如何在网页中使用人脸识别？

迈不过友情╰ 2024-04-26 10:34 138阅读 0赞

#### 文章目录 ####

*   *  项目介绍
     *  源码分析
     *   *  1.用户注册，人脸录入
         *   *  人脸录入前端代码：
             *  人脸录入后端代码：
         *  2.人脸识别登录部分
         *   *  登录界面的前端代码
             *  人脸登录后台代码:
         *  3.文件上传部分
         *   *  图片模糊前端代码
             *  图片模糊后台代码：
         *  4.人脸区域模糊功能

### 项目介绍 ###

本项目主要实现了通过人脸识别的方式进行网页的登录。
    网页采用Django框架开发，使用离线的face_recognition来进行人脸识别。登录完成后网页提供了一个人脸区域模糊的功能。

项目效果视频展示：[人脸识别网页登录][Link 1]  
项目源码：https://github.com/MicahYin/WebFaceLogin  
人脸识别模块：[face\_recognition][face_recognition]  
文件上传模块：[webuploader][]  
开发环境：ubuntu 16.04

> 关注公众号 **打代码的苏比特**，获取**java面试百问百答**和**面试高频算法题**~

### 源码分析 ###

#### 1.用户注册，人脸录入 ####

##### 人脸录入前端代码： #####

录入人脸的主要思路是先在前端进行人脸检测，使用了轻量的前端人脸检测框架，[tracking.js][]。若检测到人脸，通过javascript编写的postFace()函数将人脸图片传到后台，后台进行人脸验证成功后将图片存储。下次登录时，利用face\_recognition进行对比，即可完成人脸登录

##### 人脸录入后端代码： #####

先对前端传递来的图像进行解码,随后将人脸图片按照注册用户名保存到指定路径。

def face_entry_getface(request):
        """保存注册的截图"""
        if request.method == 'POST':
            print("begin保存注册的截图")
            register_name = request.session.get('register_name')
            strs = request.POST['message']
            imgdata = base64.b64decode(strs)
            file_path = os.path.join('users', 'static', 'face_data', 'register', register_name + ".png")
            try:
                file = open(file_path, 'wb+')
                file.write(imgdata)
                file.close()
            except Exception as err:
                print(err)
            if shape_pic(file_path):
                # request.session['register_name'] = None
                return HttpResponse('login')  # 自动跳转到登录页面
            return HttpResponse('no')

#### 2.人脸识别登录部分 ####

##### 登录界面的前端代码 #####

同注册界面一样，先使用[tracking.js][]在前端识别当前摄像头是否拍到人脸，这样可以提高程序运行的效率。同样的，捕获到人脸后通过JS的postFace()函数将人脸数据传输到后台进行对比。

<!DOCTYPE html>
    <html lang="zh-CN">
    <head>
        <link rel="stylesheet" href="/static/css/login.css">
        <link rel="apple-touch-icon" sizes="180x180" href="/static/assets/apple-icon-180x180.png">
        <link href="/static/assets/favicon.ico" rel="icon">
        <title>login page</title>
        <link href="/static/css/main.3f6952e4.css" rel="stylesheet">
        <style>
      video, canvas {
        
        margin-left: 10px;
        margin-top: 20px;
        position: absolute;
      }
        </style>
        <head/>
    <body class="minimal">
    <div id="site-border-left"></div>
    <div id="site-border-right"></div>
    <div id="site-border-top"></div>
    <div id="site-border-bottom"></div>
    
    <header>
        <nav class="navbar  navbar-fixed-top navbar-default">
            <div class="container">
                <div class="collapse navbar-collapse" id="navbar-collapse">
                    <ul class="nav navbar-nav ">
                        <li><a href="{% url 'facerecognition:index' %}">01 : 首页</a></li>
                        <li><a href="{% url 'facerecognition:face_recognition'%}">02 : 人脸模糊</a></li>
                        <li><a href="{% url 'facerecognition:about' %}">03 : 关于本站</a></li>
                    </ul>
                    <ul class="nav navbar-nav navbar-right">
                        {
        % if request.session.is_login %}
                        <li><a>你好,{
        {
         request.session.user_name }}</a></li>
                        <li><a href="{% url 'users:logout' %}">登出</a></li>
                        {
        % else %}
                        <li><a href="{% url 'users:login' %}">登录</a></li>
                        <li><a href="{%  url 'users:register' %}">注册</a></li>
                        {
        % endif %}
                    </ul>
                </div>
            </div>
        </nav>
    </header>
    
    
    
