高分影像在改进的最优分割尺度下的多层次混合分类

A Hybrid mutli-hierarchical classification based on an improved optimal segmentation parameter for high resolution remote sensing image

针对高分辨率遥感影像分类时的同谱异物和面向对象分类的最优分割尺度难以确定的问题,提出一种混合的分类方法。首先采用一种改进的目标函数来计算每种类别的最优分割尺度,实现对高分辨率影像的多尺度分割;其次,在每种类别的最优分割尺度下构建多层次分类;最后,在该多层次结构中,采取基于知识规则的分类和SVM分类器结合来进行分类。实验结果表明,本文方法总体精度和Kappa系数较其他方法都有明显的提高。

In the high resolution remote sensing image , different objects have the same spectrum and the optimal se^nentation parameter is difficult to be determined,a hybrid classification method is proposed. Firstly, using an improved objective function to calculate the optimal segnentatian parameter of each category, in order to realize the multi - resolution segmentation in high resolution image; Secondly, multi - hierarchy classification is built in each optimal segmentation scale; Finally, in the multi - hierarchy structure, the classification based on knowledge rules and the SVM classifier are combined for classification. Results show that the overall accuracy and Kappa coefficient are superior to other methods.