- 无标题文档
查看论文信息

论文中文题名:

 汽车车牌自动检测的图像识别和处理    

姓名:

 田超    

学号:

 G04020    

保密级别:

 公开    

学科代码:

 081001    

学科名称:

 通信与信息系统    

学生类型:

 工程硕士    

学位年度:

 2009    

院系:

 通信与信息工程学院    

专业:

 通信工程    

第一导师姓名:

 李白萍    

第二导师姓名:

 阎鸿森    

论文外文题名:

 Image Recognition and Processing of Vehicle Number Plate auto Detection in Transportation System    

论文中文关键词:

 车牌识别 ; 车型识别 ; 彩色图像分割 ; MPP算法 ; 图像像素    

论文外文关键词:

 Vehicle plate number recognition Vehicle recognition    

论文中文摘要:
图像识别是模式识别的重要内容之一,它的研究方兴未艾,理论体系和技术都在逐步完善中。汽车车牌的图像识别是现代交通管理的重要手段,因此这方面课题的研究,具有理论和实际技术两方面的重要意义。 论文在分析了现阶段汽车车牌识别方法的基础上,认真地研究了识别系统中各环节上,不同方法的优点和存在的问题。车牌识别系统包括:图像增强和图像转换,边缘检测,边界跟踪,阈值分割,目标的表示和描述,车牌定位,彩色图像检测与分割,车牌图像识别技术。贯穿整个论文始终的思想是找到各个环节的最合理的配合,得到更好的识别效果。 根据车牌凸角轮廓特征,采用MPP算法进行车牌倾斜校正;采用彩色图像分割方法,最大限度的去除图像噪声干扰;设计了基于神经网络的多种彩色背景车牌字符识别方法,取得了良好的效果。 基于字符的特征直接与图像像素的多少有着紧密的依赖性,作者从提高图像像素的硬件着手,经过多次反复操作,使得字符识别率得到很大的提高。 根据可限制性标志图像属性,论文提出了基于车牌识别的车牌车型一次性识别的设想。
论文外文摘要:
Image recognition is an important part of model recognition.The theories and techniques of image recognition are stil in step-by-step developments nowadays,The vehicle number plate image recognition is an important way of modern transportation management, which has both theoretical and practical meaning in research. Based on the analysis of current vehicle number plate recognition methods,this paper first discussed the merits and flows of different methods used in different stage of the recognition process in detail.The vehicle number plate recognition system mainly include image enhancement and image transform,edge detection,bound tracing,threshold segmentation,object presentation and description,vehicle plate location,color image detection and segmentation and vehicle number plate image recognition.The main idea of this paper is to find the best cooperation among the different stages of the recognition and get the best recognition effect. Second,based on the salient contour of vehicle plate,the MPP algorithm is ued to rectify the tilting plate.Third,this paper uses color image segmentation methods to erasing the noise interference to the best;Fourth,a multi-color backgrounds vehicle plate number recognition method is proposed using neural network.The recognition effect shows relatively good. The character-based feature has a close relation with the number of pixels in one picture.By enhancing the hardware,the pixel numbers in one picture are enlarged.Thus a good improvement in character recognition is obtained after repeatedly tests. Finally,based on the plate image characteristic of limitable signing,this paper gives an idea on the recognition of vehicles plate numbers and vehicles in one time.
中图分类号:

 TN911.73    

开放日期:

 2010-04-22    

无标题文档

   建议浏览器: 谷歌 火狐 360请用极速模式,双核浏览器请用极速模式