PASSIVE ELECTRO-OPTICAL SENSOR PROCESSING FOR HELICOPTER OBSTACLE AVOIDANCE

WE PROPOSE TO DEVELOP AND DEMONSTRATE AN INNOVATIVE NEW COMPUTER VISION APPROACH FOR MAPPING A MOVING VEHICLE'S 3-D SURROUNDINGS WITH PASSIVE EO SENSORS. THE PRIMARY APPLICATION IS FOR HELICOPTER OBSTACLE AVOIDANCE. IT IS THEORETICALLY POSSIBLE TO DETERMINE THE 3-D POSITION COORDINATES OF OBJECTS FROM A SEQUENCE OF 2-D IMAGES IF THE VEHICLE TRAJECTORY IS KNOWN. UP UNTIL NOW, HOWEVER, PRACTICAL APPLICATION OF THIS FACT HAS NOT BEEN POSSIBLE FORCOMPLEX, REAL-WORLD SCENCES DUE TO DIFFICULTIES IN IDENTIFYING THE SAME SET OF PHYSICAL POINTS IN SUCCESSIVE IMAGES AND THE NEED FOR ENORMOUS GENERAL-PURPOSE COMPUTATIONAL RESOURCES. OUR APPROACH COMPLETELY ELIMINATESTHE POINT IDENTIFICATION PROBLEM IN A COMPUTATIONALLY ROBUSTAND HIGHLY EFFICIENT MANNER. ALL OF THE COMPUTATIONS REQUIRED CAN BE CARRIED OUT IN REAL TIME WITH MASSIVELY PARALLEL, BULK-PROCESSING TECHNIQUES USING SEMI-CUSTOM VLSI CHIPS TO ACHIEVE EXTREMELY HIGH THROUGHPUT WITH SIMPLE, LOW-COST HARDWARE.