利用face_recognition库实现视频流中的实时人脸识别

        人脸识别技术在现代社会中扮演着越来越重要的角色，从安全监控到个性化用户体验，它的身影无处不在。在本博客中，我们将探讨如何使用face_recognition库来实现一个视频流中的实时人脸识别。

---

概述

        本文致力于实现人脸检测项目，该项目不仅能够从视频流中实时检测和识别人脸，还能进一步判断人脸是否张嘴。这项额外的功能有助于区分视频中的人脸是真实存在的还是图片、视频剪辑等伪造内容。举例来说，您可以参照本文提供的示例，为系统添加新的动作检测功能，如眨眼和转头等动作。在进行人脸验证时，您可以随机组合一组动作要求用户执行。如果用户未能在规定时间内完成这些动作，系统将判定人脸验证失败。这样的机制增强了验证过程的安全性和可靠性。

        以下内容将详细介绍项目的四个核心部分：构建人脸数据库、实现张嘴检测功能、进行人脸的实时检测与识别，以及将检测结果以可视化的方式呈现。这些环节共同构成了一个全面的人脸验证解决方案，旨在提供高效、准确的识别服务。

准备人脸图片库

        我的人脸图片库的文件结构为一级目录为图片库名称，二级目录为对应的人名，三级目录为人名对应的人脸图片（注意：图片中只能存在该人一张脸，路径建议不要有中文）。图片可以是.jpg，.png格式。

制作人脸数据库的makeFaceEncodingLib.py

   makeFaceEncodingLib.py是一个用于创建人脸编码数据库的脚本。它通过分析一组已知身份的图片，生成每个人脸的特征编码，并将这些编码存储在数据库中，以便后续的识别过程使用。

import os
import face_recognition
import ast
import numpy as np
import json

class FaceEncodingLib:
    def __init__(self,lib_path):
        directory = os.path.dirname(lib_path)
        if not os.path.exists(directory):
            # 如果目录不存在，则创建目录
            os.makedirs(directory)
        if not os.path.exists(lib_path):
            # 如果文件不存在，则创建文件
            open(lib_path, 'a').close()
            # print(f"文件 '{lib_path}' 已创建。")
        self.lib_path = lib_path
        self.face_encodings = {}
    
    def make_face_lib(self, face_folder):
        for person_name in os.listdir(face_folder):
            person_folder = os.path.join(face_folder, person_name)
            encodings = []
            for image_file in os.listdir(person_folder):
                image_path = os.path.join(person_folder, image_file)
                image = face_recognition.load_image_file(image_path)
                encoding = face_recognition.face_encodings(image)[0] #每张图片只有一个人脸
                encodings.append(encoding.tolist())
            self.face_encodings[person_name] = encodings

        with open(self.lib_path, 'w') as file:
            json.dump(self.face_encodings, file, indent=4)
        
    def load_face_lib(self):
        with open(self.lib_path, 'r') as file:
            self.face_encodings = json.load(file)
        # print(f"已加载 '{self.lib_path}'。")

    def add_face_to_lib(self, face_folder):
        self.load_face_lib()
        new_names = []
        for person_name in os.listdir(face_folder):
            if person_name in self.face_encodings:
                continue
            else:
                new_names.append(person_name)
                person_folder = os.path.join(face_folder, person_name)
                encodings = []
                for image_file in os.listdir(person_folder):
                    image_path = os.path.join(person_folder, image_file)
                    image = face_recognition.load_image_file(image_path)
                    encoding = face_recognition.face_encodings(image)[0] #每张图片只有一个人脸
                    encodings.append(encoding.tolist())
                self.face_encodings[person_name] = encodings

        with open(self.lib_path, 'w') as file:
            json.dump(self.face_encodings, file, indent=4)
        if new_names:
            print(f"新增{new_names}数据已写入 '{self.lib_path}'。")

# libpath = "/home/zhoukang/GithubProject/face_reg_master/face_encoding_lib/lib.txt"
# face_folder = "/home/zhoukang/dataset/image/known"
# faceEncodingLib = FaceEncodingLib(libpath)
# faceEncodingLib.make_face_lib(face_folder)

