车牌识别系统关键技术的研究与实现

发布时间：2018-06-12 05:02

本文选题：车牌识别 + 车牌定位　；参考：《西华师范大学》2017年硕士论文

【摘要】：车牌识别系统(License Plate Recognition,LPR)作为智能交通系统的主要组成部分,一直都是国内外学者的研究热点。目前,它已广泛应用于高速公路收费管理系统、小区智能停车管理系统、道路车辆监控系统及“电子眼”系统等多个领域。因此,对其进行研究有着重大的现实意义。本文就LPR中的三大核心技术进行了探讨和研究,所做的主要工作如下:(1)车牌定位。作为整个识别系统中关键的第一步,定位的正确与否将决定后面的字符分割与识别结果。本文通过研究和分析目前常用的一些车牌定位方法,提出了一种结合梯度和饱和度颜色特征的定位算法。此算法主要包括四个步骤:基于梯度特征的粗定位、基于饱和度特征的粗定位、精确定位及去除伪车牌。利用综合特征进行定位不仅可以有效地避免使用单一特征所导致的定位误差,而且比单一特征定位更符合人的视觉要求。(2)字符分割。根据我国车牌的字符分布特点及目前常用的车牌字符分割方法,本文采用了一种结合先验知识与投影的字符分割算法。此方法充分利用了车牌字符的分布规律,根据字符的垂直投影来找出分割点,进而分割出车牌的单个字符。此方法原理虽然简单、容易实现,但实验表明:它可以有效地避免图像中因车牌倾斜、光线不均等因素所造成的影响。(3)字符识别。作为识别系统中的核心步骤,字符识别的速度及准确度将决定系统的性能。本文通过研究前人的字符识别方法,提出了一种基于PSO-SVM的车牌字符识别算法。其主要思想是:首先通过粒子群优化算法(Particle Swarm Optimization,PSO)来对支持向量机(Support Vector Machine,SVM)中核函数参数进行寻优操作;然后根据所得参数设计合适的SVM分类器来对车牌字符进行识别。实验结果表明:与一些其它的参数寻优算法相比,此算法的识别准确率更高。
[Abstract]:License Plate Recognition (LPR), as the main component of intelligent transportation system, has always been the research hotspot of scholars at home and abroad. At present, it has been widely used in expressway toll management system, intelligent parking management system, road vehicle monitoring system and "electronic eye" system and so on. It is of great practical significance to study it. In this paper, the three key technologies in LPR are discussed and studied. The main work is as follows: (1) license plate location. As the first step in the whole identification system, the correct location will determine the following character segmentation and recognition results. This paper is studied and analyzed at present. This algorithm mainly includes four steps: coarse location based on gradient feature, coarse location based on saturation feature, accurate location and removal of pseudo license plate. The location error caused by a feature is more consistent with the human visual requirement than the single feature location. (2) character segmentation. According to the character distribution characteristics of the license plate in China and the current common license plate character segmentation method, a kind of character division algorithm combining prior knowledge and projection is adopted in this paper. This method makes full use of the character of the license plate. According to the vertical projection of the character, the segmentation point is found out and the single character of the license plate is separated. The principle is simple and easy to be realized, but the experiment shows that it can effectively avoid the influence caused by the incline of license plate and the inhomogeneous factors in the image. (3) character recognition. As the core step of the recognition system, character recognition. The speed and accuracy of recognition will determine the performance of the system. In this paper, a character recognition algorithm based on PSO-SVM is proposed by studying the character recognition method of predecessors. The main idea is: first, the kernel function of the support vector machine (Support Vector Machine, SVM) is obtained by the particle swarm optimization (Particle Swarm Optimization, PSO). Several parameters are optimized, and then a suitable SVM classifier is designed to identify the characters of the license plate according to the parameters. The experimental results show that the recognition accuracy of the algorithm is higher than some other parameter optimization algorithms.
【学位授予单位】：西华师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U495;TP391.41

【参考文献】