Real-time classification of multiple non-separated battlefield ordnance events using ELMO

With long range remote sensors observing high intensity conflicts on a battlefield, it is possible to observe multiple non-temporally separated ordnance events in a single pixel. ELMO (Event Location and classification in Multiple Ordnance profiles) is a real-time algorithm for classifying these multiple ordnance events. The algorithm has three basic steps. The first step estimates delays of events in a composite signal by computation of a peak function. The second step classifies the events by a least squares fit involving a banded matrix developed using the delay information. The third step allows for an error correction based on comparison of the computed composite signal with the measured composite signal. Trials show ELMO to exhibit good performance in cases where events have reasonably small variances from the class mean, and where the class means are linearly independent.