基于神经模糊算法的工业消防安全计算机视觉实现(IJITCS-V7-N4-2)

I.J.InformationTechnologyandComputerScience,2015,04,14-27PublishedOnlineMarch2015inMECS()DOI:10.5815/ijitcs.2015.04.02Copyright©2015MECSI.J.InformationTechnologyandComputerScience,2015,04,14-27ImplementationofComputerVisionBasedIndustrialFireSafetyAutomationbyUsingNeuro-FuzzyAlgorithmsManjunathaK.C.M/s.PrakashSpongeIronandPower(P)Ltd(ERMGroup),Chitradurga-577501,IndiaEmail:[e-mail],[e-mail].S,Dr.P.AVijayaProfessor,Dept.ofIT&ECMalnadCollegeofEngineering,Hassan-573201,IndiaEmail:[e-mail],[e-mail]-Acomputervision-basedautomatedfiredetectionandsuppressionsystemformanufacturingindustriesispresentedinthispaper.AutomatedfiresuppressionsystemplaysaverysignificantroleinOnsiteEmergencySystem(OES)asitcanpreventaccidentsandlossestotheindustry.Arulebasedgenericcollectivemodelforfirepixelclassificationisproposedforasinglecamerawithmultiplefiresuppressionchemicalcontrolvalves.Neuro-Fuzzyalgorithmisusedtoidentifytheexactlocationoffirepixelsintheimageframe.Againthefuzzylogicisproposedtoidentifythevalvetobecontrolledbasedontheareaofthefireandintensityvaluesofthefirepixels.Thefuzzyoutputisgiventosupervisorycontrolanddataacquisition(SCADA)systemtogeneratesuitableanalogvaluesforthecontrolvalveoperationbasedonfirecharacteristics.Resultswithbothfireidentificationandsuppressionsystemshavebeenpresented.Theproposedmethodachievesupto99%ofaccuracyinfiredetectionandautomatedsuppression.IndexTerms-OnsiteEmergencySystem,SCADA,PLC,WeightedCentroid,FirePixelNumber,Neuro-FuzzyAlgorithm.I.INTRODUCTIONVision-basedfiredetectionandautomatedsuppression(VFDAS)systemsareoneofthemostimportantmechanismsinmanufacturingandprocessindustries.Itismorecriticalinindustrieswhichuseoil,gasandpetrochemicalsasfuels.Afastautomateddetectionsystemhastobereadyinordertopreventanyfireaccidentsandavoidlossoflifeandproperty.VFDASisnewlydevelopedtechniquebasedoncomputervision,imageprocessingandneuro-fuzzyalgorithms.Vision-basedfiredetection(VFD)hasmanyadvantagesovertraditionalmethodssuchhasfastresponse,non-contactandnoinstallationlimitations.Currently,mostofthefiredetectionsystemsusesensorsfordetectingsmoke,riseintemperatureetc.Theyneedaconsiderabletimeforrespondingasthesensorsrequireproductoffire(e.g.,smoke,temperatureetc.)toreachthesensors.Alsotheyhavetobecarefullyplacedinselectedlocations.Suchasensor-basedfiredetectionsystemisunsuitableforlargeindustrieswithopenspaceswhenproductsofcombustionmaybecomeoutofreachofthesensorsandcanreducethepossibilityofdetection.Vision-basedfiredetectionandautomatedsuppressionsystemoffersseveraladvantages.First,thecostisless,asthissystemisbasedoncamerasandindustriesaremostlyequippedwithCCTVsforsurveillance.SCADAandPLCmayalsobepresentifthereisprocessautomation.Second,theresponsetimeisfasterasitdoesnothavetowaitforanyproductofcombustion.Firesuppressionchemicalcontrolvalvewillbeoperationalatthestartoffireitselfthusreducingthescopeforspreadingoffire.Finally,incaseoffalsealarm,confirmationcanbedonefromacontrolroomwithoutrushingtothelocation.Asitisimportanttohaveafastfiredetectionandsuppressionsystem,acomputervisionbasedtechniqueisproposedinthispaper.Thispaperinitiallyfocusesonvideoandimageprocessingforflamepixelsdetection.Oncethefireisconfirmedthefocusisonthecomputationoflocationandintensityofthefireusingneuro-fuzzyalgorithms.Finally,theattentionisonthefiresuppressionchemicalcontrolvalveoperationthroughSCADA,PLCandI/Oconfiguration.Inindustries,criticalareaswhicharepronetofireaccidentsareidentifiedbyasurvey.CCDcamerasareinstalledintheseareaswithpropersceneplanning.Alsoanumberoffiresuppressionchemicalcontrolvalvesareinstalledwhichareconnectedtocentralizedchemicalpumpingstation.Seriesofimageframesfromthecontinuousvideostreamisacquired.Theseframesareprocessedwithastoredimagetakenwhenthesituationisnormal-consideredasthebackgroundimage.Successiveframes(ithandi+1th)ofthevideostreamandnormalbackgroundimage(j)areprocessedatatime.Firerulesareappliedtoidentifythefirepixels.Insecondstage,theparticularlocationwherefirehasoccurredisidentifiedbyusingneural-network.Finally,thefiresuppressionchemicalcontrolvalveinthatlocationisoperatedwiththehelpoffuzzylogic.ThecontrolsystemcommunicationisdonethroughPROFIBUSandI/Osignalconditioningcircuits.Thispaperisstructuredinsuchawaythatliteraturepertainingtorecentdevelopmentsincomputervisionandtheirapplications.ProblemisdefinedbasedonImplementationofComputerVisionBasedIndustrialFireSafety15AutomationbyUsingNeuro-FuzzyAlgorithmsCopyright©2015MECSI.J.InformationTechnologyandComputerScience,2015,04,14-27fundamentalstudyandpracticalexperience,thefirecharacteristicsarestudiedinprofoundmannerfortheimplementation.Methodologiesareframedfortheautomatedfiresuppressionsystemandsequentialimplementationhasbeendonetoachievetheanticipatedresults.Allthesequentialresultsarepresentedanddiscussedinthevisionofperformanceoftheimplementedsystem.II.RELATEDWORKSSeveralapproacheshavebeensuggestedinliteraturetoidentifyfirebyusingvarietyofimageprocessingtechniqu