基于光流法的煙霧圖像運動特征判別模型及多信息融合探測算法研究.pdf

1、中國科學(xué)技術(shù)大學(xué)博士學(xué)位論文基于光流法的煙霧圖像運動特征判別模型及多信息融合探測算法研究姓名：于春雨申請學(xué)位級別：博士專業(yè)：安全技術(shù)及工程指導(dǎo)教師：張永明20100419AbstractIIABSTRACTWiththerapiddevelopmentoftheeconomyhomethehighbuildingsundergroundconstructionstunnelconstructionscomplexbuildingsare

2、increasingdaybyday.Oncethereisfireintheseplacesitoftencausesheavycasualtieseconomiclosses.Howeverasthemostwidelyusedfiredetectiontechnologypointtypephotoelectricsmokedetectionbelongstocontactpassivedetectionfmwhichisspat

3、iallimitedsoitcannotsatisfythefiredetectionrequirementoftheseplaces.Withthedevelopmentofmachinevisioncomputertechnologythevideofiredetectiontechnologyiscomeintobeinginthelatestyearswhichhasremarkableadvantagesovertraditi

4、onalmethodssuchasfastresponsewidedetectionarealittleenvironmentpollutionitcanbeusedfthefiredetectioninhighbuildingsevenoutdoenvironmentitrepresentsrelativelyhighlevelofthefiredetectiontechnologynowadays.Theearlystageofth

5、edevelopmentofvideofiredetectiontechnologybeganwithvideoflamedetection.Basedonthepeople’spriiknowledgeoffiremostfiresbeginwithsmoulderingstageinwhichthesmoketurnupearlierthanflamesothatthevideosmokedetectioncouldrealizev

6、eryearlyfiredetection.Severalclassicsmokeimagefeaturesitsextractingmethodsincludingcolblurtexturecontourareanalyzedinthepapertherealtimesmokeimagedetectionbasedontexturefeaturesisrealizedatthesametime.bymeansoftheresults

7、oftherecognitionofvideolibraryanalyzeeachofthesesmokeimagefeatureextractionmethods’goodpointbadpointapplicablescenes.Studytheapplicationofopticalflowmethodinvideofiresmokerecognition.Extracttheopticalflowofthefeaturepoin

8、tsinsmokeimageregionsusingtheopticalflowmethodestablishthemotionvelocityfieldindertoreflectthevariationtrendofeachgraypixelpointsintheimageplane.Analyzethemotionfeaturesvelocityfield’sdistributionacteristicsofsmokeimages