【作者】 李云华；

【导师】 李宪民； 肖梅；

【作者基本信息】 长安大学 ， 交通运输规划与管理， 2010， 硕士

【摘要】 基于图像处理和模式识别技术的牌照识别系统(license plate recognition system,LPRS)是智能交通管理中的重要研究课题之一,它利用计算机图像处理、模式识别和人工智能技术,对视频交通图像进行处理、分析和识别,从而提取车牌信息,为交通管理、收费、调度、统计提供依据,它的应用非常广泛。概括起来说,牌照识别系统可以有以下几方面的应用：①车辆收费管理；②出行时间测量；③公共停车场安全防盗管理；④机场、港口等出入口车辆管理；⑤道口检查站车辆监控；⑥小区车辆管理；⑦闯红灯等违章车辆监控；⑧交通流量检测；⑨交通控制与诱导；⑩被盗车辆及特种车辆的鉴别。本文在分析了近年来一些典型识别算法的基础上,最终提出了一有效的算法对车牌进行识别。车牌自动识别系统分为车牌定位、车牌字符分割和字符识别三个部分。在车牌定位方面,首先,介绍了车牌定位前的预处理技术,包括图像的灰度化、二值化、图像的边缘检测和滤波处理,这些处理可以提高图像质量,突出车牌信息,有利于车牌的定位。接着介绍了几种常用的车牌定位方法,对这几种方法进行分析,最后本文提出一种改进的投影法对车牌进行定位。在字符分割方面,介绍了车牌的二值化、几何校正等各种算法,然后分析了目前常用的车牌字符切分方法,最后提出了基于垂直投影法的字符切分方法,对车牌字符进行两次切分,达到了很好的效果。在字符识别方面,介绍了目前常用的字符识别方法,在研究了基于BP神经网络的字符识别方法的基础上,对其进行了改进,提出了一种改进的BP神经网络的字符识别方法。实验证明,本文提出的方法是有效的,具有较强的理论指导意义和实用价值。更多还原

【Abstract】 The license plate recognition system based on image processing and pattern recognition is the intelligent traffic management in one of the important research topics. It uses computer image processing, pattern recognition and artificial intelligence technology to process,analyze and identify the video traffic image.and then to extract license plate information. It provides the basis for traffic management, fees, scheduling, statistics. It is widely used. In a nutshell, license plate recognition system can look at several aspects of the application:①vehicle charging management;②ravel time measurement;③public car park security management;④airports,ports and other import and export traffic management;⑤crossing checkpoints vehicle monitoring;⑤community transport management;⑦running red lights and other vehicle violation monitoring;⑧the detection of traffic flow;⑨traffic control and guidance;⑩the identification of stolen vehicles and special vehicles.This paper studies the license plate recognition and analyzes the key technologies of the typical recognition algorithm in recent years,finally a series of effective algorithm is found to determine the identification plate. License plate recognition system is divided into plate location, license plate character segmentation and character recognition of three parts.Its research involves digital image processing, pattern recognition, computer vision, artificial intelligence and artificial neural networks and many other subject areas. In the plate positioning,this paper introduces the pretreatment before the license plate location,including graying, binarization, edge detection and filtering processing of images. Such treatment can improve the image quality and highlight the license plate information.All of these are conducive to the positioning plate. Then this paper introduces some commonly used positioning method and analyzes them. Finally, the improved positioning projection of the plate is used. In the character segmentation areas, this paper introduces the binary license plates, geometric correction algorithms and analyzes the current commonly used methods of character segmentation of license plate.Finally the vertical projection of the character segmentation method is used. Achieved very good results. In the character recognition area,the paper introduced the normal methods of character recognition.Based on BP neural network character recognition method, a modified BP neural network character recognition is used. Experimental results show, the proposed method is reasonable and has strong theoretical guidance and practical value.更多还原