Rapid extraction of doubly-curved scattering centers for automobile detection in the presence of clutter

This paper presents an algorithm for low-order automobile detection based on doubly-curved surface scattering phenomenon. As reported in recent works, detection based on doubly-curved scattering features is possible because automobiles consist of scattering centers that are characteristic of canonical shapes. Using a basis consisting of frequency dependent functions from well-known shapes, the proposed algorithm exploits frequency invariance of doubly-curved scattering centers to locate possible automobile scatterers in wideband radar data. The algorithm compares the magnitudes of two coarse images constructed from a single set of wideband data. Details of the basis, size of the partitions, and a demonstration with measured radar data are presented. While more work remains, the results begin to establish a performance baseline for rapidly detecting automobiles in time-sensitive applications.