Detection of Residential Buildings to Estimate Population in Lebanon using GeoEye Images

Abstract

Scholars in urban planning and Geography are increasingly interested in grasping demographic information using Remote Sensing data. The accurate detection of residential buildings from satellite images seems to be essential in this domain. This paper has a dual purpose: It aims firstly at developing an automatized method for residential buildings extraction, then, evaluating the relationship between residential building characteristics (number, area, and volume) and demographic data. To do so, a dual phasic methodology is proposed. During the first phase, the extraction of residential buildings has been done using a transformation into HSI representation where the buildings corresponds to the higher values of band I. After that, the image has been transformed into vector and the forms of the buildings have been adjusted using convex hull tool in ArcGIS. The identification of residential buildings has been done using statistical data. The volumes of buildings has been calculated using MATLAB script. During the second phase, a multivariate regression has been established and a strong relationship (R2 =0.87) has been found between the volume of buildings and the population data.