基于Django印刷体维吾尔文识别系统的设计与实现  

Design and Implementation of Printed Uyghur Recognition System Based on Django

作　　者：熊黎剑 吾守尔·斯拉木[1,2,3] 许苗苗 XIONG Lijian;WUSHOR Silamu;XU Miaomiao(School of Information Science and Engineering,Xinjiang University,Urumqi 830046,China;Xinjiang Multilingual Information Technology Laboratory,Urumqi 830046,China;Xinjiang Multilingual Information Technology Research Center,Urumqi 830046,China)

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830046 [2]新疆多语种信息技术实验室,新疆乌鲁木齐830046 [3]新疆多语种信息技术研究中心,新疆乌鲁木齐830046

出　　处：《郑州大学学报（理学版）》2021年第3期9-14,共6页Journal of Zhengzhou University:Natural Science Edition

基　　金：国家自然科学基金项目(61433012);国家“973”重点基础研究计划基金项目(2014CB340506)。

摘　　要：光学字符识别(optical character recognition,OCR)技术在图书数字化、文献管理等诸多领域得到了广泛应用,而相比于已十分成熟的中文、英文印刷体识别系统,小文种(维吾尔文)印刷体识别还有研究空间和实际应用需求。针对传统识别方法特征表示不足等问题,结合日益兴起的深度学习技术,采用Python语言编程,选用经改进的卷积循环神经网络作为识别算法核心,并利用Django设计系统框架。实验表明,印刷体维文识别系统的精度达到95.7%,平均速度达到12.5 fps。该系统实现了端到端的维文整词识别。Optical character recognition(OCR)has been widely used in many fields such as book digitization and document management.Compared with the more mature Chinese and English printed recognition system,there is still room for research and practical application of Uyghur printed recognition.Aiming at the problem of insufficient feature representation of traditional recognition methods,the rising deep learning technology was combined,the Python language programming was used,the improved convolutional recurrent neural network as the core of recognition algorithm was selected,and Django was used to design the system framework.The experimental results showed that the accuracy of the system was 95.7%and the average speed was 12.5 fps,which realized the end-to-end Uyghur whole word recognition.

关 键 词：卷积循环神经网络 门控循环单元 连接时序分类器 印刷体维吾尔文 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]