图像处理耦合模板定位的答题卡识别研究与应用

摘要

The recognition algorithm used in most current machine scoring is based on fuzzy identification,and it is very difficult to accurately deal with this situation of transforming kinds of test papers,or different collection methods the same test papers for certain types of papers.Therefore,this paper proposed a card recognition mechanism of OpenCV coupling template location based on the answer.Firstly,the delimited student ID and the obj ective questions district were obtained based on human computer interaction;then the fill position was obtained by the image processing algorithm to evaluate the filling results for completing answer card identification.This system templates creating module is programmed by C#,the answer card recognition is achieved by C++ and OpenCV.The performance of the mechanism was tested.And the results show that,compared with general method based on fuzzy recognition,this mechanism had better localization effect and high-er recognition accuracy.%当前大多数机器阅卷中采用的识别算法基于模糊识别，即针对某类型的试卷，更换多种试卷或者同种试卷不同采集方式下很难准确对应，具有一定的局限性。对此，本文提出一个基于 OpenCV 耦合模板定位的答题卡识别机制。首先基于人机交互划定学号区与客观题区；然后基于图像处理算法定位得到填涂位置，评价填涂结果，完成答题卡识别。本系统模板制作模块由 C#编程实现，答题卡识别由 C++和OpenCV实现。最后测试本文机制性能，结果表明：与基于模糊识别的普通方法相比，本文机制具有更好的定位效果和识别准确度。