针对塌陷地的光谱特征和空间结构特征的特点,采用一种面向对象的技术,实现了塌陷地的信息提取。首先,采用形态学方法建立滤波器,对影像进行形态学重建,然后对影像进行分水岭分割,将影像进行多尺度分割后,影像被分割为多个相互独立的闭合区域,并且每个区域作为一个独立的对象,计算每个对象的光谱、形状、纹理等等各种特征,并把它们作为分类的依据;其次,根据提取专题的不同进行特征选择;接着根据所选择的特征选择分类器进行塌陷地提取。
To extract subsided land is a matter of ecological restoration. Making full use of spectral information and spatial texture feature, subsided land extraction is realized based on object - oriented method . Firstly, filter is reconstructed by morphology, and image is reconstructed by morphology, after multi - scale segmentation by watershed segmentation, image is segmented into closure and Independent regions, and the spectral feature, form feature and texture feature are calculated for the basis of classification; then features are selected on the basis of thematic extraction. Lastly, subsided land is extracted by classifier in the light of selected features, object - oriented method has better accuracy than othermethods.