视觉导航智能车的道路识别及障碍物检测算法分析-road identification and obstacle detection algorithm analysis of visual navigation intelligent vehicle.docxVIP

AbstractThevisualinformationisoneofthemostimportantinformationthatcanbeobtainedfromthetrafficenvironment.Themaincontentofthispaperisapproachingtoroadboundaryinformationextractionandrecognitionbyusingthemachinevisionmethod.Besides,thereisamothedtodetectionvechilesbasedontheHaar-likefeatures.Bydoingthiscanprovidethenecessarybasistointelligentvehicles.Inthispaper,amodelofcamershasbeenworkdout.Preprocessingtheimageoflandsbythecolorspacestransform,histogramequlizationandothermethos.Denosingimagesusingtheneighborhoodaveragemethod.UsingtheCannyoperatortogetbinaryimages.IncasethattherewillbemuchmorelinesbyHoughtransformonbinaryimages,soweusethemethodthatgroupsseekingtofindtherequiredmaximumtogottheboundarysofroad.ftertheinitiallaneboundaryhasbeendetected,theROIcanbeobtainedbasedontheinitiallaneboundary.ThentheHoughtransformontheROIhasbeentaken,thelaneboundarycanbefoundout.Inthisway,thealgorithmcanspeedupcomputationandimprovingtheaccuracyofthelaneboundaryidentification.Attheendofthispapertheaccuracyofthealgorithmisanalyzed.Theroadobstacledetectionmostlybethedetectionofothervehiclesontheroad.Therearetwomethodshadbeenworkedout,oneofthemisbasedonthefeaturesofthefrontviewandbackviewofavehicle,theotheroneisbasedontraining.Inthefirstmethod,threefeaturesofvehiclehavebeenused.Thefirstoneistheshadowbelowthevehicle,andtheshadowhasbeendetectedusingthresholdonthegrayspace.Thesecondfeaturehasbeenusedisthesymmetryofavehicle,asymmetrymeasurefunctionisusedtodeterminedisthereavehicleintheimage.Thelastfeatureisthetaillightofavehicle,anditcanbedetectiveeasilyintheHSIcolorspace.ThesecondwaytodetectiveavehicleissomekindoftrainingthatbaseonHaar-likefeatures.Togetacascadeclassifierthathasbeencompensativebyanumberofweakclassifier,thetrainingusedahugenumberofsamples.Inthemethodtheimagethattobeprocessingistransformedtointegralmap.Afterthataseriesofdifferentdetectivewindowsscanningoverthewholeimage,andgetlossofregionsthatcontainingavehiclewithdifferentsize,bytheendthosesmallregionshavebeencombinedtoab