【作者】 林男；

【导师】 杨国来；

【作者基本信息】 兰州理工大学 ， 机械电子工程， 2007， 硕士

【摘要】 本论文是《智能交通系统(ITS)中车牌识别方法》研究内容的一部分。车辆牌照自动识别是智能交通系统中的关键技术。作为ITS研究重要组成部分之一的自动车牌识别是一个值得深入研究的课题。本论文针对自动车牌识别中的关键部分——车牌定位作了较为深入的研究，研究探讨了几种车牌定位方法，以期得到定位精度高、计算速度快的车牌定位方法。首先介绍基于扫描行的方法进行车牌定位。此方法对二值化图像作简单的相邻像素相减，得到图像边缘，此时车牌每个字符都出现了外边缘和内边缘，然后用扫描行的方法对图像进行扫描，利用车牌区域字符的跳变个数及跳变距离，我们就可以得到车牌区域。其具体的操作是在Visual C++6.0环境中实现的。接下来介绍利用神经网络对车牌进行定位。神经网经自适应学习能力和容错性强，这是它用于车牌定位的最大特点。该方法先对神经网络进行训练，使它对车牌区域产生敏感，然后用滑动窗口对图像进行搜索，得到车牌区域。然后介绍结合数学形态学和综合颜色特征进行车牌定位。数学形态学对图像的特征提取和形状分析具有较强的功能。用形态学的闭运算和开运算，可以使牌照区域内部连通并与干扰区域分离出来，再根据车牌区域的面积大小进行筛选，最后利用颜色特征选取车牌区域。其具体的操作是在MATLAB 6.5中实现的。最后本论文总结了上述各方法的优缺点，并概要介绍了车牌字符分割和字符识别的方法。更多还原

【Abstract】 This paper is a part of research of methods of vehicle-license-plate recognition system that is one part of Intelligent Transport System. Automatic vehicle-license-plate recognition system that is an important part of Intelligent Transport System, is an attractive subject. This paper pays attention to the key part—vehicle-license-plate location. It does some deep research and attempts several methods. We hope to find some location methods that have high location accuracy and great computing speed by the research.Firstly, the location method of license-plate based on line scanning is researched. After subtracting between two adjacent pixels in a binarization image, the image become an edge image. Then scanning the edge image with the method of line scanning, we will find the license plate area based on the count of leap and the distance between two characters in the license plate area. The method is implemented with Visual C++ 6.0.Secondly, the location method of license-plate based on neural network is researched. The ability of adaptive learning and Fault-tolerance of neural network is very strong, which is the best characteristic for locating the license plate. After training neural network, the neural network is sensitive with the area of license plate. Then search the image with a glide window and fine the license plate area.Thirdly, mathematical morphology and color characteristics on license plate are combined to locate the license-plate. Mathematical morphology is good at extracting the edge image and analyzing the shape of image. After opening operation and closing operation with mathematical morphology, the area of license plate is connected inside and separated from the other areas. Then filtrating some areas bases on the area of license plate, and locating the license plate bases on color characteristics of license plate. The method is implemented with MATLAB 6.5.Lastly, this paper summarizes the advantage and disadvantage of all above location methods. It also introduces some segmentation and recognition methods of characters in brief.更多还原