LAND-COVER CLASSIFICATION OF TEHRAN USING L- AND C-BAND SYNTHETIC APERTURE RADAR IMAGERY

摘要

Monitoring land-cover of urban areas is a main issue in several fields such as urban planning and seismic risk assessment. Detecting built-up and bare land areas in arid or semi-arid regions is quiet difficult by using multi-spectral optical images because of similarity of the spectral characteristics of grounds and building materials. On the contrary, synthetic aperture radar (SAR) images have possibility to overcome this issue because the backscatter depends on the material and geometry of different surface objects. The use of L- and C-band SAR images may have possibility to provide more information of the same objects in urban areas. In this paper, dual polarized data from ALOS-2 PALSAR-2 (HH, HV) with 6.2-m resolution and Sentinel-1 C-SAR (VV, VH) with 13.9-m resolution were used for an unsupervised classification analysis of land-cover in Tehran city, Iran, which has been growing very fast recently. Although the result of classification from the SAR images was better than that from optical images, some noise still remained in the result. Hence texture information was added to improve the classification. The result, which showed less noise by combining the texture measures with the backscattering intensity, was then compared with the visual inspection result of a high-resolution optical image and a reasonable level of accuracy was confirmed.