Inclusion of filter modulation in synthetic-discriminant-function construction

Abstract

A technique in which the filter modulation is included in the synthesis of a synthetic-discriminant-function matched spatial filter (SDF MSF) is presented. In the filter synthesis, a system of simultaneous nonlinear equations is solved with an iteration procedure. A computer simulation of the new method using thresholded images of the Space Shuttle over a range of aspect angles was performed for phase-only filters (POFs) and binary-phase-only filters (BPOFs). The filters constructed are capable of obtaining the specified peak-correlation response to within 1% with a high signal-to-clutter-ratio for the one-class problem, the two-class problem, and the multilevel problem. In contrast, conventional projection SDF POFs and BPOFs are unable to produce the desired peak-correlation response.

Computation of arbitrarily constrained synthetic discriminant functions

“…In contrast to case 1a2, they contain no information about the problem. 1c2 The same starting point used in the Jared and Ennis algorithm, 11 i.e.,…”

Higher-order neural networks applied to 2D and 3D object recognition

“…One demonstrated approach for 3D object recognition is optical correlation using binary synthetic discriminant function (BSDF) filters (Jared & Ennis, 1989). Optical correlation is based on the Fourier transform operator, which is invariant over the object's position in the input field.…”

The design of a composite wavelet matched filter for face recognition using breeder genetic algorithm

“…The CWMF can be constructed if the target and the antitarget images are known a-priori. A CWMF has been designed earlier for invariant pattern recognition using the delta learning rule [21]. The CWMF is defined in the frequency domain as [11,12] F ðu; vÞ ¼ X n i¼1 a i T i ðu; vÞ !…”