Abstract :
The optimum design of synthetic aperture radar (SAR) systems intended to classify randomly reflecting areas, such as agricultural fields, characterized by a reflectivity density spectral density is studied. Assuming areas of known shape and location, the binary case, and a certain Gaussian signal field property, and ignoring interfield interference, the problem solution is given. The optimum processor includes conventional matched filter processing, but is nonlinear; a coherent optical system realization is outlined. The performance is approximated using a x2 assumption and bounded by the Cernov bound. A fundamental design problem involves the system bandwidth analogously, in a special case, as in diversity communication systems; a solution is given based on the Cernov bound. A set of summary design curves is given and exemplified by a satellite SAR system design. Also discussed is the measurement of reflectivity spectral density amplitude with imaging sidelooking (synthetic or ¿brute-force¿) radars and the maximum likelihood estimator´s accuracy and realization with a coherent optical system. It is also shown that a CW modulation is useable if the random reflectivity is, effectively, isotropic. Finally, the reflectivity density spectral density amplitude, when constant over the spatial bandpass of the measuring system, is related to the scattering cross-section density commonly measured.
