高分辨率遥感影像中提取特定地物的分区多阶段混合分类方法

遥感图像分类技术发展过程中出现了各种不同的分类方法,本文提出一种对分类过程控制和分类中间数据管理的综合分类方式,通过分类方法应用方式的改进用于提取局部的特定地物。应用多阶段循环分类过程,在一个阶段的分类中控制参与分类的像元范围,在一次分类中混合应用监督分类、非监督分类和区域生长分类法,以及各种分类后处理技术,目的在于综合应用目视解译、计算机自动分类等各种技术的优点,结合地物的光谱数据,用于局部区域的特定地物信息提取和分类。

A Study on Segmentation-based, Multi-stage and Mixed Classification for Land Type Identification on High Resolution Remotely Sensed Imagery

There are many classification methods available with the development of remotely sensed imagery application. In the present study, we put forward and realized a new approach to the improvement of the traditional classification method through classification process control and classification data management. The way is improved on the application of the classification methods on the local imagery interpretation. The iterative method focuses on segmentation-based classification data, multi-stage classification process, mixed classification methods including supervised, un-supervised and area increase and some post-classification processes. It aims to combine the advance artificial interpretation with computer aided automation so as to extract ground objects in the regional mapping.