The Research Of Land Cover Information Extraction Based On GF-1 Satellite Image

In recent years,With the development of remote sensing technology, how to faster and accurater extract information from remote sensing date has become one of the focus with the experts.However,as the first launch satellite of "GF Special Project",the use of GF-1 satellite image to extraction information has a very important significance for the satellite that launched later.This article take GF-1 satellite image as the experimental data,hangzhou city as the experimental area.Research the technology of the information extraction from the high-resolution image,have bring forward an work flow of object-oriented method and construct an rule of suburbs classification from high-resolution remote sensing image,also compare the difference of the new method and the co nventional methods.Main cotent and production are as follows:(1) Research the object-oriented method and the key technology of it,According to the feature of GF-1 satellite image,and put forward the process that registration at first,then merge,finally correction.and have proposed a work flow that based on the sift registration and Envi Feature Extraction workflow.(2) Research the based-on edge segmentation.Analyzed the effect of the scale and the segmentation parameter on segmentation result,and select a best segment value and parameter that suit for experimental area and other similar regions.(3) Construct the classification rule of suburbs area.By analyzing the segmentation image,select color,shape,texture,density,standard deviation and the relation o f objects as the classification characteristic,counstruct the best classification rule,to produce a high accuracy.(4) Compare the difference of the object-oriented method and the conventional method.The object-oriented have the advantage as follow:it make full use of the spectral information and the space information of the high-resolution remote sensing image,and increase the classification accuracy;Second,the result of the new method is object- level,have a good vision effect.