基于面向对象的高分影像地表覆盖典型要素快速提取对比研究

摘要

全国第一次地理国情普查工作正在开展，涉及大量影像解译工作，高分辨率遥感影像是当前解译工作中应用最广泛、最基础的影像产品。然而，由于高分辨率遥感影像地物颜色多样、纹理结构复杂，传统的基于像元的解译方法对高分辨率遥感影像进行地物要素自动提取的效果不能满足日益工程化项目要素提取效率与分类精度要求。因此，针对基于像元的影像解译方法存在的不足，本文采用面向对象的解译方法，对几种常见影像处理软件进行对比试验，并对精度进行验证。结果表明：基于 eCognition（易康）软件的遥感影像自动分割分类解译精度较 ENVI EX、Titan Image 软件解译精度高。%The first nationwide general survey of geographic conditions which involves massive image interpretation is being carried out.High-resolution remote sensing imagery is the widely used and the most basic imagery product in image interpretation.However,because of its varied colors and complex texture structure,the conventional pixel-based image processing technique is not applicable.This paper compared and analyzed commonly used remote sensing image processing software.Experimental results show that accuracy of the automatic image segmentation and classification based on eCognition software is higher than that of ENVI EX and Titan Image.