ABSTRACT
CONTENTS 4.1 Introduction ................................................................................................. 59 4.2 Study Area and Dataset............................................................................. 60 4.3 Mapping of Impervious Surface with Landsat ETMþ Data ............... 61 4.4 Mapping of Impervious Surface with IKONOS Data .......................... 66 4.5 Discussions................................................................................................... 69 4.6 Conclusions.................................................................................................. 71 References ............................................................................................................. 72
Research on impervious surface extraction from remotely sensed data has attracted interest since the 1970s. During the 1970s and 1980s, much research on impervious surface extraction was based on aerial photographs [1,2]. In the past decades, research was shifted to develop more advanced approaches for quantitative impervious surface extraction from satellite images. The approaches include per-pixel image classification [3-5], subpixel classification [6-9], and decision tree modeling [10-13]. Extractions of impervious surface have also been conducted by the combination of highalbedo and low-albedo fraction images [14-16] and by establishing the relationship between impervious surfaces and vegetation cover [17,18]. Because of the complexity of impervious surface materials and the spectral confusion between impervious surfaces and other land covers, extraction of impervious surface from remotely sensed data is still a challenge. The objective of this research is to extract impervious surface from Landsat ETMþ image through the integration of land surface temperature (LST) and fraction
images. Specifically, the combination of the LST and low-albedo fraction is used to extract the dark impervious surface contents existing in the lowalbedo fraction images. Another objective is to explore the approach to extract impervious surface areas from IKONOS data, through the use of a hybrid approach based on the combination of decision tree classifier (DTC) and unsupervised classification to extract the dark impervious surface and other shadowed impervious surface areas.
