ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001 
CONSTRUCTION OF 3D MODELS FOR ELEVATED OBJECTS IN URBAN AREAS USING AIRBORNE SAR 
POLARIMETRIC DATA 
Yalkun YUSUF 1) , Masashl MATSUOKA 1) , Fumio YAMAZAKI 1) 
Seiho URATSUKA 2 ’, Tatsuharu KOBAYASHI 2) , Makoto SATAKE 2) 
Earthquake Disaster Mitigation Research Center 
National Research Institute for Earth Science and Disaster Prevention (NIED) 1) 
2465-1 Mikiyama, Miki, Hyogo 673-0433, Japan. 
Tel: +81-794-83-6632, Fax: +81-794-83-6695 
Email: yalkun@edm bosai.ao.ip 1) 
Communications Research Laboratory (CRL) 2) 
Ministry of Public Management, Home Affairs, Posts and Telecommunications 
4-2-1 Nukui-Kitamachi, Koganei, Tokyo 184-8795, Japan 
Tel: +81-423-27-7465, Fax: +81-423-27-6665 
Keywords: airborne SAR, Synthetic Aperture Radar, 3D construction, polarization, segmentation, classification, GIS, speckle noise 
filtering 
ABSTRACT 
Airborne survey methods can be an effective solution since they can provide spatial information with a large-scale area quickly and easily. 
Using remote sensing data to construct 3D urban area is very useful for urban planning. This paper presents the elevation extraction in an 
urban area using airborne SAR (Synthetic Aperture Radar) polarimetric data. Simple trigonometric models and the SAR geometry 
parameters were used to derive relative heights of elevated objects from the shadow of the elevated objects (Kirscht and Carsten, 1998). 
A Lee polarimetric filter was applied to reduce speckle level while preserving the polarimetrtic properties and is given to be useful to 
improve the classification accuracy (Lee et al., 1999). In classification of SAR image, Bayes maximum likelihood classification algorithm is 
applied to SAR filtered data, which is based on the complex Wishart distribution of the multilook covariance matrix (Lee et al., 1994). 
Finally a contour map and VRML image were given using the result of the constructed 3D elevated object of the study areas. 
1.INTRODUCTION 
Imaging radars are rapidly becoming one of the major tools for 
observing the earth’s surface and its cover. SAR is an active radar 
system due to its capability to image in all weather, day and night, 
and independence of sun illumination. It can provide spatial 
information with a large-scale area quickly and easily. Airborne 
SAR provides high-resolution images, which can be used for 
identifying individual buildings. Therefore the applications of a 
SAR embrace a wide range of fields including geology, natural 
disasters, agriculture, detection of snow and ice, and 
oceanography. 
Using remote sensing data to construct 3D urban area is very 
useful for landscape planning. As an attempt, in this paper we 
presented construction of 3D elevated objects in urban area using 
airborne SAR polarimetric imagery data. 
Radar images area composed of many dots, or picture elements. 
Each pixel (picture elements) in the radar image represents the 
radar backscatter for the area on the ground, dark areas in the 
image represent low backscatter, bright areas represent high 
backscatter. Backscatter for a target area at a particular 
wavelength will vary for a variety of conditions: size of the scatters 
in the target area, moisture content of the target area, polarization 
of the pulse, and observation angles. Flat surfaces that reflect 
little or no microwave energy back towards the radar will always 
appear dark in radar images. Surfaces inclined towards the radar 
will have a stronger backscatter than which slope away from the 
radar and will appear brighter in a radar image. Some areas are 
not illuminated by the radar, like the back of building, are in 
shadows, and will appear dark. The shadows of the objects are 
very helpful to judge the size and shape of objects by providing an 
impression of convexity and concavity. Since the shadows are 
assigned to suitable elevated objects, then relative heights of 
elevated objects from SAR data image can be derived from the 
shadow using trigonometric models and parameters of the SAR 
geometry (Kirscht and Carsten, 1998). 
In most case, images produced by radar have noisy in 
appearance than comparable pictures produced optically due to a 
phenomenon called speckle. The radar echoes reflected from 
individual scatters within a pixel either randomly reinforce one 
another if they happen to be in-phase or reduce the return signal 
if they are out-of-phase (fading). Speckle complicates the image 
interpretation problem by reducing the accuracy of the image 
segmentation and classification. In this study, to reduce the 
speckle noise of SAR polarimetric imagery and improve the 
classification accuracy, a Lee polarimetric filter was applied to 
reduce speckle level, which has the feature of smoothing the 
image data without removing edges or sharp features in the 
images (Lee et al., 1999). 
To construct 3D image from SAR polarimetric data needs 
classification and segmentation of SAR data. A classification 
algorithm was applied to the filtered SAR data based on the 
complex Wishart distribution of the multilook covariance matrix 
(Lee et al., 1994). 
After filtering speckle noise, segmentation and classification of 
SAR images, elevated objects and shadow regions are 
segmented in the SAR image. Since the shadows are assigned to 
suitable elevated objects, then relative heights of elevated 
objects from SAR data image can be derived from the shadow 
using trigonometric models and parameters of the SAR geometry. 
339