SBIR-STTR Award

The Midcourse Target Identification (MCTID) Algorithm for Midcourse Threats Using Bayesian-Topological Feature Fusion and Classification program will extend the previously developed Enhanced Target Identification (ETID) algorithm developed under a previous MDA program to the midcourse threat regime. ETID was built to identify boost phase threats by performing both topological threat classification and Bayesian probabilistic threat classification, and then fusing the two classification results. The ETID achieved a predicted 95% classification accuracy for a broad range of possible strategic and theater threat boosters. The MCTID will employ the same dual mode classification and target ID approach, but apply it to midcourse threat and decoy types. The goal of the MCTID is a 99% accuracy of correct classification. The proposed program will develop the algorithm, test it against real and simulated data, and report the results. The algorithm development will utilize dedicated software coding, as well as generalized COTS software packages for building the topological and Bayesian classifiers.

Target identification for the boost and early ascent (BEA) phase, and the midcourse (MC) phase is one of the most challenging problems for the BMDS. The Multi-Classifier Target Identification project employs two synergistic classifier techniques - a probabilistic Bayesian Classifier and a deterministic Topological Classifier – to develop target classifications using weighted target features. These individual classifications, together with outside classifications, are then fused, based on adaptive weightings, to produce a final classification. Phase I testing against realistic simulated data showed promise for both BEA and MC threats. In this Phase II effort, the MCTID technology will be developed by The CORE Group team, and tested in the MDA X-Lab against real recorded and live BEA data. Development and testing will be accomplished in two threads- one using Overhead Non-imaging Infrared (ONIR) data only as provided by the Common Mission System (CMS) from Lockheed Martin IS&S, and the second using a more robust set of real and simulated sensor data sources from current and planned BMDS sensors. Development will include a TID operator interface employing a state-of-the-art human machine interface (HMI). The MCTID will be hosted on a Field Test Unit (FTU) for testing at multiple sites.

Keywords:
Classification, Missile Defense, Target Identification, Onir, Bayesian, Data Fusion, Topological