基于机器学习的车牌识别技术研究  

Research on License Plate Recognition Technology Based on Machine Learning

作　　者：陈政 李良荣 李震 顾平 CHEN Zheng;LI Liang-rong;LI Zhen;GU Ping(School of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China)

机构地区：[1]贵州大学大数据与信息工程学院,贵州贵阳550025

出　　处：《计算机技术与发展》2020年第6期13-18,共6页Computer Technology and Development

基　　金：国家自然科学基金(61361012);贵州省科技计划项目(黔科合平台人才[2017]5788号)。

摘　　要：随着经济的快速发展和城市扩张,交通量逐年增加,交通管理也变得复杂多样。针对隧道环境下高速行驶车辆的车牌识别问题,提出了一种车牌分割和识别的算法。算法分为四个部分:图像预处理,车牌定位,车牌分割和字符识别。采用选择更新法拦截行车辆视频进行关键帧处理;在车牌定位中选用边缘检测与形态学相结合的算法,以消除噪声干扰,提高定位准确率;又用阈值分割法进行字符分割,以解决投影分割法等传统算法出现的字符黏贴和汉字不连通等问题;再通过HOG算法对分割后的字符图像进行特征提取,基于SVM算法实现字符识别。针对训练模型,则采用PSO算法对SVM分类器的参数设置进行优化,以获得最佳分类精度。利用MATLAB平台对优化后的SVM算法进行检验,通过实验数据说明该方法能够提高字符识别的准确率。With the rapid development of economy and urban expansion,the traffic volume has increased year by year,and the traffic management has become complicated and diverse.Aiming at the problem of license plate recognition of high-speed vehicles in tunnel environment,we propose a license plate segmentation and recognition algorithm which is divided into four parts:image preprocessing,license plate location,license plate segmentation and character recognition.The selection update method is used to intercept the vehicle video for key frame processing.In the license plate location,the algorithm combining edge detection and morphology is used to eliminate noise interference and improve the positioning accuracy.In addition,the threshold segmentation method is used for character segmentation to solve the problem of character sticking and Chinese characters disconnection in traditional algorithms such as projection segmentation method.Then the feature extraction of segmented character images is conducted by HOG algorithm,and character recognition is realized by SVM algorithm.For the training model,the PSO algorithm is used to optimize the parameter settings of the SVM classifier to obtain the best classification accuracy.The optimized SVM algorithm is tested by MATLAB platform.The experimental data shows that the proposed method can improve the accuracy of character recognition.

关 键 词：隧道 机器学习 智能交通 车牌识别 SVM MATLAB 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]