基于大数据标签与机器学习的车牌识别算法  

License Plate Recognition Algorithm Based on Big Data Label and Machine Learning

作　　者：林庆新[1] Lin Qingxin(Zhicheng College, Fuzhou University, Fuzhou, Fujian 350002, China)

机构地区：[1]福州大学至诚学院

出　　处：《黑龙江工业学院学报（综合版）》2019年第9期49-54,共6页Journal of Heilongjiang University of Technology(Comprehensive Edition)

基　　金：福建省中青年教师科研项目(编号:JAT170853);福州大学至诚学院本科教学工程项目“信息服务综合平台建设”(编号:Zj1715)

摘　　要：为了提高机器学习算法对实际场景中车牌识别能力,克服天气、车牌种类对识别的干扰,以实际场景拍摄的车牌为识别对象,设计并开发了一种基于大数据标签与机器学习的车牌识别系统。首先,在不同启用场景、不同天气下采集两万帧车牌,涵盖各省市汽车,作为大数据标签和学习基础,建立车牌图像和车牌文字字段的训练集数据库;然后,利用图像降噪、自动阈值分割、形态学处理以及边缘检测,完成车牌定位与分割;随后,通过建立支持向量机与随机森林分类器,对训练集样本进行模型学习,完成机器学习算子,完成针对车牌的最终识别;最后基于Windows系统、C++语言与OpenCV开源学习框架,编码实现算法和系统。经过与对照组的对标测试,所提算法具有更高的车牌识别能力。In order to improve the recognition ability of machine learning algorithm for license plate in real scene and overcome the interference of various weather and license plate types, a license plate recognition system based on big data tag and machine learning is designed and developed. Firstly, 20,000 license plates are collected in different scenarios and weather, covering all provinces and cities. As the basis of big data label and learning, the training set database of license plate image and license plate text field is established. Then, image denoising, automatic threshold segmentation, morphological processing and edge detection are used to complete license plate location and segmentation. Subsequently, through the establishment of support vector machine and random forest classifier, the training set samples are model-learned, the machine learning operator is completed, and the final license plate recognition is completed. Finally, based on Windows system, C ++ language and OpenCV open source learning framework, coding algorithm and system are implemented. The algorithm has higher license plate recognition ability after the benchmarking test with the control group.

关 键 词：大数据标签 随机森林 支持向量机 机器学习 车牌识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]