【作者】 杜建波；

【导师】 任洪娥；

【作者基本信息】 东北林业大学 ， 计算机应用技术， 2006， 硕士

【摘要】 汽车牌照自动识别技术应用于智能交通系统是我国交通事业的必然发展趋势,对国民经济的快速发展起着十分重要的作用。 车牌自动识别系统主要分为图像预处理、车牌定位、字符分割和字符识别等主要模块,也包括后续应用程序的开发。针对不同的模块,本文研究分析了现有的理论算法,并提出了具有实际应用意义的解决方案。 1.在图像预处理模块,因为人眼对于不同颜色分量的敏感度不同,图像灰度化采用加权平均值法;二值化过程中阈值的选取至关重要,本文采用动态自适应阈值法,效果理想;边缘提取利用了拉普拉斯算子;去噪过程采用的是中值滤波方法; 2.车牌定位模块包括粗定位和细定位,本文通过分析车牌的尺寸、类型、颜色,得到不同的特征向量,即车牌的几何特征、灰度分布特征、投影特征和字符排列特征等,利用这些特征进行车牌定位; 3.在车牌字符分割模块,提出了双向对比垂直投影分割法,该方法基于车牌的垂直投影,能够将字符准确的分割开,利于车牌字符识别: 4.本文对车牌数字和车牌字母及汉字提出了不同的处理方法,数字识别采用投影技术,汉字和字母识别应用CC4神经网络技术,兼顾了识别准确率和识别速度; 5.对于应用程序的开发,数据库设计是个重点,大型的基于卫星和网络的系统适于用Oracle数据库;小区车辆管理等小型系统适于用SQL Server等数据库。 根据上述方法原理,以VC++6.0为开发平台,进行程序设计,编制了车牌自动识别软件。更多还原

【Abstract】 The application of automatic identification technology for license plate in intelligent transport system is the inevitable trend of development, and plays a very important role to the rapid development of the national economy.Automatic identification system consists of image processing module, plate positioning module, characters division module and characters recognition module, including follow-up application development. According to different modules, this paper researches and analyzes the existing theories and algorithms, and puts forward a practical solution scheme of application significance.1.In image processing module, we use weighted average method to gray the image because people have different sensitivity to different colors;the selection of threshold is critical in two- value process, the dynamic adaptive threshold method is used in this paper, and the results are satisfactory;edge extracting use Laplace;noise removing process we adapt the median filtering method;2. Plate positioning module includes rough positioning and careful positioning. In this paper we analyze plate sizes, types and colors to get different characteristics vector such as geometric characteristics, gray level distributional characteristics, projection characteristics and rank characteristics of characters, then find the position of plate with these characteristics;3. In characters division module, the division method of vertical projection in two-way is put forward. This method is based on the vertical projection of the cars plates, and is able to divide the characters accurately, advantageous for characters recognition of cars plates;4. This paper raises different methods to figures recognition and letters as well as characters recognition. Figures are recognized using the projection technology, while letters and characters are recognized by CC4 neural network, giving the same attention to accuracy and speed of recognition;5. For applications development, database design is a focus, large-scale or satellite-based systems apply to use Oracle database;SQL Server database is fit for small system like district management system.Based on the above principles and methods, with the development platform of VC++6.0, program, and produce automatic recognition software of license plate.更多还原