PHASE CALIBRATION OF IMAGING RADAR POLARIMETER STOKES MATRICES

Phase calibration is a required step in the utilization of imaging radar polarimeter data, as errors in the phase relationship between polarization channels may lead to false conclusions about the nature of scatterers on the surface. The Stokes matrices measured by such an imaging radar polarimeter provide in themselves adequate information for the accurate phase calibration of the observed polarimetric characteristics of a surface, where by phase calibration we refer to correcting for instrumentally induced errors in the relative phase relationships between the elements of the measured scattering matrices. This property allows the data to be reduced in volume in an operational synthetic aperture radar (SAR) correlator with no prior knowledge of the conditions at the surface. No ground calibration equipment is necessary, as all important parameters are derived from the data themselves. We have implemented a phase calibration procedure which may be used to calibrate imaging radar polarimeter data and have applied it to data collected by the NASA DC-8 radar system. We illustrate the technique here by using an image of the Cima volcanic field in California, which contains in addition to the cinder cones and lava flows several trihedral corner reflectors, plus a mining area containing metal buildings we may use for validation of the technique. The trihedral reflectors may be expected to exhibit odd-bounce scattering geometry, while the buildings exhibit double-bounce geometry. We show that before calibration the polarization signatures do not at all resemble theoretical predictions, while after phase calibration the theoretical and observed signatures agree quite well. We note that the calibration factors are derived independently of these specific targets, and the targets are examined solely to illustrate the success of the technique. We have combined the phase calibration with a data-compression technique. The particular data-compression algorithm currently used at the Jet Propulsion Laboratory (JPL) reduces data volume of the output products by a factor of 12.8, resulting in a product that is of relatively manageable size (10 versus 128 megabytes for each data set). Any amount of averaging to reduce data volume at the expense of resolution may be achieved, without loss of phase calibration fidelity. The algorithm is invertible; thus if the data set is incorrectly calibrated once, the data may still be phase calibrated correctly without the use of external equipment. © 1990 IEEE