Data volume reduction for imaging radar polarimeter

Abstract

0.Introduction.Imaging radar polarimetry is becoming a widely used tool for remote sensing studies. Several systems are now in use, each of which produces very large volumes of data which limit the usefulness of fully polarimetric observations of the Earth's surface. In 1985, an L-band SAR polarimeter was flown on a NASA aircraft recording data which were subsequently processed and stored at JPL. This system was improved in 1988 by the addition of P-and C-band systems, increasing the data rate and volume. One difficulty with these experiments is the large inherent data volume required for each image; this limits the number of data sets that can be stored efficiently. Furthermore, the computer time necessary to synthesize a picture in an arbitrary polarization is devoted primarily to cumbersome data transfers. Clearly, the volume of data must be reduced in order to provide the user with a more flexible investigation tool, yet the data volume reduction must not impact the system noise level by introducing additional errors.In this paper, a description of the operations necessary to generate a synthetic polarization image is first presented for clarity. Then, a method reducing the data volume as well as the synthesis time is described and the errors introduced by this operation are quantified and analyzed. This method is currently used at the Jet Propulsion Laboratory where the output of this operation is called the compressed data format.The terminology and definitions used in this paper are fully developed in [l], and will not be repeated here. Working with t h e complete set of d a t aAn imaging radar polarimeter operating at L band wae flown in 1985 on a NASA CV990 aircraft [2] and subsequently replaced in 1988 by a combination P-, L-and Cband system. The synthetic aperture radar in each system is monostatic and emits alternately horizontally and vertically polarized waves and receives both returns with two co-located antennas, one horizontally polarized and the other vertically polarized. The system is therefore able to measure the full scattering matrix for each resolution cell.A typical radar scene contains 4096 single-look pixels in the along-track direction and 1024 pixels in the across-track direction. Each pixel represents about a 4 X 12 meter ground area. A scattering matrix is stored for each of these pixels. Since a scattering matrix contains four complex numbers and since each complex number is 8 bytes long, the storage requirement for a typical scene is 128 Megabytes. If the reciprocity principle is assumed then the matrix should be symmetrical and only three elements of the scattering matrix need to be stored. In fact, this property is assumed when phase calibrating the data [2].In order to create a synthetic polarization image from this data set, the polarization vectors of both the emitting and the receiving antennas of the simulated system are chosen. The received signal, V is synthesized for each pixel:where h, and ht are the polarization vectors of the receiving and transmitting antennas and S ...

Hybrid-Polarity SAR Architecture

“…All information in the backscattered field is retained. In the quad-pol case, the Stokes parameters (often expressed in the 3x3 compressed format pioneered by JPL [11]) are interchangeable with those from a conventional linearlypolarized SAR. Thus, all analysis tools developed for conventional quad-pol SARs carry over to data products from a hybrid-polarity quad-pol SAR.…”

Polarimetric Portraits

“…Those elements are comprised of two pairs, each pair having two orthogonal received polarizations in response to one (of the two) orthogonal transmitted polarizations [vanZyl et al,1987]. In the event that the transmitted and received polarization bases are the same [Emmons and Alexander, 1983]-which in conventional remote sensing radars are the linear polarizations (H, V)-those data may be compressed by exploiting reciprocity between orthogonally cross-polarized counterparts [Zebker et al, 1987;Dubois and Norikane, 1987]. The resulting quadrature polarized (quad-pol) data products are set up for analysis by users in either a coherence or covariance matrix, each derived from the scattering matrix [Cloude and Pottier, 1996], or the Stokes (Kennaugh) 4x4 matrix operator [vanZyl et al, 1987].…”

Hybrid Dual-Polarization Synthetic Aperture Radar

“…If those data are retained for subsequent operations, such a SAR is termed fully polarimetric. However, it is customary to reduce that array to a 3 × 3 matrix, by taking advantage of symmetries that are true in almost all known situations [63]. This form is known as quadrature polarimetry (QP).…”