面向对象的高分辨率遥感影像道路自动提取实验

传统的遥感影像道路提取,主要是利用数理统计与人工解译相结合的方法。这种方法不仅精度相对较低,效率差,而且依赖参与解译的人,在很大程度上不具备重复性。本文采用面向对象的分类技术,充分利用高分辨率遥感影像中丰富的空间结构信息和地理特征信息,针对实验区中的不同道路类型,在不同尺度下自动提取出道路信息。同时,本实验还提供了一种提取道路的普遍性规则集,提高了道路提取的自动化水平。通过实验表明:该方法提取速度快、精度高。

An Experiment of Automatic Road Extraction from High-resolution Remote Sensing Image Based on Object-oriented Technology

The traditional road extraction of remote sensing mainly makes use of the method combining statistics with man power interpretation,which is low in precision and inefficient.Meanwhile,the method depends on the people who participates in the interpretation doesn’t have repetition to some extent.According to different styles of the road,this paper adopts a classification technology of object-oriented to extract the road information from high-resolution remote sensing image whose information is abundant in the space structure and geographical features on different scales.Meanwhile,this paper also provides a universal ruleset of road extraction,which improves the automation degree of road extraction.Also,the experiment indicates that this method has a precise result and short course.