Green Change Detection In Beijing Urban Area Based On CBERS Remote Sensing Image

This paper mainly studied green land changes in Beijing main urban area,by Remote Sensing and other methods,aimed to support planning and construct of "Green Beijing".This paper mainly studied as the following several aspects:The CBERS CCD data of summer in Beijing for consecutive 6 years are selected to analyze change of green land in urban area of Beijing,and study the urban area in Beijing with coverage of 856km²(a scope equivalent to that within the 5^th Ring Road).By help of remote sense treatment software,we have conducted radiating correction,geographic correction,injection change,and a serial pre-treatments on the images in 6 different time phases,as per the specific situation of the studied districts,the images of 2001-2006 were classified into 4 categories by means of Neural Net:green land(including farm land where crops grow),water body,building(house,road),and bare land.From the study,we can draw the following conclusion:1) Compared with traditional classification method,manual neural net has the advantage of non-linear classification,more precise classification results can be acquired,and remote sense information distributed in complex space can be treated and analyzed.Compared with common linear regression method,it can better simulate complex non-linear system;it can well combine inaccurate or disturbed training data,so as to acquire higher precision level.2) The classification of Green is influenced by the no-integrity of CBERS CCD images.The detailed information drawing capability of the classified results of CBERS-02 CCD image is much more superior to CBERS-01 CCD.The selected domestic-made satellite data can completely meet the demand of city greening construction investigation.3) By 2006,the green land coverage has reached 38.2%.Considering difference in study area location,and the statistical method of green coverage,the target of "greening coverage of urban area before 2007 reaching 40%" has basically realized.4) The classification-before-comparison method can be used to detect the green land change,by compare NDVI of images.Thus the NDVI can be influened by the time of images,also the quality of images.By radiometric correction,we calculated NDVI after the apparent reflectance transform,the results of changes detection was still not very good.The study results show,the green land areas in 2001-2006 are respectively 25.5%,26.8%,30.0%, 32.6%,34.1%and 38.2%.During 2001-2003,the increase in green land area mainly concentrated near 4^th Ring Road and 5^th Ring Road;the change in urban area green land is not obvious,mainly there are increases in park and landscape green land area;in 2003-2004,the increase in green land area is mainly with the greening at two sides of urban road;while in 2004-2006,it was mainly with the in-piece greening in urban area.This study fully indicates,the construction achievement of "Green Beijing" is obvious,while the green land area in study area increases year by year.