Design And Implementation Of Intelligent Proctoring System Based On CCDADSL-YOL

Nowadays,informatization and digitalization are developing at a high speed,campus teaching and life are also developing towards intelligence.So far,many universities in China have followed the development and changes of the times and started to construct and implement smart classrooms.Smart Classroom is an innovative digital and intelligent auxiliary teaching system,which is an important component of the future in-depth construction of smart campuses.The construction of an intelligent invigilation system is a preliminary task for the construction of smart classrooms and smart campuses.In this regard,this article proposes a method for detecting the behavior of exam takers and a feasible high-precision behavior recognition model CCDADSL-YOLO.This model is improved based on YOLOv5x model structure,which compresses and optimizes the source model structure,and reduces the occupation of memory space and runtime consumption.The model CCDADSL-YOLO on the COCO2017 dataset with mAP₀.5 reached 68.77%,3%year-on-year growth,with an accuracy of 74.57%.After adding SELayer,it reduced the parameter count by 38.3%and the computational load by 0.7%compared to Baseline.This paper also designs and implements an intelligent invigilation system based on CCDADSL-YOLO to identify and analyze the behavior of candidates during exams.The front-end framework of the system uses PyQt5,improving recognition accuracy.The system is divided into three major modules:cheating detection,silent live face registration,and substitute exam detection.Firstly,the real-time video stream processing technology is used to decode the input video stream.Then,the human body and its border are detected by CCDADSL-YOLO algorithm.Then,an improved AlphaPose system algorithm is used for human pose detection.Finally,detect and analyze the human actions in the video,and at the same time warn the abnormal behaviors.The face registration of the system uses silent live detection face recognition technology to establish the face information database of candidates in the examination room,and dynamically compares the face information of candidates in the examination room in real time to detect the phenomenon of substitute examination.The experimental results show that the system can run in real time,accurately identify the examinee’s body movements and head posture,and can run normally in specified scenes,with an accuracy rate of 69.0%for identifying cheating.This system is conducive to promoting school standardization and the construction of digital classrooms,reducing the burden on invigilators while improving the effectiveness of invigilators,and striving to ensure the authenticity and fairness of exams.