摘要

人脸识别技术是建立在大数据、人工智能等先进信息技术基础之上的一种生物识别方法，它通过捕捉和分析个体的面部特征数据来实现对人脸的精确识别。近年来，对人脸识别快递柜系统的研发与应用，已经引发了广泛的研究兴趣，并在实际操作中得到了深入的推广和应用。

本文通过分析传统回归树及 ERT 算法的原理，结合 Dlib 库中的人脸检测器和人脸特征点检测器两个预训练模型，设计了一款适用于物流行业的人脸识别快递柜控制系统。其主要功能如下：

（1）基于 Python 编写上位机软件，用于处理人脸图像数据，执行识别算法，并与嵌入式控制模块进行通信；

（2）采用 ESP8266 作为主控芯片，负责接收上位机软件的人脸识别结果，并根据指令控制快递柜的门禁系统；

（3）采用 PC 端自带的摄像头用于实时采集人脸图像，并将这些图像传输给上位机软件进行人脸识别；

（4）采用 sg90 舵机用于模拟快递柜的门禁动作，根据上位机的指令进行开启或关闭；

（5）采用 LCD 显示模块来实时显示门禁状态，提供给用户直观的反馈。

本设计能够将采集到的人脸信息与事先录入的人脸数据进行比对，以控制柜门进行动作。实验结果表明，本文实现了设计所需功能，达到了预期设计目标，且响应速度快，稳定性高。

关键词：人脸识别；回归树；ERT 算法；Dlib 库

Abstract

Face recognition technology is a biometric identification method based on advanced information technologies such as big data and artificial intelligence. It achieves precise face recognition by capturing and analyzing individual facial feature data. In recent years, the research, development, and application of face recognition express locker systems have garnered widespread interest and have been extensively promoted and applied in practical operations.

By analyzing the principles of traditional regression trees and the Ensemble of Regression Trees(ERT) algorithm, and incorporating two pre-trained models from the Dlib library: the face detector and facial landmark detector, this article has designed a face recognition express locker control system suitable for the logistics industry. Its main functions included the following:

(1) Upper computer software is programmed based on Python, which processes facial image data, executes recognition algorithms, and communicates with the embedded control module.

(2) An ESP8266 is used as the main control chip, responsible for receiving face recognition results from the upper computer software and controlling the access control system of the express locker based on instructions.

(3) The PC's built-in camera is used to collect real-time facial images, which are then transmitted to the upper computer software for face recognition.

(4) An sg90 servo is used to simulate the access control action of the express locker, opening or closing based on instructions from the upper computer.

(5) An LCD display module is utilized to show the access control status in real-time, providing intuitive feedback to users.

This design has been compared the collected facial information with pre-recorded facial data to control the locker door's action. Experimental results have demonstrated that the design has achieved the intended functionality, met the expected design goals, and exhibited fast response time and high stability.

Keywords: face recognition; regression trees; ERT algorithms; Dlib library

摘要

Abstract

目录