基于YOLOv8作弊行为检测系统的设计与实现  

Design and Implementation of Cheating Behavior Detection System Based on YOLOv8

作　　者：孙秀娟[1] 方园 孙洪迪[1] 杨民峰 SUN Xiujuan;FANG Yuan;SUN Hongdi;YANG Minfeng(School of Information Engineering,Bejjing Polytechnie College,Beijing 100042,China)

机构地区：[1]北京工业职业技术学院信息工程学院,北京100042

出　　处：《北京工业职业技术学院学报》2024年第4期17-22,共6页Journal of Beijing Polytechnic College

基　　金：2024年北京工业职业技术学院科研课题(BGY2024KY-22);全国工信行指委—中国电子学会电子信息职业教育2024年度教学研究科研课题(DZ24118)。

摘　　要：在考试监控领域,深度学习技术的应用对于提高作弊检测的准确性和效率至关重要。结合YOLOv8算法和AlphaPose姿态识别模型,构建一个既能快速检测异常行为又能精准判断作弊行为的考试作弊行为检测系统。通过改进数据结构和优化网络结构,仅使用身体数据进行模型训练,使得模型更加专注检测与作弊行为相关的身体动作,从而提高系统的鲁棒性和适用性。实验结果表明:系统对作弊检测的实际准确率在95%左右,能够有效区分正常行为和作弊行为,为考试监管提供了强有力的技术支持。In the field of exam monitoring,the application of deep learning technology is crucial to improve the accuracy and efficiency of cheating detection.Combining YOLOv8 algorithm and AlphaPose pose recognition model,a cheating behavior detection system which can quickly detect abnormal behavior and accurately judge cheating behavior is constructed.By improving the data structure and optimizing the network structure,only using the body data for model training,the model is more focused on detecting the body movements related to cheating behavior,thus improving the robustness and applicability of the system.The experimental results show that the actual accuracy of cheating detection by the system is around 95%,which can effectively distinguish between normal behavior and cheating behavior,providing strong technical support for exam supervision.

关 键 词：YOLOv8算法 目标检测 AlphaPose姿态识别 作弊行为检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]