Object based building extraction by QuickBird image for population estimation: A case study of the City of Waterloo

Abstract

This paper used QuickBird high resolution image to estimate the population of the city of Waterloo, ON, Canada. Two approaches of object based classification were compared to extract buildings from the original image. One is rule based classification and the other is example based classification. We chose two districts which are Lakeshore and Columbia as our testing areas. Rule based result is better than example based. The overall accuracy of rule based classification in Lakeshore District and Columbia District are 92.5% and 85.5%. With census data, the average area per person is about 38.8 m2 and the estimated population of the city of Waterloo is about 109589.