Video Analysis Using Perceptual Organization and Machine Learning

ABSTRACT: The goal of this project is to develop, demonstrate, and implement at EAFB the EVASE:"Extensible Video Analysis of Symbology Events"Matlab toolkit. This system is based on the premise that it will learn symbology events during training by an expert operator, and then autonomously detect this symbology in long video sequences, resulting in a list of key events to assist the operator in further analysis. During the Phase I, we developed the EVASE Matlab toolkit which includes a GUI, video preprocessing and noise removal, symbol event detection, symbol recognition, and operator training. Finally, we demonstrated the EVASE toolkit using F-16 HUD video with a focus on symbology from armament, radar, and avionics. During the Phase II we will expand the EVASE Matlab toolkit to include a comprehensive list of events from armaments, radar, and avionics and along with the F-16 include other types of aircraft. BENEFIT: The EVASE system will enhance the ability of the Air Force to manage flight test and simulation projects by providing software to see and recognize symbolic information from the HUD and instrumentation.