Robust Real Time Detection and Recognition ofTraffic Sign Boards
Akash Kumar, Akshay Dutt, Gautam Saini · 2014
Automatedrecognitionoftrafficsignboardsfaci litatesthegrowthofemergingtechnologyforaut onomousvehicles.Here,weimplementanovelfe atureextractionstrategyfortrafficsignboardsr ecognition.AnoveltreestructuredmulticlassAd boostalgorithmisproposedtouseasaclassifier. AnewdatasetoftrafficsignboardsaroundVellor ecityiscreatedandusedinourexperiment.Thed atasetcoverswholerealtimescenarioswithlotofvariations,e.g.,varyin glightingconditions,dissimilarbackground,m otionblur,dayandnightetc.Ourproposedmetho dexhaustivelyscansallthefeaturesandselectth ebestoptimumfeaturetoreducefalsealarmandi mproverecognitionrate.Comparisonwiththee xistingmethodsisdonetodemonstratetheperfor manceoftheproposedmethod.Furtherimprove mentsarealsodiscussed.