基于知识蒸馏的考场异常行为识别  

ABNORMAL BEHAVIOR RECOGNITION IN EXAMINATION ROOM BASED ON KNOWLEDGE DISTILLATION

作　　者：姚捃[1] 郭志林 Yao Jun;Guo Zhilin(The Engineering&Technical College of Chengdu University of Technology,Leshan 614000,Sichuan,China)

机构地区：[1]成都理工大学工程技术学院,四川乐山614000

出　　处：《计算机应用与软件》2025年第3期156-161,共6页Computer Applications and Software

基　　金：教育部产学合作协同育人项目(201802047024);成都理工大学工程技术学院青年科学基金项目(C122019008)。

摘　　要：在实际监控的边缘设备中使用GAN或者3DCNN等网络很难实现实时的、相对准确的监控任务。提出一种基于知识蒸馏的考场异常行为识别算法。相对于以提取空间、时序特征并进行融合为主流思想的异常行为识别,利用视频帧进行目标检测和知识蒸馏的异常行为识别方法更加快速准确。算法借助知识蒸馏策略使用预训练的teacher网络监督student网络学习,进行正常行为的推理并检测异常行为。结果表明该算法达到了主流数据集的中上水平,并在考场环境具有良好的高效性与准确性。It is difficult to achieve real-time and relatively accurate monitoring tasks by using GAN or 3DCNN in the actual monitoring edge devices.An abnormal behavior recognition algorithm based on knowledge distillation in examination room is proposed.Compared with the abnormal behavior recognition which takes extracting spatial and temporal features and fuses as the mainstream idea,the abnormal behavior recognition method using video frames for target detection and knowledge distillation is faster and more accurate.With the help of knowledge distillation strategy,the algorithm used pre-trained teacher network to supervise student network learning,infer normal behavior and detect abnormal behavior.The results show that the algorithm reaches the upper level of the mainstream data set,and has good efficiency and accuracy in the examination room environment.

关 键 词：目标检测 注意力机制 知识蒸馏 异常行为识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]