利用面向对象的分类方法从IKONOS全色影像中提取河流和道路

传统的基于像素的统计特征的分类方法在处理高分辨率影像的分类问题上遇到了很大的困难。本研究利用面向对象的影像分析方法对IKONOS全色影像进行了河流与道路的分类,包括利用影像对象的光谱特征的初次分类和利用子目标对象的线特征的二次分类两个过程;其中初次分类由于仅依据光谱信息,不能很好地将河流或道路与其他光谱特征相似的地物区分开,而通过引入子目标的形状特征进行二次分类,就可以准确地将河流与道路提取出来。试验结果表明,面向对象的分类方法能够满足高分辨率或纹理影像的分类需要,具有很大的应用潜力。

Extracting the river and the road using an object oriented technique from IKONOS panchromatic imagery

The result will be unacceptable when traditional classification approach, which is based on the grey value of pixels, classify the high resolution image data.In this study,we extractd the river and road from IKONOS imagery using object oriented image anlysis technique, in two phases: the first classification using spectral information of image objects and the second classification using line features based on subobjects. In the first phase,the river and road are not classified exactly, because only spectral information is considered, but by introducing the shape features of subobjects, they are classified accurately in the second classification. The results prove that the object oriented approach can satisfy the classifications of high resolution or texture imagery, which is a potential image anlysis technique.