AMPAC, under contract to the Navy, has developed a deterministic, software-only Decoy Detection Algorithm, based on a mathematically proven if-and-only-if theorem, which uses mathematical concepts that rely on geometric features (points and lines) to deduce planar qualities object is 2-dimensional or not. It uses linear algebra on small matrices (~ 20 x 20) and is mathematically exact limited only by image resolution and quality. The algorithm requires two sets of correlated straight line segment endpoint data obtained from two images of the object of interest taken from sufficiently different perspectives. It has been developed to be used with EO sensors, is based on the pin-hole camera model, and is applicable to visible bands, IR, UV, etc. No pre-mission object data is required: databases, templates, etc. No camera data location, orientation, construction is required. Images do not need to be orthorectified. A prototype software system (where a human operator specifies the object of interest) is being developed to accept images from a host/sensor, generate inputs for the Decoy Detection Algorithm, call that algorithm to obtain an answer, and then pass that answer to the host/operator.
Benefit: Our vision of the benefit of this technology is to transform the last mile to the target 0x9D from a sensor data choked bottleneck into a smaller bit stream of enriched information providing a clearer overview of the tactical situation. For example, by showing decoy annotations, generated on the sensor platform via our machine vision approach, on a lower resolution image, the operators assets can be focused on the remaining higher resolution images of genuine targets. Commercially, this system could be expanded to address various aspects of Homeland Security and anti-terrorism efforts for purposes of identification, tracking and threat management.
Keywords: Object/Target Recognition and Identification, Object/Target Recognition and Identification, real-time image processing, Decoy Discrimination, Non-Template Based Target Identification, Camera Orientation Independence
