Sparse Vegetation Extraction And Change Analysis Based On Spectral Differentiation

As an important part of the desert ecosystem,vegetation is the “indicator” of the changes in the ecological environment in arid and semi-arid regions.The extraction of desert vegetation in Shule River Basin can not only determine the scope of spatial distribution of vegetation,but also reveal the process of ecological environment change by analyzing its changes.On the basis of reviewing and summarizing the extraction method of desert vegetation,the paper uses mediumresolution multispectral imagery as the data source and uses the method of spectral differentiation to extract the vegetation information in the study area.Finally,the vegetation change in the plain of the Shule River Basin in the past 31 years was analyzed.Firstly,the differential results of the measured desert vegetation and bare soil reflectivity were analyzed.It was found that the spectral differentiation can better highlight the vegetation information.Among them,in the first derivative,the "red edge" feature of the vegetation is obvious,and in the second derivative and the third derivative,the main band that can distinguish the desert vegetation from bare soil is the visible light band.Comparing the results of the non-synchronous long differential results,it is found that as the differential step length increases,the difference in the first-order differential curve is small,but as the differential order increases,the differential step required to distinguish between desert vegetation and bare soil can be obtained.The length also increased,with the best differential step size of 5 nm for second-order differentials and 7 nm for third-order differentials.The analysis of the differential characteristics of the measured spectrum provides the basis for selecting the optimal step size and band for extracting vegetation information from hyperspectral images.Multispectral imagery has fewer information bands and less information redundancy.In differentiation,it only needs to process each band as discrete data without the steps of band selection.The differential results of multispectral remote sensing images have shown that higher-order differential results of images do not necessarily increase the amount of information.Therefore,it is necessary to build high-precision remote sensing indicators to improve the accuracy of information extraction.In particular,the characteristics of the scatterplot and the second-order differentials of the third-order differential bands are higher,but the increase in the amount of information of the second-order differential bands is also less than that of the first-order differentials.This result is consistent with the measured features.The result of spectral differentiation is consistent.Vegetation differential index(VDI)and Gobi differential index(GDI)were used to extract the desert vegetation,showing a better advantage.However,due to the use of two different sensor images,vegetation extraction methods are also different.In the Landsat5 TM image,more accurate vegetation distribution information is obtained by GDI greater than zero.In the Landsat8 OLI image,by combining GDI and VDI,VDI is less than 0 to obtain the land surface where all the vegetation and the Gobi are mixed,and then the Gobi surface is removed by GDI greater than 0 to obtain a high precision sparse vegetation range..At the same time,a regression analysis was conducted on the VDI and GDI indicators and the vegetation coverage respectively.It was found that the linear relationship between GDI and vegetation cover(FVC)obeys a strict linear relationship,whereas the relationship between VDI and vegetation coverage obeys an inverse proportional function.However,because the correlation of VDI is lower than that of GDI,the complexity of the model is also higher than the former.Because of the versatility of GDI in the two sensor images,GDI is used as a method that can both extract sparse vegetation and can also Remote sensing indicators for inversion of vegetation cover changes and other vegetation-related biophysical parameters.Analysis of the vegetation changes in the plain area of the Shule River Basin found that the total range of vegetation distribution has changed little in the past 31 years,and the vegetation distribution has remained basically stable.However,in the study period,desert sparse vegetation and high-coverage vegetation showed Different trends.Specifically,during the period from 1986 to 1995,the area of sparse vegetation in the plains of the corridor increased while the area of highcoverage vegetation decreased.From 1995 to 2017,the distribution of sparse vegetation decreased,and at the same time,the vegetation coverage of high coverage For the gradually increasing trend,desert sparse vegetation and high-coverage vegetation show completely opposite trends.The vegetation cover change during the study period was obtained during the adjacent years.It was found that the overall vegetation cover change from 1986 to 1995,ie,the vegetation coverage showed a decreasing trend,and the vegetation coverage after 1995 became better.