MULTILEVEL MOTION PROCESSING FOR AUTONOMOUS HELICOPTERS

A FUNDAMENTAL PROBLEM WITH AUTOMATING NAP-OF-THE-EARTH HELICOPTER FLIGHT IS THE EXTRACTION AND REPRESENTATION OF INFORMATION ABOUT THE RAPIDLY CHANGING RELATIONSHIP BETWEEN THE HELICOPTER AND THE SURROUNDING ENVIRONMENT. WE TREAT MOTION ANALYSIS AS INVOLVING SEVERAL EXPLICIT LEVELS OF PROCESSING AND REPRESENTATION. THE DIFFERENT LEVELS CONSTRAIN AND DIRECT EACH OTHER SIGNIFICANTLY FOR DETERMINING IMAGE MOTION AND BUILDING ENVIRONMENTAL INFERENCES. THE LEVELS RANGE FROM VERY LOCAL AND SIMPLE DESCRIPTIONS OF CHANGES IN IMAGE INTENSITY OVER TIME TO QUALITATIVE AND POTENTIALLY ROBUST DESCRIPTIONS OF THE RELATIVE DEPTHS OF COHERENT ENVIRONMENTAL SURFACES AND OCCLUSION BOUNDARIES. THIS RANGE OF REPRESENTATION AND CONSTRAINTS CAN EFFECTIVELY ADDRESS PROBLEMS ASSOCIATED WITHTHE TYPICALLY BRITTLE ENVIRONMENTAL INFERENCES ASSOCIATED WITH MOTION PROCESSING. THE RESULTING FRAMEWORK WILL SUPPLYNASA WITH AN EXTENSIVE SET OF ALGORITHMS THAT CAN BE EXPERIMENTED WITH AND COMBINED. THIS IS PARTICULARLY IMPORTANT SINCE THE EXACT USE AND NATURE OF MOTION PROCESSING FOR HELICOPTER SYSTEMS WILL VARY BASED UPON SEVERAL EVOLVING FACTORS.