Discrimination

We propose to create a discrimination engine for the Ballistic Missile Defense System (BMDS) based on the concept of manifold learning algorithms. Manifold learning algorithms have come to prominence within the computer vision community as a type of feature extraction algorithm which does not require specific features to be learned a priori. High dimensional sensor data is input to these algorithms which find a lower dimensional representation of the same data, illustrating differences between potential target types. The algorithm can explore regions of the data space for which there are no feature extraction algorithms. We propose to apply these algorithms to high resolution range profiles of target features and embed these results in a novel inference engine to create an on-line discrimination system. Our team consists of staff who have contributed to the development of a prototype system based on these algorithms, as well as radar engineers who can build complex simulated data modeling real objects. In additional Decisive Analytics has a long history of transitioning SBIR technologies to end-users and will work with our partner in this effort, Raytheon, to integrate this discrimination system into an operational radar platform.