车牌识别系统设计与实现

发布时间：2018-06-22 20:11

本文选题：车牌识别 + 卡尔曼滤波　；参考：《苏州大学》2009年硕士论文

【摘要】： 车牌自动识别系统(License Plate Recognition System)是智能交通系统中的重要组成部分,它有着广泛用途和良好的应用前景。目前已有不少研究者投身这一研究领域,并作了一些富有成效的工作。本文研究了车牌识别系统的四项关键技术及其相应的实现方法:车牌图像预处理技术、车牌图像定位技术、车牌字符提取技术、车牌字符识别技术。在车牌图像预处理技术中,针对阴雨天情况车牌的处理,首先采用分段灰度线性拉伸与卡尔曼滤波相结合的车牌图像的去噪方法,实验结果表明,该方法可以有效地滤除图像的白噪声。在车牌图像预处理基础上对车牌定位和分割,本文采用基于数学形态学和连通域标记的车牌图像定位方法。首先采用9×1的纵向结构元素对车牌图像进行腐蚀,去除一些噪声点,在此基础上,根据车牌字符的长宽比,采用19×17的结构元素对图像进行闭运算,使得车牌所在的连通域与其他可能与之粘连的相对独立的连通域分开,并基于子图像进行连通域判断的字符提取。采用改进的BP人工神经网络的字符识别方法来设计车牌识别系统。BP网络结构为输入层128个结点,输出层结点有6个,隐含层18个结点。在这个网络结构上字母和数字的识别精度分别为91.85%、96%。本文在Visual C++6.0开发平台上,对以上方法进行了开发和实现,构建了一个车牌自动识别系统。最后,本文对车牌自动识别系统的进一步发展方向提出了自己的看法。
[Abstract]:License Plate recognition system (Licensing Plate recognition system) is an important part of Intelligent Transportation system (its). It has a wide range of applications and good application prospects. At present, many researchers have devoted themselves to this research field and done some fruitful work. This paper studies four key technologies of license plate recognition system and their corresponding implementation methods: license plate image preprocessing technology, license plate image location technology, license plate character extraction technology, license plate character recognition technology. In the pre-processing technology of license plate image, aiming at the processing of license plate in rainy and cloudy weather, the method of de-noising the license plate image based on segmented gray linear stretch and Kalman filter is adopted. The experimental results show that, This method can effectively filter the white noise of the image. Based on the preprocessing of license plate image, this paper adopts the method of license plate image location based on mathematical morphology and connected domain marking. First, 9 脳 1 longitudinal structural elements are used to corrode the license plate image and some noise points are removed. Based on this, according to the aspect ratio of the license plate characters, the 19 脳 17 structure element is used to close the image. The connected domain in which the license plate is located is separated from other relatively independent connected domains with which the license plate is located, and the characters of the connected domain judgement are extracted based on the sub-image. The improved character recognition method of BP artificial neural network is used to design the vehicle license plate recognition system. The BP network structure is 128 nodes in the input layer, 6 nodes in the output layer and 18 nodes in the hidden layer. The recognition accuracy of letters and numbers on this network structure is 91.85 / 96, respectively. In this paper, the above methods are developed and implemented on the platform of Visual C 6.0, and a license plate automatic recognition system is constructed. Finally, this paper puts forward my own views on the further development of license plate automatic recognition system.
【学位授予单位】：苏州大学
【学位级别】：硕士
【学位授予年份】：2009
【分类号】：TP391.41

【引证文献】