥感影像多特征联合分类方法。该方法主要分为3部分。首先,提取影像的多种特征进行联合处理,减少数据处理运算量;然后,利用极限学习机对样本子类训练多个弱分类器,通过分类器获得初始类别概率影像;最后,利用多特征联合影像构建边缘保持滤波器,对初始类别概率影像进行滤波处理,通过投票表决的方法确定每个像素的类别。通过实验验证了该方法的有效性。

Multi-feature Joint Classification of Remote Sensing Imagery Based on Edge Preserving Filter

In order to solve the problems of“massive data disaster”,“salt and pepper”phenomenon and the indiscernibility enhancement of the edge of ground objects in the classification of high-resolution remote sensing images,a multi-feature joint classification method based on edge preserving filtering is proposed.This method is mainly divided into three parts.Firstly,extract multiple features of the image for joint processing to reduce the amount of data processing operations.Then,use the over limit learning machine to train multiple weak classifiers for sample subcategories,and obtain the initial category probability image through the classifier.Finally,use the multi feature joint image to construct the edge preserving filter to filter the initial category probability image,and then determine the category of each pixel by voting.Experimental results prove the effectiveness of this proposed method.