Clutter suppression algorithm for nonimaging data

A clutter suppression algorithm based on principle component analysis for overhead non-imaging infrared (ONIR) focal plane data is presented and its performance evaluated. The algorithm is novel in that (1) the principle components are derived from a small subset of the pixel histories and (2) probable target pixels are reevaluated in a way that substantially reduces the “lost energy” usually associated with subspace projection. The algorithm scales very favorably to larger focal planes because the principle components are derived from a small number of pixel histories. The algorithm is highly vectorizable and very well suited for parallel processing. A series of 43 actual data collections was analyzed. The raw data had mean standard deviations ranging from 4 to 4375. Clutter was reduced by over a factor of 10 using 8 principle components and over 30 for high clutter cases using 24 components