基于YCbCr颜色空间的火焰图像分割方法-陈天炎

TransducerandMicrosystemTechnologies20113010YCbCr1221．3500072．350108。。、、。YCbCr。、。YCbCrTP391．4A1000—9787201110—0062—03FlameimagesegmentationmethodbasedonYCbCrcolorspaceCHENTian-yan1ZENGSi-tong2WUHai-bin21．DepartmentofAutomobileApplicationandEngineeringMachineryFujiantransportationTechnologyCollegeFuzhou350007China2．SchoolofMechanicalEngineeringandAutomationFuzhouUniversityFuzhou350108ChinaAbstractVideofiredetectionhasimportantpracticalsignificanceforfiresafety．Theflamecolorinformationplaysanimportantroleinvideofiredetectionandmanyscholarshavepresentedavarietyofflamecolordetectionalgorithmsbasedondifferentcolorspace．Aimedatlowdetectionratehighfalse-alarmrateandpooradaptabilityincurrentflamecolordetectionalgorithmaflamecolorsegmentationmethodbasedonYCbCrcolorspaceisproposed．TheconstraintsoftheflamepixelsissummedupbystudyingthedistributionoftheflameinYCbCrcolorspaceandtherelationshipbetweentheYCbandCrcomponentsofflamepixels．Theperformanceoftheproposedmethodistestedunderdifferentscenariostheresultsarecomparedwiththreeothermethodsintheliteratureandtheproposedmethodisshowntohaveahigherdetectionratealowerfalserateandbetteradaptability．KeywordsfirerecognitioncolordetectionYCbCrcolorspaceimagesegmentation0、。。。。。HorngWBPengJWHSI12ChenTH34HuangPH5RGB2011—01—18HSILeeBM6、ChenJuan7CelikT8RGB。、。YCbCr。。1YCbCrYCbCrJPEGMPEG。YCbCrY2CbCr26DOI:10.13873/j.1000-97872011.10.02310YCbCr。YCbCrRGB19YCbCr=0．25680．50410．0979－0．1482－0．29100．43920．4392－0．3678－0．0714RGB+16128128．1YCbCrYCbCrHSI。YCbCrRGBY。HSIYCbCr。YCbCrYCbCr。2YCbCrYCbCrRGB1。YCbCrY1623516240。YCbCrYCbCr。1。(a)%原图像(a)%original%image(b)%Y通道(b)%Ychannel(c)%Cb通道(c)%Cbchannel(b)%Cr通道(b)%Crchannel1YCbCrFig1YCbandCrcolorseparationoffireimage1YCrCb。YCbCrCbCr。YmeanCbmeanCrmean。YYYmeanCbCbCbmeanCrCrCrmeanYxy＞YmeanCbxy＜CbmeanCrxy＞Cr{mean．2YmeanCbmeanCrmean3Ymean=1K∑Ki=1YxiyiCbmean=1K∑Ki=1CbxiyiCrmean=1K∑Ki=1Crxiyi3K。。2。(a)%原图像(a)%original%image(b)%人工分割(b)%manually%segmentation2Fig2Manuallysegmentationimageinflameregion2YCrCb1。1YCbCrTab1MeanvaluescalculationofYCbandCrinfireregionYmeanCbmeanCrmean1159．8396．098151．092163．5682．166165．18YCbCrYCbCrCb。RGBRxy≥Gxy＞BxyRxy≥Rmean10YCbCr45Yxy＞Cbxy4Crxy＞Cbxy5YxyCbxyCrxyxy。1CbCrCbCrCb3630“”Cr“”。|Crxy－Cbxy|≥τ．6τ。τ1～100τ。τ。τ3596．5%15%。YCbCr7rule1Yxy＞Cbxyrule2Crxy＞Cbxyrule3Yxy＞Ymeanrule4Cbxy＜Cbmeanrule5Crxy＞Crmeanrule6|Cbxy－Crxy|≥τ．73。。3RGB-S10RGB2HSI。。3。(a)%原始图像(a)%original%image(b)%RGB-S法(b)%RGB-S%method(c)%RGB法(c)%RGB%method(d)%HSI法(d)%HSI%method(e)%本文方法(e)%the%proposed%method3Fig3Camparisonofflamedetectionresults34RGB-SHSIRGB-SRGB。YCbCr。3Fig3Continued4、。1HorngWBPengJW．Image-basedfiredetectionusingneuralnetworksC∥ProceedingsoftheJointConferenceonInformationSciencesJCISKaohsiungTaiwan2006．2HorngWBPengJWChenCY．Anewimage-basedreal-timeflamedetectionmethodusingcoloranalysisC∥ProcedingsofIEEENetworkingSensingandControlTucsonArizonaUSA2005100－105．3ChenTHWuPHChiouYC．Anearlyfire-detectionmethodbasedonimageprocessingC∥ProceedingsofIEEEInterna-tionalConferenceonImageProcessingICIP’04Singapore20041707－1710．4ChenTHKaoCLChangSM．Anintelligentreal-timefire-de-tectionmethodbasedonvideoprocessingC∥The37thIEEEInternationalCarnahanConferenceonSecurityTechnologyTai-wan2003104－111．5HuangPHSuJYLuZMetal．Afire-alarmingmethodbasedonvideoprocessingC∥InternationalConferenceonIntelligentInformationHidingandMultimediaSignalProcessingIIH-MSP’062006359－364．6LeeBMHanDongilReal-timefiredetectionusingcamerase-quenceimageintunnelenvironmentC∥ProceedingsofICIC20071209－1220．7ChenJuanHeYapingWangJian．Multi-featurefusionbasedfastvideoflamedetectionJ．BuildingandEnvironment2010451113－1122．8CelikTDemirelHOzkaramanliH．AutomaticfiredetectioninvideosequencesC∥ProceedingsofEuropeanSignalProcessingConferenceEUSIPCO2006FlorenceItaly20069．9PoyntonCA．ATechnicalIntroductiontoDigitalVideoM．NewYorkWiley1996．10CelikTDemirelHOzkaramanliHetal．FiredetectionusingstatisticalcolormodelinvideosequencesJ．JournalofVisualCommunication＆ImageRepresentation200718176－185．1981－。46