Electro-Optic (EO) Feature-Based Target Combat Identification (CID)

We propose to design, develop and evaluate a deterministic approach to combat identification (CID) of stationary and moving objects using a single video stream. From a video containing a moving target, we detect and track the target using an appearance-based tracker. Stationary targets in the video are extracted using a shape-driven operator. Using fingerprints extracted from the segmented target chip and synthesized views of targets generated from a homgraphy derived from the given video sequence, quasi-invariants and a generalized Hausdorff metric, we design a target CID algorithm. A special case of our algorithm handles the target CID problem for stationary targets. Using the results of target segmentation, we will develop an efficient compression scheme for transmitting the target chip over noisy channels. During Phase I, we will assess the potential performance of a stationary and motion-based video CID capability by integrating a suite of algorithms for the tasks mentioned above. The resulting working software prototype will help us to identify challenges to be addressed during Phase II. During Phase II, we will demonstrate and evaluate a prototype system capable of robust verification/identification in video imagery. Our goal is to sue as few a frames as possible in determining CID.