139-基于DSP的车牌识别系统

整理文档很辛苦,赏杯茶钱您下走!

免费阅读已结束,点击下载阅读编辑剩下 ...

阅读已结束,您可以下载文档离线阅读编辑

资源描述

基于DSP的车牌识别系统目录一、内容简介····································································3二、选题··········································································4三、系统主要特点······························································4四、系统方案、实现原理·····················································44.1图像采集及存储格式······························································64.2图像的灰度化········································································64.3图像的二值化········································································64.4车牌的去边框········································································64.5图像的梯度锐化·····································································74.6字符分割算法········································································74.7字符的归一化········································································84.8字符特征提取·········································································84.9字符识别算法········································································94.9.1BP神经网络法·································································94.9.2模板匹配法···································································10五、操作说明与硬件框图····················································11六、实验程序··································································12七、数据统计分析····························································322八、结果分析··································································32九、实验心得··································································32十、参考文献··································································343基于DSP的车牌识别系统一、内容简介摘要伴随着世界各国汽车数量急剧增加,城市交通状况日益引起人们的重视,如何有效地进行交通管理,已成为越来越多人关注的焦点,解决这些问题的关键就是建立智能交通系统。车牌识别是智能交通系统的重要组成部分,它在交通控制与监视中有着多种用途,目前已广泛应用于各种领域。本文将TMS320C54XX作为核心器件用于车牌自动识别系统中,完成车牌图像的采集、数字图像的处理、提取车牌信息并针对提取的特征对字符进行识别。首先分析了车牌识别系统实现的背景以及其实现意义。然后对实现车牌识别的硬件环境作简要介绍。接着对车牌识别过程中所涉及的边缘检测、字符分割、大小归一化等一系列数字图象处理技术进行进一步的详细分析。之后,对几种字符特征提取算法进行了对比分析,最后选取最适合的网格特征提取法,以此为基础进行模扳匹配,最终识别出车牌号码。关键词:车牌提取;图象处理;车牌识别;DSP;模扳匹配AbstractWiththeincrementofvehicleallovertheworld,thesituationofcitytraffichasattractedtheattentionofpeople.Howtocontrolthetrafficeffectivelyhasbecometheproblemwhichmoreandmorepeoplepaycloseattentionto.ThewaytosolvethisproblemistoestablishtheITS—IntelligentTransportationSystem.VehiclelicenseplaterecognitionsystemisthecrucialpartoftheITS.Itiswidelyusedinvehiclemonitoringandtrafficcontrol.ThisexperimentattemptstousethenewgenerationDSP—DigitalSignalProcessortoimplementtherecognitiontask.TheDSPchipTMS320C54XXisusedtoprocessthepictureofthevehicle,distillinformationofthelicenseplateandrecognizethevehiclelicenseplate.Thefirstpartofthethesisisaboutthebackgroundandmeaningofthevehiclelicenseplaterecognitionsystem.Consequentlyweanalyzetheenvironmentoftheexperiment.Inthenextpartweanalysisthefundamentaltheoryandtechniqueoftheimageprocessing,includingthecollectionofpictureofvehicle,distillofthelicenseplate,segmentationalgorithmofcharacter.Thenweputforwardseveralmethodstodistillthefeatureofthecharacters.Onthebasisoftheabove-mentionedresearch,wemakethetemplatematchingandrecognizethecharacters.Keywords:VehicleLicensePlateLocation;ImageProcessing;CharacterRecognition;DSP;templatematching4二、选题改革开放以来,我国的交通运输业迅速发展。但伴随着其发展,也出现了一系列问题,如交通堵塞、交通事故和环境污染等等。虽然可以靠建设更多的道路设施来满足交通运输增长的需求,但在资源、环境矛盾越来越突出的今天,道路设施的增长将受到限制,这就需要依靠提供除了设施以外的技术方法来满足这一要求。交通的迅速发展使得全世界的研究者不断采用先进的电子和计算机视觉技术来检视超速车辆、掌握车辆行驶,或者用于收费站,以提高车辆的通过速度等等。智能运输系统ITS(IntelligentTransportationSystems)是解决这一矛盾的途径之一,而车牌识别LPR(LicensePlateRecognition)又是ITS中的关键技术之一。传统的车牌识别系统,必须依赖于PC,识别的关键算法全部在计算机中实现,计算机因为其有较高的处理速度和较大的内存,而传统的识别和预处理算法又需要大量的存储空间和较快的cpu处理速度,但是由计算机作为识别主体的系统成本高,而且体积庞大,不易于批量安装和随身携带,当今嵌入式系统的发展日新月易,嵌入式处理器(DSP)的速度不断提高,在某些方面已经超过了传统的PC,而且最DSP主要的优点就是不依赖于任何操作系统,具有高稳定性,成本低,体积小。正因为以上优点,本课题选择了采用DSP来实现车牌识别。三、系统主要特点车牌识别系统能将输入的车牌图像经过处理识别,输出为几个字节大小的车牌字符串,无论在存储空间的占用还是与管理数据库相连方面都有无可比拟的优越性,有着广泛的应用前景。车牌识别系统的成功开发将大大加速智能运输系统的进程。系统采用DSP-EXP-IV实验箱,静态视频图象采集卡,摄像头,该系统的工作过程如下:第一步,由摄像头拍下车牌图像,完成图像采集并将采集来的图像存在视频采集卡的SRAM中。第二步,将图像通过DSP1进行处理,图像处理包括滤波、二值化、倾斜度调整和去除车牌边框、字符分割以及归一化等等。第三步,提取字符的特征,并按照模板匹配法进行识别。第四步,进行双机通信,将分析数据传到DSP2,然后控制液晶屏显示结果。四、系统方案、实现原理汽车牌照自动识别系统工作原理车辆牌照识别系统一般可分为车辆图像获取、车辆牌照子图像区域定位分割、车牌照内字符切分、字符识别(OCR)四大部分:在第一部分主要通过摄像头与计算机的视频捕捉卡直接相连接来完成图像采集,可以实时在线监控图像,重点抓取到含有牌照的图像;该部分功能可简单调用计算机视频捕捉卡厂商提供的各种软件开发工具即可实现。由于车辆牌照自动识别系统前端的车辆位置检测、图像捕捉部分都有了相应的软硬件5刹良好的实现了,所以牌照自动识别系统的关键在于第二、三、四部分,而后续的几部分是串行化工作的,即没有前一步正确快速的识别、切分出相对较清晰的字符,其后的识别阶段根本就不能够进行,所以从这个意义上来讲,牌照子区域定位自动识别切分技术是整个系统的重中之重,它处理工作进行的好坏直接影响后续工作,对字符的拒识率和误识率以及识别速度的实时性有很重要的影响;而从算法的相对难易程度上来说,字符具体的识别部分要大的多,牌照子区域内的字符的切分相对最为容易些。本课题采用DSP作为核心处理器来完成识别过程的算法。总体方案图1系统框图图像预处理的必要性:由于车牌图像是在室外自然背景下拍摄,其背景往往很复杂,可能包括自然场景中的人、其他车辆、树木、建筑物等,拍摄图像时的光照条件也因拍摄时间、地点、天气等条件的不同而不同,因此,考虑到车辆牌照具有不因外部条件变化而变化的特征,即牌照区域与汽车背景在灰度分布上存在着明显的差异,而且车牌的底色和车牌照字的颜色也形成强烈对比,在一相对小的范围内变化频繁,据这一特征,可以对车牌进行二值化处理,提取边缘特征,定位出车牌。图像的二值化处理必须保留车牌区域的信息,即二值化后车牌字符要与底色有明显的区别。这里的图像的预处理阀值T的选取至关重要,二值化后,车牌的背景大部分被去除,这样处理之后将大大有利于后续车牌区域的搜索定位。图像捕捉图像预处理去除牌照边框预处理后字符切分字符识别输出识别字符牌照子图像定位64.1图像采集及存储格式用摄像头对准车牌,按下视频采集卡上的复位键,图像便被存储在采集卡的SRAM空间。存在SRAM中的图像在电脑显示器上显示时,有27

1 / 34
下载文档,编辑使用

©2015-2020 m.777doc.com 三七文档.

备案号:鲁ICP备2024069028号-1 客服联系 QQ:2149211541

×
保存成功