基于视线估计的考场作弊检测方法  

Cheating Detection Based on Line of Sight Estimation

作　　者：刘雪 陈亚军 马登 彭名杨 LIU Xue;CHEN Yajun;MA Deng;PENG Mingyang(Collegeof Electronic and Information Engineering,West China Normal University,Nanchong 637000,China)

机构地区：[1]西华师范大学电子信息工程学院,四川南充637000

出　　处：《太原师范学院学报（自然科学版）》2022年第3期47-52,96,共7页Journal of Taiyuan Normal University:Natural Science Edition

基　　金：西华师范大学英才基金(463177).

摘　　要：为保证考试的公平性,促进诚信考场建设,解决现有考场异常行为检测方法中对左右偷看作弊动作的低识别率、容易误判等问题.根据考场环境特异性,运用视线估计对考场帧图像中考生头部特征提取,预估注视位置,并在此基础上,融合人脸检测与识别、目标检测,结合考生视线角度与考生到视线目标的距离,对考场帧图像进行作弊检测,将考场图像分为未作弊、可能作弊、已作弊三种行为,在输出结果中将可能作弊行为与已作弊行为连同作弊考生名字一起保存.通过大量实验证明,该方法能够卓有成效地排开相似行为干扰,准确检测出左右偷看作弊行为.In order to ensure the fairness of the examination,promote the construction of the integrity examination room,solve the problems of low recognition rate and easy misjudgment in the existing examination room abnormal behavior detection methods.According to the test environment of the specificity of the uniqueness,cheating in the examination room action,use the line of sight to estimate the examinee in the examination room frame head feature extraction,estimated at position,and on this basis,the fusion of face detection and recognition,target detection,the target line of sight Angle and candidates to the line of sight distance,evaluate the cheating in the examination room frame,The images in the examination room were divided into three kinds of behaviors:non-cheating,possible cheating and already cheating.The possible and already cheating behaviors were saved together with the names of cheating examinees in the output results.A large number of experiments show that this method can effectively eliminate the interference of similar behaviors and accurately detect the cheating behaviors.

关 键 词：考场 注视目标估计 人脸检测与识别 目标检测 作弊检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]