DEVELOPMENT OF MACHINE VISION OPTICAL FIRE DETECTOR SYSTEM

THE CONCEPT PROPOSED UTILIZES MACHINE VISION TECHNOLOGY, PATTERN RECOGNITION AND ARTIFICIAL INTELLIGENCE TO IDENTIFY "CHANGES" IN VISUAL SCENES THAT HAVE BEEN STORED IN A COMPUTER. CHARACTERISTICS OF EACH LABELLED "CHANGE" AREA, SUCH AS GROWTH AND INTENSITY (BRIGHTNES) ARE MONITORED ALONG WITH THE SPECIAL SIGNATURES IN THE VISIBLE. THE SYSTEM IS DEIGNED TO BE IMMUNE TO IR AND UV FALSE ALARM SOURCES AND CAN LOCATE AND DETERMINE THE SIZE OF POSSIBLE FIRE EVENTS.THE RESPONSE TIMES TO "FIRE DECISIONS" AND SCENE CHANGE/EDGE DETECTION ANALYSIS ARE IN THE MILLISECONDS RANGE, THUS INCREASING CONSIDERABLY THE RELIABILITY OF PRESENT UV/IR DETECTORS AND PROVIDING VERY FAST REACTION TIMES. THE SYSTEM OPERATES IN THE SAME "VISUAL AND REASONING PROCESS" AS A HUMAN. THE FIRE IS SEEN IN THE VISIBLE AND OBSERVED FOR SOME BRIEF PERIOD TO DETERMINE ITS CHARACTERISTICS IN COLOR, GROWTH, SIZE, AND OTHER FEATURES AND TO COMPARE THESE ATTRIBUTES/FEATURES AGAINST KNOWN EXPERIENCE/STORED KNOWLEDGE. COMPARISONS ARE DONE ON WHAT OTHER ITEMS MAY CAUSE SUCH AN EVENT, TO REDUCE POSSIBILITIES OF A WRONG DECISION. THE SYSTEM USES OTHER SPECTRAL DATA INPUT IN ITS DECISION-MAKING PROCESS TO FURTHER ASSURE A HYDROCARBON FIRE IS PRESENT. IT CAN ALSO DISTINGUISH IF THE EVENT IS ASSOCIATED WITH THE AIRCRAFT AND WHERE IT IS LOCATED, THUS ALOWING ZONAL SUPPRESSION TO BE APPLIED.