一种适用于移动设备在线阅卷的答题卡自动识别算法

An automatic recognition algorithm for the online marking of mobile devices

摘要:

基于智能移动设备的阅卷方式无需专用设备即可快速阅卷,降低成本的同时还能增加阅卷工作的可移动性。为解决移动设备拍照阅卷时,出现的阴影、反光、倾斜(-45°~45°)等情况,设计并实现了一套答题卡自动识别算法,算法主要分为三部分：图像预处理、待识别区域定位与分割、答题卡内容识别。不同操作系统(PC、Android、IOS)下测试结果表明：该算法正常填涂采集识别率为100%,异常填涂采集识别率为93.6%；识别速度小于2s,满足实时性要求；无需修改就能在不同操作系统下编译运行,提高了程序的通用性和兼容性。目前该算法已成功应用于某教育企业上线APP中。

Abstract:

The marking method which based on intelligent mobile device can work fast without special equipment, reduce costs and increase the mobility of marking work. To solve the problem when using a mobile device for marking, such as shadows, reflections, tilted -45° to 45° and so on, this article designed and implemented a set of sheet automatic identification algorithm. The algorithm is mainly divided into three parts: image preprocessing, identify regional orientation and segmentation, sheet content recognition. Experimental results show in different operating systems (PC, Android, IOS): the recognition rate of the algorithm is 100%, and the detection rate of abnormal filling is 93.6%; the recognition speed is less than 2s, satisfying the real-time requirement; it can be compiled and run under different operating systems without modification, thus improving the universality and compatibility of the program.At present, this algorithm has been successfully applied to a education enterprise on-line APP.