搜索
扬州大学本科生毕业设计基于BP神经网络的车型识别摘要车型分类识别技术作为智能交通系统中的关键技术，对提高道路运输效率，改善车辆收费检测等方面有着重要的理论与现实意义。本文基于视频检测技术，首先通过图像预处理、车辆分割、轮廓提取得到车辆的轮廓图，从中获得车辆的外形几何参数，并做相关性分析，提取特征向量。然后利用提取的特征向量，构建BP神经网络的车型分类系统进行车型识别。主要研究内容包括：(1)车辆检测研究。本文采用一种基于背景差分的车辆分割方法，较好地解决了复杂交通情况下车辆的检测问题。(2)车型特征提取。根据车型分类的需要，分析了车型特征参数的选择问题，为车辆分类奠定了基础。本文最终选取了顶长比、顶高比、前后比作为特征向量。(3)车型分类研究。研究了基于BP神经网络的车型分类，通过选择合适的特征参数，获得了较高的分类正确率。应用效果与仿真结果表明，基于BP网络的车型分类技术的实时性、精确性和分类识别性能等关键指标得到明显的改善，达到系统设计的预期要求。同时,我们采取的方法具有提取的特征简单、量少，并且所构成的具有分类功能的BP网络简单、便于硬件实现、有利于BP网络的分类识别等优点。关键词：智能交通系统；车型识别；车辆检测；特征提取；BP神经网络扬州大学本科生毕业设计VehicleRecognitionBasedonBPNeuralNetworkAbstractAsthekeytechnologyofIntelligentTransportationSystem(ITS),vehiclerecognitionhasallimportanttheoreticalandpracticalsignificanceinimprovingtheefficiencyofroadtransportationandtestingofvehiclecharging．Firstly,thepaperbasedonvideodetectiondiscusseshowtogetthevehiclecontourmapthroughtheseoperationssuchasimagepre-processing,vehiclesegmentationandcontourextractiontoderivegeometricalparametersofvehicleswhichareusedtoestablishthevectorbyacorrelationanalysis．Secondly,weusethesefeaturevectorstobuildthesystemofvehicleclassificationbasedonBPNeuralNetworktorecognizethevehicles．Themaintasksareasfollows:(1)Vehicledetection．Thispaperpresentsthevehiclesegmentationmethodbasedonbackgroundsubtraction．Itcansolvetheproblemofvehicledetectionincomplextrafficsituationseffectively.(2)Featureextraction．Accordingtotheneedsofvehicleclassification，weanalyzetheselectiveproblemsoftheparameterstolaidthefoundationofvehicleclassification．Thispaperfinallyadoptsthevectorsoftheratiooftopandlength,topandheight,forwardandback．(3)Vehicleclassification．ThispaperstudiesvehicleclassificationbasedonBPneuralnetworkandobtainshigherclassificationaccuracybyselectingtheappropriateparameters．Theresultofapplicationandsimulationindicatesthatthereal-timequality,accuracyandotherperformancesimprovedandthevehicleclassificationsystemachievesprospectiveobjectives．Atthesametime,ourapproachhasfollowingadvantages:Theextractedfeaturesaresimple,theaccountissmall,andtheBPnetworkposedbytheclassificationfunctionissimpleandeasytoimplementhardware,whichwillhelpclassificationandrecognition．Keywords:ITS;VehicleRecognition;VehicleDetection;FeatureExtraction;BPNeuralNetwork扬州大学本科生毕业设计目录第1章绪论..............................................................................................................................11.1课题研究的背景和意义..............................................................................................11.2国内外车型识别技术的研究现状..............................................................................21.3论文的主要内容..........................................................................................................3第2章车辆目标检测..............................................................................................................42.1基于视频图像的车型识别系统简介..........................................................................42.2视频图像序列采集......................................................................................................42.3车辆目标检测的常用方法..........................................................................................52.3.1图像预处理........................................................................................................52.3.2背景差分............................................................................................................62.3.3阈值分割............................................................................................................82.3.4形态学处理......................................................................................................102.3.5连通区域标记及区域填充..............................................................................14第3章车辆目标的特征提取................................................................................................163.1目标特征的提取及描述............................................................................................163.2基于轮廓特征的边缘检测........................................................................................163.3基于轮廓特征的选择与提取....................................................................................193.3.1车型特征值的选择...........................................................................................193.3.2车型特征值的提取...........................................................................................20第4章BP网络的设计与车型识别......................................................................................234.1BP神经网络简介.......................................................................................................234.1.1多层前馈神经网络..........................................................................................244.1.2BP网络学习规则............................................................................................264.2BP网络在本实验中的设计与应用...........................................................................274.2.1BP网络的设计.................................................................................................274.2.2车型识别结果...........................................