文献类型：期刊文献

中文题名：基于图像处理和深度学习的题库生成系统

英文题名：A Question Bank Generation System Based on Image Processing and Deep Learning

作者：张瑞[1];张希厚[1];曹嘉玲[1];罗彬杰[1];万永菁[1]

机构：[1]华东理工大学,上海200237

年份：2021

卷号：31

期号：4

起止页码：142

中文期刊名：计算机技术与发展

外文期刊名：Computer Technology and Development

收录：CSTPCD

基金：国家级大学生创新创业训练计划项目(201910251070)。

语种：中文

中文关键词：图像处理;深度学习;残差神经网络;题目切割;智能化题库

外文关键词：image processing;deep learning;ResNet;question segmentation;intelligent question bank

摘要：随着学生的学业压力不断加重,学生在复习时需要消化整理大量做过的试卷,传统的方法是手抄或剪贴题目,耗时耗力。为了帮助学生更高效便捷地整理旧试卷,研究了如何对一张批改后的试卷进行自动处理,同时生成题库的方法。由学生在题号处画圈作为标记,首先通过图像处理技术,在HSV颜色空间获得红色的题号标记和批改痕迹,根据题号对试卷题目和批改痕迹进行分割,得到每道题的批改痕迹和题干,再利用ResNet深度学习网络判断批改痕迹的类型,最后依此将题目分类保存。为了验证算法的有效性,选定小学数学试卷作为测试样本。实验结果表明,该系统对测试样卷有较好的分割效果,能够较为清晰地将题目与批改痕迹进行切割;同时可以相对准确地判断批改类型,实现了高效的分类题库生成。
As the academic pressure of students increase,students need to digest and sort out a large number of done papers when reviewing.Traditional way is to transcribe or cut the papers and paste.In order to help students sort out the written test papers more efficiently and conveniently,we study how to automatically process a corrected paper and generate a question bank.The papers are already marked by students.Through image processing technology,the red marks at the question number and correction marks are separated from the whole papers in HSV color space,and thus the questions and correction marks are automatically divided according to the question number,and ResNet,the deep learning network,is used to judge the type of correction marks,according to which the questions are classified and saved.To testify the efficiency of arithmetic,we test on primary school math papers.Experiment shows that this system has an ideal segmentation effect on the test sample and can cut down the question and correction marks clearly.At the same time,the correction type can be judged relatively accurately,and the efficient classification database generation can be realized.