汽车牌照自动识别技术研究

【作者】 周敏；

【导师】 张重雄；

【作者基本信息】 南京理工大学 ， 信号与信息处理， 2008， 硕士

【摘要】 车牌自动识别系统是以汽车牌照为特定目标的专用计算机视觉系统。一个典型车牌识别系统又分为车牌定位、车牌字符分割及字符识别三部分,它的研究主要涉及到了模式识别和人工智能、计算机视觉、数字图像处理、人工神经网络等众多的学科领域。本课题的目的是对输入的汽车图像进行预处理,提取车牌区域,并对车牌字符进行分割,然后对字符逐个进行识别,最后按顺序以文本的形式表示出来,得到车牌号码的识别结果并将其保存在数据库中。在对车牌识别关键技术进行研究的基础上,完成了车牌识别系统软件的设计。车牌定位是车牌识别系统处理的第一步,定位的准确与否直接关系着车牌识别的成败。首先对图像进行预处理,然后采用行扫描方法进行水平定位,再根据投影法确定车牌左右边界,在准确率上基本满足要求。车牌字符的切分效果直接影响到字符识别。在字符切分之前,先对二值化处理后的车牌图像进行线扫描,确定字符纹理区域上、下边界线的水平倾斜角度,据此对车牌进行倾斜校正,并去除残余边框和铆钉。然后通过投影法确定字符左右边界,进行字符切分,并进行归一化处理。车牌字符的识别是整个系统中最重要的一步,也是系统的设计目的所在。根据车牌字符的特点,采用BP神经网络进行分类识别。将归一化处理后的独立字符图像分别送入汉字识别网络、字母识别网络、字母数字识别网络和数字识别网络进行识别,最后把识别结果按原顺序组合,以文本形式输出,并存放到数据库中。实验结果表明,本课题设计的车牌识别软件能比较准确地定位车牌区域、分割车牌字符并进行字符识别,且性能良好。更多还原

【Abstract】 Automatic License Plate Recognition System is a specific computer vision system based on the specific goals of the vehicle license. A typical License Plate Recognition System is composed of License Plate Detection, Plate Character Segmentation and Character Recognition. The LPR system involves numerous discipline domains, such as Pattern Recognition and Artificial Intelligence, Computer Vision, Digital Image Processing, Artificial Neural Network etc.The purpose of the paper is analyzed as follows. The image inputted to the computer is disposed and analyzed in order to recognize the characters on the license plate and complete the character segmentation and recognition. At last the vehicle license plate number is printed in text string according to the order of original images, and sent to the database. Based on the research of key technologies in LPR, a tentative LPR system is set up.Locate the plate is the first step of the LPR system, whose accuracy directly affects the success of the character recognition result. Firstly, image pretreatment is processed. Secondly, row-scanning is adopted as the horizontal location method. Then, vertical projection is adopted to determine the left and right boundary of the plate. The result proved that, it can satisfy the demand in accuracy.The result of character segmentation affects the recognition directly. Before character segmentation, line-scanning method is taken to determine the horizontal gradient of the top and bottom of character grain area. According to the angle, incline-revising is adopted, and at the same time remaining frame and rivets are also wiped off. Then, vertical projection is adopted to determine the left and right boundary of every character and segment the characters. After that, normalization is taken.Character Recognition takes the most important part in the whole system, which is the purpose of the system design. According to the feature of characters on the plate, BP Artificial Neural Network is adopted as a classified recognition method. Single normalized character is delivered to Chinese character recognition sub-network, letter character recognition sub-network, letter and number recognition sub-network, and number recognition sub-network separately. Finally, the results are printed in text strings according to the order of original images, and sent to the database.Experiment results show that, the LPR system discussed in the paper can accurately locate to license plate, segment and recognize characters. And the system’s performance is good.更多还原