Effect of aspect angle normalized PolSAR data on urban building detection

Abstract

Variation in the building aspect angle, which is defined as the angle between the flight direction of a satellite (azimuth direction) and the vertical wall of a building, is a significant cause of accuracy reduction in building detection and terrain classification from polarimetric synthetic aperture radar (PolSAR) data due to the structure-sensitive characteristic of PolSAR data. This variation also makes comparison or change detection within time-series PolSAR data on the same buildings questionable and unsatisfactory. In this paper, an aspect angle normalization method based on the relationship between the aspect angle of the building and the orientation angle shift is introduced to remove the disadvantageous influences caused by the aspect angle. From a comparison between the double-bounce building detection results of the original and aspect angle normalized RADARSAT-2 spaceborne and E-SAR airborne PolSAR data, buildings with any aspect angle can be detected effectively from aspect angle normalized data. In addition, the aspect angle normalization method can be used to improve the capability of building change detection.