Aquatic Vessel ATR using Structural Traits (AVAST)

To maximize situational awareness and survivability of Navy assets in littoral environments, Navy submarines must be able to quickly distinguish between hostile targets and similarly-sized non-hostile vessels. Currently, skilled personnel determine if each contact is hostile. This approach is limited by available operators, the classification speed and accuracy of each operator, the number of recognizable ship types, and operator attention spans. Automation is the key to reducing operator workload, and vision-based automatic target recognition (ATR) techniques will allow less skilled personnel to more accurately identify a wider variety of marine targets without loss of focus in a variety of challenging conditions. Model-based classification methods hold much promise for solving the ship identification problem in the varying conditions faced by Navy submarines. We propose a model-based ATR approach called Aquatic Vessel ATR using Structural Traits, or AVAST. AVAST extracts ship silhouettes from images, and derives a skeleton to model the ship’s main structures. The relative geometry and real-world measurements of these structures are computed and then used to identify matching ship types in a database of known vessels in real-time (less than 1 second per target). The matching ship types are then displayed for operator confirmation and, if necessary, refinement.