【作者】 陆铮刚；

【导师】 戚飞虎； 朱伟民；

【作者基本信息】 上海交通大学 ， 计算机技术， 2007， 硕士

【摘要】 作为现代智能交通系统中的一项非常重要的技术,汽车牌照自动识别技术是近年来的研究热点。智能交通系统在车辆跟踪与提取、高速公路自动收费、停车场自动计费以及交通流量统计等方面能发挥重要的作用,而牌照是作为交通系统个体的车辆的唯一标志,因此车牌自动识别技术在整个系统中处于核心的关键地位,具有实际而重要的研究意义。国内外的研究人员已经和正在就车牌自动识别技术展开广泛而深入地研究,提出了许多算法和方案,并有一些产品投入使用,其效果都未能达到人们所期望的水平,离真正实用的要求还有一定的差距。因此在提高车牌提取及识别算法的准确率,健壮性和实时性方面,还存在较大的研究空间。本文在充分研究前人的研究成果基础上,结合新兴的小波分析技术完成了车牌识别系统的方法研究和设计工作,主要工作有以下几个部分:(1)车牌提取(2)字符分割(3)字符识别。在车牌提取方面,本文针对各种复杂背景的车牌图像,对车牌候选区域图像进行小波分析从而准确地提取出车牌位置。经过实验证明,该算法达到了预期的效果,从原始图像中提取出车牌的准确度超过98%。在字符分割方面,本文先采用区域生长算法得到个别字符的位置,然后再利用车牌图像的竖直投影和先验知识来修正字符的位置,最后用得到字符的位置进行字符分割。该算法的准确率和抗干扰性要比传统的投影算法要好得多。在字符识别方面,本文采用发展非常成熟的BP神经网络来完成字符分类工作,但在某些细节上对BP神经网络进行优化,使本系统的的BP神经网络收敛性好,训练速度快。本文提到的所有算法皆已正确实现并仿真成功,使用从各种不同环境中采集了多幅含有车牌的图像作为数据源,数据源具有相当的代表性。经过测试,系统的整体识别率达到了92.7%。所有这些都表明本文的算法识别率较高,速度较快,并且具有相当的鲁棒性。更多还原

【Abstract】 As one of the very important techniques in modern intelligent transportation system, the automatic vehicle license plate recognizing technique is the hot spot of recent years’ research. Intelligent transportation system plays an important role in vehicle tracking and locating、high way toll collecting、automatic parking fee charging and transportation statistic. Because license plate is the only identity of each vehicle, the technique of automatic license plate recognizing stands at a key position in the modern transportation system. Researchers at home and overseas have done extensive works on license plate recognizing, and proposed many algorithms and proposals. Many products were put into practical use, but they never achieved the performance as people had expected. There are many works need to be done to improve the accuracy of plate locating and recognizing, robustness and real-time ability.On the basis of other people’s research, this paper accomplished research and design of vehicle LPR system. This paper’s theory foundation is the new theory——wavelet analysis. The system consists of three main modules: (1) Vehicle license plate extraction; (2) Character segmentation; (3) Character recognition.In plate extraction module, this paper employs wavelet analysis to extract the number plate from the complicated background image. The algorithm locates the plate correctly based on wavelet analysis. According to the experiment, the algorithm achieves the goal and the rate of plate location of this system is over 98%.In the character segmentation module, area growth algorithm is adopted, and plate image vertical projection is combined to segment the character. The algorithm is much better than the traditional methods based on projection. In the character recognition module, the traditional character recognition method based on neural network is adopted in this paper. But the BP neural network has been modified a little and the performance of the BP neural network is improved.更多还原