Building density estimation using PolSAR images based on adaptive volume scattering model

Abstract

Building density, which is ratio of building area to basal area, is of great significance in infrastructure planning and management for cities. Polarimetric synthetic aperture radar (PolSAR) images, delivering abundant information of detected areas, make the building density detection more convenient and accurate. The estimation of the density depends largely on the precision of building detection, which is a tough problem in PolSAR image interpretation because of the confusion of forest and buildings. Since the existing interpretation methods cannot distinguish buildings from forest accurately, an adaptive volume scattering model for the model-based decomposition is proposed in this study to help detect the building area. Together with the support vector machine algorithm, marker-controlled watershed algorithm and regression analysis, the ratio of building density can be calculated more precisely and comprehensively. Experiments on the ESAR L-band PolSAR data of the Oberpfaffenhofen have been taken out. The results demonstrate that the proposed method has a better performance in division of building areas and forest and can detect the building density with a higher degree of precision.