基于统计学习的机场跑道异物检测

《基于统计学习的机场跑道异物检测》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、ABSTRACTABSTRACTFODdetectionforairportrunwayisanemergingissueandhasbeenproposedinrecentyears.AllresearchesaboutFODdetectionincludemillimeterwaveradartechnologyandimageprocessingtechnology.Ifweadoptmillimeterwaveradartechnology,thecostofFODdetectionwillbeveryhigh;thesiz

2、erequirementsofFODwillbehigh;therequirementsofhardwaredevicewillbeveryhigh.Moreover,Chinadoesn’thavematureandrelevantcommercialsystems.Ifweadoptimageprocessingtechnology,thecostofFODdetectionwillberelativelylow;therequirementsofhardwaredevicewillberelativelylow.Moreove

3、r,imagemonitoringsystemhasbeeninstalledinthevicinityofrunwayinsomeairports.Thus,inthispaper,weadoptimageprocessingtechnologytoachieveFODdetection.Inthispaper,thetargetdetectionmethodbasedonstatisticallearningisappliedtoFODdetectionsystem.Wemakeuseofmaturefacedetections

4、ystem;keepAdaboostclassifiermethodcommonlyusedinfacedetectionsystem;improveorexcludefeaturescommonlyusedinfacedetectionsystem;findasuitablefeatureforFODdetectionsystem.FirstweintroduceLBPfeaturecommonlyusedinfacedetectionsystem;applyLBPfeaturetoFODdetectionsystem;verif

5、ythefeasibilityofLBPfeaturethroughexperimentalresults.TheexperimentalresultsshowthatLBPfeatureisnotfeasible.Accordingtothecharacteristicsofairportrunwaypictures,weproposeanewhistogramfeaturewhichisbasedonSUSANfeature.Thentheexperimentalresultsshowthatnewfeatureisfeasib

6、le.Secondlyinthepicturestodedetected,therearealotoffalsealarmsaroundtherunwaylines.Tosolvetheproblemweproposeawaybasedonimagesegmentationtoremovethefalsealarmsaroundtherunwaylines.Wefindthepositionsoftherunwaylinesthroughmarkingedgepoints.Accordingtothepositionsoftheru

7、nwaylines,wedividethepicturesintopartscontainingtherunwaylinesandpartswithoutcontainingtherunwaylines,respectivelymakeuseoftargetdetectionbasedonKirschfeatureandstatisticallearning.FinallyweintroducetheprincipleandtrainingprocessesofAdaboostclassifier.Themaincontributi

8、onsofthispaperaresummarizedinthefollowingthreepoints:(1)AnewhistogramfeaturebasedonSUSANfeatureisproposedWeproposeane