欢迎访问《图学学报》 分享到:

图学学报

• 图像与视频处理 • 上一篇    下一篇

一种答题卡客观题识别算法

  

  1. (1. 北京林业大学信息学院,北京 100083; 2. 北京市虚拟仿真与可视化工程技术研究中心,北京 100871; 3. 北京大学信息科学技术学院,北京 100871; 4. 沈阳理工大学艺术设计学院,辽宁 沈阳 110159)
  • 出版日期:2019-10-31 发布日期:2019-11-06
  • 基金资助:
    国家重点研发计划项目(2017YFB1002705);中央高校基本科研业务费专项基金(2015ZCQ-XX);北京林业大学“北京市大学生科学研究与 创业行动计划”项目(S201710022068)

A Novel Recognition Algorithm of Objective Questions  for Exam Answer Sheets

  1. (1. School of Information Science and Technology, Beijing Forest University, Beijing 100083, China;  2. Beijing Virtual Simulation and Visualization Engineering Center, Beijing 100871, China;  3. School of Electronics and Computer Science, Peking University, Beijing 100871, China;  4. Shenyang Ligong University, School of Art and Design, Shenyang Liaoning 110159, China)
  • Online:2019-10-31 Published:2019-11-06

摘要: 在保证阅卷质量的前提下,网上阅卷系统不仅极大地减少了教师的工作量,而且 降低了对试卷纸张质量的要求,节约能源。但是,网上阅卷系统中的客观题识别效果对答题卡 图像质量和排版有很强的依赖。为此,提出一种鲁棒的客观题识别算法。首先,考虑到用户填 涂时可能偏离填涂区域,或者用户图像和模板图像位置匹配出现的误差,提出了滑动窗口策略 重新定位实际的填涂区域,消除相关的偏差。然后,通过分析各选项的直方图,并引入加权平 均灰度消除单个选项中填涂不均匀的影响。对同一题下的每个选项进行比较,使得识别算法有 很强的局部适应性,克服使用全局识别策略带来的参数选择困难。实验结果表明,该算法兼容 性好,可以适用于不同排版类型的答题卡客观题识别,鲁棒性强,识别精度高,适用于各种扫 描质量和不同填涂质量的答题卡。

关键词: 网上阅卷系统, 客观题识别算法, 滑动窗口, 加权平均灰度

Abstract: Under the premise of ensuring the quality of the marking, the online marking system not only greatly reduces the workload of the teachers, but also lowers the requirement of high quality exam paper and saves energy. However, the results produced by automatic judgment by the online marking system for objective questions heavily depends on the image quality of high-speed scanning and layout of the exam answer sheets. This article proposes a robust recognition algorithm of objective questions for exam answer sheets. Firstly, considering the possibility that the user may deviate from the filling area, or the possible error in the match between the present image and template image, a sliding window strategy is proposed to relocate the practical filling area in order to eliminate the related deviation. Then the histogram of each option is calculated, and a weighted average intensity is introduced to remove the effect of the uneven filling between different options. The comparison between each option for the same question enables the recognition algorithm to have strong local adaptability. At the same time, this strategy overcomes the difficulty of parameter selection caused by the global recognition strategy. The experimental results show that owing to good compatibility, our algorithm is suitable for different typesetting types of exam answer sheets in the recognition of objective questions. In addition, the algorithm is characteristic of high recognition accuracy as well as strong robustness, thus applicable to exam answer sheets of varying scanning quality and different filling quality.

Key words:  online marking system, recognition algorithm of objective questions, sliding window, weighted average intensity