基于嵌入式AI技术的课堂考勤系统（YOLO和FaceNet） 毕业论文+任务书+开题报告+中期检查表+外文翻译及原文+答辩PPT+参考文献

摘  要

课堂考勤是高校课堂的必要环节，但往往占用师生大量宝贵的课堂时间。传统的人工点名考勤方式效率较低，而自动点名考勤和手机口令考勤等方式存在容易导致假出勤、替出勤的现象。因此，开发一套识别速度快、准确率高、稳定性高的智能课堂考勤系统具有一定现实意义。

从对课堂出勤统计的实际需求出发，以人脸检测与特征识别为基础，论文实现了一套基于嵌入式AI技术的课堂考勤系统。该系统由人脸识别考勤设备与PC端上位机软件两部分组成。首先，利用IR-RGB双光谱摄像头获取出勤者人脸图像信息并送入嵌入式AI处理器K210芯片，经基于YOLO的人脸检测算法与近红外活体检测算法检测到活体人脸后，使用基于FaceNet的人脸识别算法对人脸图像进行特征向量映射，通过特征向量比对获取识别对象的身份信息，从而实现课堂出勤信息统计。利用网络通信处理器ESP32芯片实现出勤信息的实时无线传输。最终，利用PC端上位机实时接收出勤数据，并对出勤统计结果进行数据存储和实时显示。测试结果表明，该系统可准确、快速地完成课堂考勤任务，并能有效地避免假出勤、替出勤等现象。

关键词：课堂考勤系统；嵌入式AI；人脸识别；YOLO；FaceNet

Abstract

As a necessary part of college classroom, classroom attendance often takes up a lot of valuable classroom time of teachers and students. Because the efficiency of traditional manual attendance is low, and automatic attendance and mobile phone password attendance are easy to lead to false attendance and substitute attendance. Therefore, a set of intelligent classroom attendance system with fast recognition speed, high accuracy and high stability has been developed, which has certain practical significance.

Starting from the actual demand of class attendance, this paper implements a set of class attendance system based on embedded AI technology and face detection and feature recognition. The proposed system is composed of face recognition attendance equipment and PC upper computer software. Firstly, the face image information of the participants is obtained by IR-RGB dual-spectral camera and sent to the embedded AI processor K210 chip. After living face being detected by the face detection algorithm based on YOLO and the near infrared living body detection algorithm, the face recognition algorithm based on FACENET is used to implement feature vector mapping on the face image, through which the identity information of the target object is obtained according to comparison. The class attendance information statistics is realized through the steps above. The real-time wireless transmission of attendance information is achieved by using network communication processor ESP32 chip. Finally, the real-time attendance data is received, stored and displayed by the upper computer of the PC terminal. Experimental results show that the proposed system can complete the task of class attendance accurately and quickly, and avoid phenomenon of the pretended and substitute attendance effectively.

Key Words: Class attendance system; Embedded AI; Facial Recognition; You Only Look Once; FaceNet

目  录

第1章  绪论

1.1  研究背景及意义

1.2  国内外研究现状

1.3  本文的主要工作及论文结构

第2章  系统实现方案设计

2.1  引言

2.2  系统功能需求分析

2.3  硬件实现方案设计

2.4  软件实现方案设计

2.5  本章小结

第3章  人脸识别考勤算法设计及实现

3.1  引言

3.2  基于YOLO目标检测算法的人脸检测

3.3  基于近红外活体检测算法的活体判断

3.4  基于FaceNet人脸处理算法的人脸识别

3.5  算法流程及步骤

3.6  本章小结

第4章  系统硬件设计及实现

4.1  引言

4.2  人脸识别核心板

4.3  通信与供电底板

4.4  外观结构设计

4.5  本章小结

第5章  系统软件设计及实现

5.1  引言

5.2  PC端上位机软件

5.3  K210芯片固件

5.4  ESP32芯片固件

5.5  本章小结

第6章  实验测试与数据分析

6.1  引言

6.2  人脸识别考勤设备测试

6.2.1  算法性能测试

6.2.2  硬件性能测试

6.3  系统整体功能测试

6.4  本章小结

第7章  结论

参考文献

致  谢