Geophys. J. Int. zyxwvutsrqpon (1992) zyxwvutsrq 108, 423-432 zyxwvutsrqp Inversion of synthetic aperture radar data for surface scattering J. S. Won and W. M. Moon Geophysics, The University of Manitoba, Winnipeg, Manitoba, Canada R3T 2N2 Accepted 1991 July 23. Received 1991 July 15; in original form 1991 March 27 SUMMARY The conventional image-formation technique of high-resolution SAR synthetic aperture radar (SAR) data has utilized correlation in the range-Doppler domain. An alternative, more recent approach, in the SAR image-formation algorithm exploits downward extrapolation of the wavefield in the zyxw f-k domain to perform not only azimuth compression but also simultaneous range-curvature corrections, with improved quality of the final image. In this paper, a new approach to SAR data processing, based on the inverse scattering and the Kirchhoff approximation, is described and tested. The complex backscattering coefficient can also be estimated by this new approach provided the surface scattering is dominant. The final inversion formula is designed to exploit f-k domain computation in an analogous manner to seismic Born inversion. Digital simulations using one- and multiple-point target models are presented to dem- onstrate the performance of the proposed method. Key words: Kirchhoff approximation, surface scattering, synthetic aperture radar. 1 INTRODUCTION Since early 195Os, synthetic aperture radar (SAR) has been developed as an imaging radar into one of most valuable remote sensing tools. Fig. 1 shows schematic SAR geometry. The dimension corresponding to the aircraft flight or spacecraft orbit direction is called along-track or azimuth (x-axis in Fig. l), and the horizontally normal to azimuth is called ground-range (y-axis in Fig. 1). The slant-range is defined by the radial distance from the aircraft to the surface scatterers. Fig. 2 shows the relationship between slant-range and ground-range. The slant-range determines the two-way traveltime, which is directly related to the system's measurement, of the transmitted pulse. The SAR signal data is analysed into a linear frequency-modulated chirp signal in both range and azimuth dimension (Kovaly 1972). Thus, conventional approaches to processing synthetic aperture radar signal data are mostly based on matched filter theory and are well described in various publications such as Wu (1976), and Bennet zyxwvut & Cumming (1979). The azimuth compression * of these approaches utilizes a 1-D or limited 2-D matched filter in *The term azimurh compression conventionally represents the correlation of along-track data with a azimuth reference function because a conventional SAR processor utilizes a matched filter. It is a correlation in offset dimension according to seismic processing. However, it can be any kind of processing that can reconstruct along-track impulse-response from raw SAR signal. the range-Doppler domain, and consequently the processing efficiency and performance depends very much on how the range-migration correction is implemented in the range- Doppler domain. Two different approaches to SAR data processing, in which the azimuth compression is performed in the f-k (frequency-wavenumber) domain, have been recently developed by Rocca, Cafforio & Prati (1989), and Raney & Vachon (1989). These two algorithms are quite different in their basic approach. The former was initially conceived and developed from the similarities between the reflection seismic and SAR configuration, while the latter is developed solely from the theory of the SAR system. zyx An attractive feature of the two algorithms is that the conventional range-migration correction can be more efficiently and precisely achieved by phase shifting in the f-k domain. Furthermore, the f -k domain processing simultaneously accommodates the range-curvature which is a function varying with the range. Success in applying the seismic migration idea to the SAR data processing has led us in this paper to develop a new SAR data processing method based on the inverse scattering theory (Bleistein 1984). The Born inversion method in optics has been successfully applied to the inverse scattering of seismic wavefields (Cohen & Bleistein 1979). In this paper, an inversion of the SAR image data is formulated in a similar fashion to the seismic Born inversion method. To estimate physical parameters characterizing surface pro- perties, such as the dielectric constant from SAR data, it is 423 Downloaded from https://academic.oup.com/gji/article-abstract/108/2/423/561730 by guest on 30 July 2018