Optimal Three-Dimensional Matched Filter Processing for Detection of Point-Like Moving Objects in Clutter
Abstract
A simple model of a time sequence of star images containing a moving point object (satellite) is developed. Optimal signal enhancement and detection processing theory is applied to this model and a three-dimensional Fourier matched filter implementation is derived to compute clutter-to-noise ratio (CNR) suppression, signal-to-noise ratio (SNR) enhancement, and probability of detection (Pd) and false alarm (Pfa) rate estimates as a function of input single pixel SNR. Using this theory allows one to compute the Fourier domain matched filters directly, thereby eliminating the enormous storage cost associated with large banks of three-dimensional matched filters. The model and theory are tested using computer-generated simulated data sets having known noise and clutter characteristics. The theory is then applied to real data sets collected using a Lincoln Laboratory CCD camera. Initial results indicate good agreement with matched filter theory. Detection of objects with an initial single pixel SNR approx. 1 is demonstrated.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 30, 1992
- Accession Number
- ADA259727
Entities
People
- John N. Sanders
Organizations
- Massachusetts Institute of Technology