基于人工智能的大篆字体识别系统研究与验证  

Research and verification of large seal style font recognition system based on artificial intelligence

作　　者：李凯 邓杰荣 张鑫 李勇博 习雨璇 李淄博 曹喜信[1] LI Kai;DENG Jierong;ZHANG Xin;LI Yongbo;XI Yuxuan;LI Zibo;CAO Xixin(School of Software&Microelectronics,Peking University,Beijing 100871,China)

机构地区：[1]北京大学软件与微电子学院,北京100871

出　　处：《微纳电子与智能制造》2020年第1期122-126,共5页Micro/nano Electronics and Intelligent Manufacturing

基　　金：华为类脑视觉处理技术(YBN20180825207)项目资助。

摘　　要：大篆是我国西周时期普遍采用的汉字字体,其特点为线条均匀柔和、简练生动,在我国书法文化中具有独特的艺术价值。随着人工智能的飞速发展,字体识别OCR和手写汉字识别HCCR技术得到了广泛发展,但目前大篆字体识别领域依旧处于空白。通过数据集的采集与处理、字符分割、深度学习模型选择与搭建和结果可视化4个步骤,建立了3000常用字的篆体数据库,设计并实现一款基于人工智能的大篆字体识别系统。但因篆体字数据集难以采集全,所以只选出其中常用的60字进行数据扩充和测试。实验结果表明,该系统具有较高的准确性和实时性,对考古、文学作品研究等都有重大意义。甚至有望将人工智能应用到甲骨文识别中,利用算法的决策为甲骨文识别提供帮助。Large seal style has been commonly used in the Western Zhou Dynasty.It has the unique artistic value in Chinese calligraphy culture because of its soft structures and succinct features.Besides the rapid development of artificial intelligence technology,OCR and HCCR technology has been widely developed,but there has a little research about large seal style.Based on these situation,through the four steps of data set collection and production,deep learning model selection and construction,character segmentation,result visualization,a three thousand font database was established,and an artificial intelligence-based font recognition system was designed and implemented.It is difficult to collect the full body font data set,so only 60 commonly used words are selected for data augmentation and test in this article.The experimental results show that the system has high accuracy and real-time,that led to a great significance to the research of Archaeology and literary works.It is even expected that artificial intelligence will be applied to Oracle recognition,and the decision-making of algorithm will provide ideas for Oracle recognition.

关 键 词：人工智能 深度学习 字体识别 字符分割 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]