基于CRNN-CTC的智能判题器设计  

Design of Intelligent Scoring System Based on CRNN-CTC

作　　者：黄巧洁[1] 刘思思 黎颖 刘伟俭 HUANG Qiao-jie;LIU Si-si;LI Ying;LIU Wei-jian(Guangdong Agriculture Industry Business Polytechnic,Guangzhou 510507 China;Guangzhou City University of Technology,Guangzhou 510800 China)

机构地区：[1]广东农工商职业技术学院,广东广州510507 [2]广州城市理工学院,广东广州510800

出　　处：《自动化技术与应用》2025年第4期61-65,共5页Techniques of Automation and Applications

基　　金：广东省普通高校重点科研平台和项目(2021ZDZX1121);广州市科技计划项目(202201011736)。

摘　　要：为了有效提升线上辅助教学效率,建立微信小程序判题系统实现随时随地智能判题。基于卷积循环神经网络(convolutional recurrent neural network,CRNN)和联结主义时序分类器(connectionist temporal classification,CTC),设计部署于云服务器的智能判题器,通过调用微信小程序,实现待识别图片判题功能。实验结果表明,该系统能实现十以内加减法的自动判题,准确率达99.5%以上。采用云技术的自动判题系统突破了传统主观题判题模式,能更好地调动学生的学习积极性,也能大大减少教师的重复判题工作量,实现了教与学的双赢发展。In order to effectively improve the efficiency of online assisted teaching,an intelligent scoring system based on WeChat mini pro-gram is established,which can achieve intelligent question answering anytime and anywhere.Based on convolutional recurrent neural network(CRNN for short)and connectionist temporal classification(CTC for short),called CRNN-CTC for short,an intel-ligent question solver deployed on cloud servers is designed,which could achieve the function of identifying images by invoking WeChat mini program.The experimental results show that the system can achieve automatic problem determination using addi-tion and subtraction within ten,with an accuracy rate of over 99.5%.This system adopts cloud technology,breaking through the traditional objective question answering mode,which can better mobilize students'learning enthusiasm and greatly reduce teach-ers'repetitive question answering workload,achieving a double-win development between teaching and learning.

关 键 词：CRNN-CTC 智能判题器 微信小程序 二值化 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]