基于属性形态学剖面多特征超高分辨率遥感影像分类研究

Research on Attribute Morphological Profiles Based on Multifeature Ultra-high Resolution Remote Sensing Image Classification

针对近年来出现的超高分辨率遥感卫星数据,本文提出了一种基于形态学属性剖面多特征分类方法。首先针对超高分辨率多光谱影像提取属性形态学剖面,提取相应的细节信息;然后结合多光谱影像的光谱信息,训练分类器。其次,对Worldview2城镇区域影像进行了分类,可以看出,应用形态学属性剖面多特征分类的算法可以有效地将地物进行区分,目视结果和定量结果都达到了较高精度。

A multi-feature classification method based on morphological attribute profile is proposed for ultra-high resolution remote sensing satellite data.Firstly, we extract attribute morphological profiles from ultra-high resolution multi-spectral images which can extract corresponding details.Then the random forest classifier is trained by combining the spectral information of multispectral images and the attribute morphological feature.We classify the Worldview2 image of urban area.It can be seen that the multi-feature classification algorithm based on morphological attribute profile can effectively distinguish the objects, and the visual and quantitative results have achieved high accuracy.