Fast Space-Time Clutter
|Steven T. Smith and James Ward
MIT Lincoln Laboratory
244 Wood Street
Lexington, MA 02173-9108
tel: (781) 981-3106
Abstract Initial analysis of adaptive radar system performance is oftentimes performed using ideal covariance matrix and time-series analysis. If the radar must operate in the presence of clutter interference, then STAP algorithms are appropriate and the space-time clutter covariance matrix must be computed. Because this matrix's size is determined by the product of the aperture and CPI length, huge matrices result for even moderately sized radars. Nevertheless, space-time clutter covariance matrices containing hundreds of thousands of elements may be computed in a few seconds by exploiting Toeplitz structure and a trick involving the Fourier-Bessel expansion of the ground's reflected energy, i.e., a discretization in frequency. This method is typically several orders of magnitude faster than a direct approach of discretization in azimuth. The only assumption necessary is independent patch-to-patch clutter interference, i.e., no terrain scattered interference. Aside from this restriction, arbitrary illumination and reflectance patterns are treated. Furthermore, the following cases are easily incorporated into this framework at limited or no computational cost: clutter with zero pitch, clutter with nonzero pitch, internal clutter motion on the ground (Billingsly's model is used as one example), and nonzero signal bandwidth. The consideration of such details can add integrity to a performance study. As an additional benefit, a block Toeplitz data structure is easily obtained, resulting in potentially large savings in storage cost, and matrix-vector and matrix inversion operations. This talk presents these results, which are greatly expanded and enhanced from previously published work, and uses them to explore some implications for STAP performance of pitch, ICM, and bandwidth.
Direct comments and questions to: email@example.com
© MIT Lincoln Laboratory. All rights reserved.