摘要
为实现答题卡精准计数,设计了基于机器视觉的答题卡计数系统,对该系统所采用的卷纸机构和纸张计数算法进行研究。首先,设计了机械卷纸系统,增大了答题卡边缘条纹间距。其次,基于多个感兴趣区域的多次测量数量,设计了对其进行概率统计的计数算法。实验结果表明,通过卷纸机构卷折,答题卡边缘图像中黑白条纹间距可由非卷折状态下的2个像素,增大为卷折后的5个像素;结合基于概率统计的计数算法,系统对28~31张答题卡的测量准确率可达100%。
In order to count the answer sheets accurately,a paper counting system based on machine vision is designed,in which the rolling mechanism and the counting algorithm are studied.Firstly,a rolling mechanism for answer sheets is developed to increase the spacing between the black and white stripes formed by the edges of the answer sheets.Secondly,a counting algorithm based on the statistical analysis for multiple counts calculated from multi-region of interests(ROI)is proposed.The experimental results show that the rolling mechanism can broaden the black and white stripes from 2 pixels to 5 pixels.The counting algorithm achieved the accuracy of 100%with measuring the answer sheets of 28 to 31.
作者
叶慧雯
郭新年
季陈
杨玉东
Ye Huiwen;Guo Xinnian;Ji Chen;Yang Yudong(College of Computer Science and Technology,Nanjing University of Technology,Nanjing 211816,China;Faculty of Electronic Information Engineering,Huaiyin Institute of Technology,Huaian 223003,China)
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2020年第5期608-613,共6页
Journal of Nanjing University of Science and Technology
基金
江苏省高校自然科学基金(18KJB140001)。
关键词
机器视觉
答题卡计数
机械卷折
纹理
概率
machine vision
answer sheet counting
rolling mechanism
texture
probability
