Regional Population Spatialization Based On Integration Of Optical And SAR Data

Population density is an important indicator for social economy,urban development and ecological environment.So far,the commonest way to obtain population data is census data,which is authoritative,systematic and normative and can be used to reflect the population condition of a certain administrative unit.However,on account of its inability to represent inner variation,census data is unable to be used in loss assessment of accidental nature disaster,research on population change and social economy and development planning.To solve these,population spatialization can make up for the shortages of census data and provide spatial population data.The objective of this study is to development a method for regional population spatialization.Through the analysis and summarization of current methods,we divide those methods into two categories: methods based on information layers and methods based on image features.However,the currently existing methodologies have difficult to meet the demands of scale of regional population spatialization or to develop a direct relationship with population density.Therefore,this study proposed a method for population spatialization based on building density,which can both satisfy the scale requirement and has obvious relationship with population density.This study chose Jing-Jin-Ji urban agglomeration as study area,combine optical and SAR data and proposed a series of methods for extracting built-up,estimating building density and spatializing population.Firstly,this study came up with a method to extract rural and urban built-up based on improved variogram algorithm and applied it to the whole study area.Secondly,this study estimated the building density using CART algorithm and features from optical and SAR data.Finally,this study conducted the population spatialization based on building density and different scale unit and make up for the lack of three dimension information with the employment of social economic data.Firstly,this study analyzed the features of rural settlement and urban built-up in middle-high resolution SAR image and the reason for false classification of rural settlement caused by traditional variogram algorithm.An improved method which can highlight rural settlement,restrain farmland and decrease false classification was proposed based on the above analysis.The improved method was applied to the whole study area and eight sample areas indicated that the average detection rate was 86.81%,the average false error was 15.62% and the average missing error was 13.19%.Secondly,this study analyzed the shortages of using mono-source data to estimate building density and employed the combination of optical and SAR image,i.e.,spectral reflectance,the normalized indices and backscatter intensity,and CART algorithm to estimate the building density of the whole study area.The result indicated that the R2 between estimated building density and actual building density is 0.7831.Finally,population spatialization was conducted on three different scale level based the built-up extraction and building density estimation results and was evaluated by the census data of 176 counties of Jing-jin-ji agglomeration.The result indicated that the model which using the mayor level had the highest accuracy,i.e.,R2 of 0.6001,among models that merely based on building density.However,considering that building density only represents for two dimension information but population generally distributed over three dimension space,social economy data was added to assign weight to each county.The result showed that the model combining building density and social economy data had a higher R2 of 0.7515.