        生成的人脸数据库文件内容如下，不需要提前创建空文件，代码会检查指定路径下是否有文件，如果没有，会自动创建：

        face_encoding_lib.txt

{
    "AoBaMa": [
        [
            -0.04698127135634422,
            0.11568425595760345,
            0.07694509625434875,
            0.012189071625471115,
            0.003552430309355259,
            -0.015561307780444622,
            -0.10225330293178558,
            -0.08539304882287979,
            0.17924216389656067,
            -0.1594679206609726,
            0.2602607309818268,
            0.06947591155767441,
            -0.19648624956607819,
            -0.13786761462688446,
            0.0800618827342987,
            0.11740974336862564,
            -0.18707014620304108,
            -0.09171440452337265,
            -0.10898447036743164,
            -0.07599669694900513,
            -0.030940856784582138,
            0.012973425909876823,
            0.11444678157567978,
            0.055909279733896255,
            -0.12910205125808716,
            -0.4061811566352844,
            -0.04442097991704941,
            -0.14777310192584991,
            -0.0025415634736418724,
            -0.1710074096918106,
            -0.09442616254091263,
            0.03824620321393013,
            -0.15249015390872955,
            -0.057147182524204254,
            -0.020095467567443848,
            -0.0017956988885998726,
            -0.01978248730301857,
            -0.0010614925995469093,
            0.21213343739509583,
            0.042255908250808716,
            -0.14911791682243347,
            0.05194048583507538,
            -0.03325088322162628,
            0.22742372751235962,
            0.26070332527160645,
            0.0851793885231018,
            0.02810218557715416,
            -0.031441718339920044,
            0.15681341290473938,
            -0.2645900845527649,
            0.04644179344177246,
            0.1558387279510498,
            0.07189097255468369,
            0.03015509434044361,
            0.10073961317539215,
            -0.1961713582277298,
            -0.07440462708473206,
            0.06522148102521896,
            -0.08218405395746231,
            0.04435121268033981,
            0.01078705582767725,
            -0.05520990490913391,
            0.007997343316674232,
            0.035012856125831604,
            0.22852802276611328,
            0.11237428337335587,
            -0.1355539858341217,
            -0.02520371600985527,
            0.12039197236299515,
            -0.08181384950876236,
            -0.009902472607791424,
            0.025467559695243835,
            -0.16078172624111176,
            -0.19973687827587128,
            -0.24500666558742523,
            0.046205684542655945,
            0.3267871141433716,
            0.1516401767730713,
            -0.18830375373363495,
            -0.04017651081085205,
            -0.16449512541294098,
            0.027364101260900497,
            0.030816862359642982,
            0.052031975239515305,
            -0.07667862623929977,
            -0.08185460418462753,
            -0.059264302253723145,
            0.029098335653543472,
            0.14400523900985718,
            0.06174206733703613,
            -0.06401927024126053,
            0.20638079941272736,
            -0.03457768261432648,
            0.09527434408664703,
            0.007656199857592583,
            0.006947179790586233,
            -0.1283358782529831,
            -0.04836852476000786,
            -0.16992738842964172,
            -0.008588451892137527,
            -0.03997272253036499,
            -0.0167672298848629,
            -0.016716938465833664,
            0.11703445762395859,
            -0.17535124719142914,
            0.12042904645204544,
            -0.003188052447512746,
            -0.008692312054336071,
            0.01684371568262577,
            0.12957079708576202,
            -0.10339426249265671,
            -0.043619245290756226,
            0.06909753382205963,
            -0.2405875325202942,
            0.2362385243177414,
            0.22492918372154236,
            0.05780321732163429,
            0.16976143419742584,
            0.061278171837329865,
            0.03657356649637222,
            -0.029212625697255135,
            -0.009622236713767052,
            -0.14237217605113983,
            -0.10185068845748901,
            0.044171467423439026,
            0.04137367010116577,
            0.05731340870261192,
            -0.007031507324427366
        ],
        [
            -0.09676194936037064,
            0.12042399495840073,
            0.05106069892644882,
            -0.051098428666591644,
            0.028641145676374435,
            -0.005404854193329811,
            -0.08761100471019745,
            -0.11248961091041565,
            0.1929294466972351,
            -0.08010173588991165,
            0.22121968865394592,
            0.10412289947271347,
            -0.1979617029428482,
            -0.19429588317871094,
            0.056788913905620575,
            0.0877159