大数据时代的车牌汉字识别-上海大学学报.pdfVIP

第 22 卷 第 1 期 （自然科学版） Vol. 22 No. 1 2016年2月 JOURNAL OF SHANGHAI UNIVERSITY (NATURAL SCIENCE) Feb. 2016 DOI: 10.3969/j.issn.1007-2861.2015.04.019 大数据时代的车牌汉字识别 沈文枫, 张建蕾, 周丁倩, 陈圣波, 邱 峰 (上海大学 计算机工程与科学学院, 上海 200444) 摘摘摘要要要: 在大数据 时代, 交通信息成为网络数据量最大 的数据来源之一, 智能交通成为必然需 求. 车牌识别是智能交通 的基础, 可广泛应用于车库管理、交通监控等工程 中, 然而识别 的准 确率还有待加强, 已有算法对于字母 、数字 的识别准确率都非常高, 而对于中国特有 的汉字识 别却效果不佳. 提 出用受限玻尔兹曼机组成的深信度 网络算法来识别车牌字符, 大大提升了汉 字识别的准确率, 使准确率达到 99.44%. 关关关键键键词词词 : 车牌汉字识别; 深信度 网络; 受限玻尔兹曼机; 深度学习 中中中图图图分分分类类类号号号 : TP 391.4 文文文献献献标标标志志志码码码 : A 文文文章章章编编编号号号 : 1007-2861(2016)01-0088-09 Recognition of Chinese characters on license plates based on big data SHEN Wenfeng, ZHANG Jianlei, ZHOU Dingqian, CHEN Shengbo, QIU Feng (School of Computer Engineering and Science, Shanghai University, Shanghai 200444, China) Abstract: Today, traffic provides sources of huge scale data sets on the network, calling for the development of intelligent traffic. The license plate recognition (LPR) techniques are an important basis of intelligent traffic, and widely applied in applications such as garage management and traffic monitoring. However, the current LPR algorithms are imperfect in terms of recognition accuracy. Although working well in recognizing English letters and digits, they are unsatisfactory in recognizing Chinese characters. This paper proposes a license plate recognition algorithm using a deep belief network (DBN) algorithm consisting of restricted Boltzmann machines (RBM). It greatly improves the qualit