资源描述:
《基于单目视觉的车辆主动安全技术分析》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、ABSTRACTABSTRACTInrecentyears,withtherapidincreaseofvehicles,trafficaccidentsalsoincreasedrapidlycausedbyvehiclecollisionandlanedeparture,thenumberofcasualtiesisrisingyearbyyear,andbecomeapublichazardinmodernsociety.Recently,vehicleactivesafedrivinghasattractedmor
2、eandmoreattentionofsomeresearchinstitutionsandvehiclemanufacturers.ComputerVisionbasedvehicledetection,vehiclecollisionwarning,lanedetectionandlanedeparturewarningisoneimportantresearchareaofvehicleactivesafety.DuetothelowpriceofVisionsensors,flexiblealgorithm,and
3、collectinginformation-rich,iswidelyusedinvehicleactivesafety.Inthispaper,wemainlystudyonthetechnologyofactivesafetybasedonmonocularvision,collectionofenvironmentalinformationbythecamerainstalledinfrontofthevehiclesandprocessingmeanwhile,realizationvehicledetection
4、,vehicledistancemeasuring,lanedetection,andLaneDepartureWarning.Themainworksandresearchresultsareasfollows:1.Inordertobuildupthemappingrelationshipbetweenimageandspatialobjectandvisionmeasurement,thecameraneedtocalibration.Atfirstweintroducedthemaincameracalibrati
5、onMethodcurrently,thenaccordingtotheresearchsubjectusingmonocularcamera,drawupasimple,practicalandhigh-accuracycalibrationmethod.2.Studyavehicledetectionmethodusingbasedongradientorientationhistogram,usinglinearsupportvectormachinetrainingtheHOGcharacteristicsextr
6、actedfromsamplesthenreceivedabinclassifier,applytheclassifiertovehicledetection.Accordingimageprocessingandgeometrictransformtheory,combinedwithvehicledetectionresultinformationtorealizedistancemeasuringofthefrontvehicles.3.Usingcannyoperatorextractionimage’sedgea
7、ndedgedirection’sinformationtoimprovethetraditionalHoughtransform,applyimprovedHoughtransformrealizelanedetection.AccordinglaneDetectionresultandthecurrentdrivingsituationdeterminesthelaneparameters.4.Firstgiveabriefintroductionofdecision-makemethodforlanedepartur
8、ewarning.II万方数据ABSTRACTPresentacalculationmethodforvehiclelateralshiftdistanceandvehicleyawangle,accordingtotheseinformationestablishmentsawarningmethod