Radiometric Calibration of Sar Image Data

Recent advances in calibration techniques have made calibrated SAR images, in which the data represents radar cross-section (0') or normalized radar cross-section (0'0), a reality. In this paper, some of these recent advances are discussed in the context of their impact on SAR sys'tem design, and in terms of the errors associated with each technique. The main sources of error in radiometric calibration of SAR data are fluctuations in the transmitted power, receiver gains, system noise and the illumination pattern of the antenna. Internal calibration loops can be used to ensure the stability of the transmitters and receivers. The noise level can be measured by turning off the transmitters and operating in a receive-only mode. One of the more difficult problems remaining is then to determine the appropriate antenna pattern illumination for each pixel on the ground. This can be achieved by a variety of techniques: first, the normal antenna pattern can be measured using a suitable uniform, distributed target such as the Amazon Rain Forest; second, the roll angle of the antenna can be measured by using some form of echotracker; then the elevation angle for each pixel can be determined if the terrain height variations in the image are known or can be bounded. For aspaceborne SAR, a curved earth model gives a good approximation for the long range at which spaceborne SAR's must operate. For an airborne SAR, a Digital Elevation Model (DEM), registered to the SAR image, may be necessary. The DEM can be derived from a standard topographic data set, registered using tie-poin ts to the SAR image, or from interferometric SAR height maps which are automatically registered to the SAR image, or by monopulse techniques, which are also in the same reference frame as the SAR image. Additionally, these three techniques may be used to determine the local incidence angle for each pixel, so that measurements of 0' may be converted to 0'0. To verify radiometric calibration, and particularly to ensure that the data is absolutely calibrated, known calibration targets such as corner reflectors or transponders, are often used. These suffer from their own calibration uncertainties, which may be a limiting factor in the overall radiometric calibration. Examples will be presented from the NASA/JPL AIRSAR and ERS-l.