计算机专业英语学术论文.pdf

ComputerTechnologyandApplication3(2012)361-367 … MonitoringFreewayIncidentDetectionUsingaHotelling ControlChart JoonseLim’一． YoungSeonJeong andYoungsulJeong 1．PolytechnicInstituteofNewYorkUniversity，SixMetroTechCenter,Brooklyn．NewYo 11201．USA 2．GeorgetownPreparatorySchool，10900RockvillePike，NorthBethesda，MD 20852，USA 3．DepartmentofIndustrialandSystemsEngineering,KhalifaUniversiytofSeienee，TechnologyandResearch，AbuDhabi，UAE 4．DepartmentofNaturalSciences，WashingtonBaptistUniversiyt，Annandale，VA22003，USA Received：March12，2012／Accepted：M arch31，2012／Published：M ay25，2012 Abstract：Inreal-liferfeewaytransportationsystem，afewnumberofincidentobservation(veryrareevent)isavailablewhilethereare largenumbersofnormalconditiondataset．Mostofresearchesonfreewayincidentdetectionhaveconsideredtheincidentdetection problem asclassification one．However，becauseofinsufficiencyofincidentevents，mostofpreviousresearcheshaveutilized simulatedincidenteventstodevelopfreewayincidentdetectionmodels．Inordertoovercomethisdrawback．thispaperproposesa wavelet-basedHotelling controlchart forfreeway incidentdetection．whichintegratesawavelettransform intoanabnorma1 detectionmethod．Firstly，wavelettransform extractsusefu1featuresrfom noisyoriginaltraffi cobservations．1eadingtoreducethe dimensionalityofinputvectors．Then．aHotelling controlchartdescribesadecisionboundarywithonlynormaItrafficobservations withtheselectedfeaturesinthewaveletdomain．Unliketheexistingincidentdetectionalgorithms．whichrequirelotsofincident observationstoconstructincidentdetectionmodels，theproposedapproachcandecideadecisionboundary givenonlynormaltraining observations．TheproposedmethodisevaluatedincomparisonwithCaliforniaalgorithm．M innesotaalgorithm andconventional neuralnetworks．Theexperimentalresultspresentthattheproposedalgorithm inthispaperisapromisingaltemativeforrfeeway automati