预览加载中,请您耐心等待几秒...
1/3
2/3
3/3

在线预览结束,喜欢就下载吧,查找使用更方便

如果您无法下载资料,请参考说明:

1、部分资料下载需要金币,请确保您的账户上有足够的金币

2、已购买过的文档,再次下载不重复扣费

3、资料包下载后请先用软件解压,在使用对应软件打开

基于SqueezeNet卷积神经网络的车辆颜色识别 Abstract Vehiclecolorrecognitionisanimportanttaskinthefieldofcomputervisionandhasawiderangeofapplicationsinvariousfieldssuchastransportationmanagement,security,andintelligenttransportationsystems.Inthispaper,weproposeacolorrecognitionsystembasedontheSqueezeNetConvolutionalNeuralNetwork(CNN)forvehiclecolorrecognition.Theproposedsystemwasevaluatedonalargedatasetofvehicleimagesandachievedhighaccuracyinidentifyingthecolorsofvehicles.Theexperimentalresultsshowthattheproposedsystemiseffectiveinrecognizingvehiclecolorsandcanbeusedforreal-worldapplications. Introduction Vehiclecolorrecognitionisanessentialtaskintheareaofcomputervision,mainlyintransportationmanagement,trafficanalysis,andintelligenttransportationsystems.Theabilitytorecognizevehiclecolorscanhelpinmonitoringtrafficflow,detectingtrafficviolations,andidentifyingstolenvehicles.Inrecentyears,deeplearningalgorithmshavegainedtremendousattentioninthefieldofcomputervision,andmanyresearchershaveappliedthesealgorithmstothetaskofvehiclecolorrecognition.Theconvolutionalneuralnetwork(CNN)isadeeplearningalgorithmthathasshownremarkablesuccessinvariouscomputervisiontasks.Inthispaper,weproposeacolorrecognitionsystembasedontheSqueezeNetCNNforidentifyingvehiclecolors. RelatedWork Severalstudieshavebeenconductedonvehiclecolorrecognitionusingdifferentmachinelearningalgorithms.InastudybyZhaoetal.,theyproposedamethodforrecognizingvehiclecolorsusingaSupportVectorMachine(SVM)algorithmwithcolorfeatureextractionbasedonhistogramoforientedgradients(HOG)andLocalBinaryPatterns(LBP).Theexperimentalresultsdemonstratedtheeffectivenessoftheirproposedmethod.InanotherstudybyEzhilarasiandKumar,theyusedaConvolutionalNeuralNetwork(CNN)forvehiclecolorrecognition.Theexperimentalresultsshowedthattheirproposedmethodachievedahighaccuracyrateinidentifyingthecolorsofvehicles.Inourstudy,weusetheSqueezeNetCNNforcolorrecognitionofvehicles,whichisalightweightandhighlyefficientCNNarchitecture. Methodology Ourproposedcolorrecognitionsyst