EOC用SAR処理設備について
A New SAR processor of Earth Observation Center

島田政信 (宇宙開発事業団・地球観測データ解析研究センター)
Masanobu SHIMADA (Earth Observation Research Center, NASDA)
E-mail: shimada@eorc.nasda.go.jp
 

要旨  2000年問題の解決と、一般利用者の高度な処理要求を満たすべく、宇宙開発事業団は高精度高速処理可能なSAR処理設備の開発を進めてきた。その鍵を握るのは高精度処理アルゴリズムであり、本開発においては、これまでEORCがJERS-1検証を目的として開発してきたSAR処理ソフトウェアをコアとして使用することとなった。これにより、利用者には従来よりも良好な精度の成果物が届くものと思われる。本発表では、本ソフトウェアの性能、精度を中心に紹介する。

ABSTRACT  This paper presents a verification processor which can be used for imaging, calibration, and interferometry of data from a spaceborne synthetic aperture radar (SAR).  This system can process the SAR on board ERS-1 and JERS-1 with various types of the computers ranging from microcomputers to the supercomputers.  This paper summarizes the functionality, configuration, and system performance of the processor.
 

INTRODUCTION
  Synthetic Aperture Radar (SAR) is an active microwave instrument that performs high-resolution observation under almost all weather conditions.  The measurements with SAR are the backscattered signals from the targets on the Earth in response to its transmission code. The optimal correlation of these signals with the ideal SAR receiving signal achieves high-resolution imaging of several meters.  These correlated signals (or simply SAR images) contain information on the targets, the wave-propagation media, the distance between the SAR and the target, and the SAR characteristics.  When the image is well modeled for these targets, the SAR observation can retrieve the target itself.  There are, however, two difficulties.  The first is how well the SAR characteristics are eliminated from the SAR images and how
accurately the signal power or the backscattering coefficient (s0) of the target is determined.  The second is how accurately the backscattering model can be built for each target.  The first difficulty is called SAR calibration.  Because erroneous calibration of SAR images causes misunderstanding of the targets, the SAR calibration  is very important.  To estimate s0 and phase from the SAR data accurately, a SAR processor, that can conduct the imaging and interferometry-process in preserving the phase, is required.  Recent innovations in microcomputer technologies enable effective development of SAR processing algorithms.  Since 1992, we have been developing SAR processors.  We have now completed a processor which can run on several types of the computers and process ERS-1 and JERS-1 SAR data.  In this paper, we summarize the processor performance.
 

SYSTEM CONFIGURATION
  The verification processor is composed of three subprocessors: SAR processor, SAR calibration processor, and SAR interferometry processor. The detailed descriptions of the SAR processor are given below.
 

SAR Processor
   Basic Processing: The SAR processor adopts the range-Doppler type algorithm. To avoid errors due to number rounding of the Fourier transformation in azimuth, double precision is always adopted. Fourier transformation can be selected from length of 2048, 4098, and 8192.
   Doppler Centroid Estimation: The Doppler centroid is estimated from i) the orbit state vector, ii) the peak response of the azimuth spectrum of raw data, or iii) the peak response of the azimuth spectrum of correlated data. The Doppler chirp rate is calculated from the state vector (Curlander and McDonough, 1991).  The state vectors for recent satellites are very accurate, so, autofocusing is not installed.
   Radiometric Corrections: Sensitivity time control (STC), automatic gain control (AGC), saturation, and interference from the ground radar can be corrected in the range and azimuth correlations (M. Shimada, et. al., 1991).  This supports several window functions (i.e., rectangle, Hanning, Hamming, three sigma gauss, and so on).
   Output Expressions: The outputs are threefold, i) the 4-byte single-look-complex data, ii) two byte four-look-unsigned-short integer data, and iii) four look single-byte-unsigned-character data.  The output is radiometrically corrected for antenna pattern and slant range.  The data can be converted to s0 using the calibration factor, which is calculated at the calibration processor.