Abstract:Inordertosolvetheproblemsofthepoornoiseimmunityandlowrobustnessinthevideolanedetection

afastlanedetectionmethodbasedonnewfeatureextractionandclassificationwasproposed.Thesetofmidpointsofoptimalbeamletswasextractedafterusingthebeamlettransformationinthepreprocessedgrayscaleimage

whichwasappliedtotheanalysisofimprovedK-meansclustering.ThepointsineveryclasswasfittedastraightlineorcurvebyusingtheRANSACalgorithmbasedonthreeordersbezierspline.Thesimplifiedbeamletdictionaryandextractingbeamletbaserapidlygreatlyspeedupthebeamlettransform.ThesearchforthebestprojectionlinesandthemultipleiterationsofclusteringcenterwereusedtoimprovetheK-meansclustering

whichsolvestheclusteringproblemofcurvelaneandthenumberoflane.Theresultshowedthatthealgorithmhasgreatreliability

real-timeandrobustnessinthestructuredorunstructuredlaneenvironment.