Statistical nonrecursive spatial-temporal focal plane processing for background clutter suppression and target detection.

Abstract

Advanced surveillance and weapon guidance systems using new mosaic sensor arrays and large scale integration (LSI) electronic data processors on the same focal plane require detection of weak targets deeply buried in background clutter noise by many tens of db's. This investigation reports on the focal plane processing techniques to accomplish these heretofore unachievable goals. Five focal plane processing algorithms are developed consisting of nonrecursive statistical spatial-temporal filters for clutter suppression followed by thresholding for initialization of target detection. These filters are based on either the minimization of mean square error criterion (for MMSE filters) or the maximization of signal to noise ratio criterion (for matched filters). Two are nonadaptive spatial filters and two are nonadaptive spatial-temporal filters. The fifth type is an adaptive spatial filter based on the minimization of mean square error criterion. These filters have been investigated analytically and by computer simulation using computer generated images containing correlated clutter noises modeled by Markov processes and also real world infrared images. Using an infrared image in the red spike spectral band, a single frame statistical spatial filter can suppress the clutter by 27 db. A five frame sequential statistical spatial-temporal filter was found to have a clutter suppression of 87 db.