    <div class="section-container">
        <div class="container">
            <div class="row-input-top">
                <div class="col-xs-12">
                    <div class="section-container-spacer text-center">
                        <h1 class="h2">欢迎登录</h1>
                    </div>
                    <div class="demo-container">
                        <video id="video" width="320" height="240" autoplay></video>
                        <canvas id="canvas" width="320" height="240"></canvas>
                    </div>
                    <div class="row-input text-center">
                        <div class="col-md-10 col-md-offset-1">
                            <form action="{% url 'users:login' %}" class="reveal-content"  method="post">
                                 {
        % if message %}
                                <div class="alert alert-warning">{
        {
         message }}</div>
                                {
        % endif %}
                                <div class="row-input-top">
                                    <div class="col-md-7">
                                        {
        % csrf_token %}
                                        <div class="form-group">
                                            {
        {
         login_form.username }}
    
                                        </div>
                                        <div class="form-group">
                                            {
        {
         login_form.password }}
    
                                        </div>
                                        <button type="reset" class="btn btn-default pull-left">重置</button>
                                        <button type="submit" class="btn btn-default pull-right">登录</button>
                                        <p>没有账户？点此<a href="{% url 'users:register' %}">注册</a></p>
                                    </div>
                                </div>
                            </form>
                        </div>
                    </div>
                </div>
            </div>
        </div>
    </div>
    
    <script>
        var video = document.getElementById('video');
          var canvas = document.getElementById('canvas');
          var context = canvas.getContext('2d');
          function getUserMediaToPhoto(constraints,success,error) {
        
            if(navigator.mediaDevices.getUserMedia){
        
                //最新标准API
                navigator.mediaDevices.getUserMedia(constraints).then(success).catch(error);
            }else if (navigator.webkitGetUserMedia) {
        
                //webkit核心浏览器
                navigator.webkitGetUserMedia(constraints,success,error);
            }else if(navigator.mozGetUserMedia){
        
                //firefox浏览器
                navigator.mozGetUserMedia(constraints,success,error);
            }else if(navigator.getUserMedia){
        
                //旧版API
                navigator.getUserMedia(constraints,success,error);
            }
        }
        //成功回调函数
        function success(stream){
        
            if ("srcObject" in video) {
        
                video.srcObject = stream;
            } else {
        
    
                video.src = window.URL.createObjectURL(stream);
            }
            video.onloadedmetadata = function(e) {
        
                video.play();
            };
            postFace()
        }
        function error(error) {
        
            console.log('访问用户媒体失败：',error.name,error.message);
        }
        function postFace() {
        
            setTimeout(function () {
        
                context.drawImage(video,0,0,320,240);
                img=canvas.toDataURL('image/jpg')
                {
        #获取完整的base64编码#}
                img=img.split(',')[1]
                //将照片以base64用ajax传到后台
                $.ajax({
        
    		        type:"POST",
                    url:'get_face/',
                    //必须添加 csrf_token
                    beforeSend:function (xhr,setting) {
        
                        xhr.setRequestHeader("X-CSRFToken","{
        { csrf_token }}");
                    },
                    data:{
        
                        message:img
                    },
                    success:function (callback) {
        
                        if(callback=='no'){
        
                            postFace()
                        }else {
        
                            window.location.href=callback
                        }
                    },
                    error:function (callback) {
        
                        postFace()
                    }
                })
            },300)
        }
        if(navigator.mediaDevices.getUserMedia || navigator.webkitGetUserMedia || navigator.mozGetUserMedia || navigator.getUserMedia)
        {
        
            getUserMediaToPhoto({
        video:{
        width:320,height:240}},success,error);
        }else{
        
            alert('你的浏览器不支持访问用户媒体设备');
        }
    </script>
    <script type="text/javascript" src="/static/js/main.70a66962.js"></script>
    <script type="text/javascript" src="/static/js/jquery.js"></script>
    </body>
    </html>

##### 人脸登录后台代码: #####

后台通过get\_face()函数接受到人脸图像，对其进行解码，随后将其保存到人脸存储的临时路径。通过check\_face()函数传入需要对比的两张图片的路径，进行人脸对比

def get_face(request):
        """人脸登录验证"""
        if request.POST:
            time = datetime.datetime.now().strftime('%Y%m%d%H%M%S')
            strs = request.POST['message']
            imgdata = base64.b64decode(strs)
            file_path = os.path.join('users', 'static', 'face_data', 'confirm', time + ".png")
            try:
                file = open(file_path, 'wb+')
                file.write(imgdata)
                file.close()
                username = check_face('./users/static/face_data/register/', file_path)
                if username:
                    # 此处会有bug，数据库用户和人脸识别用户不一致的问题
                    request.session['is_login'] = True
                    request.session['user_name'] = username
                    return HttpResponse('index')
                else:
                    return HttpResponse('no')
            except Exception as err:
                print(err)
            finally:
                delete_pic('./users/static/face_data/confirm/')
        return HttpResponse('no')

check\_face()人脸对比代码，调用的是[face\_recognition][face_recognition]库进行离线对比，如若对比成功，则返回文件名（即当前登录的用户名），否则返回None

def check_face(known_path, unknown_path):
        """
        用于人脸判别
        :param known_path:
        :param unknown_path:
        :return:
        """
        picture_of_known = []  # 已知的图片文件
        known_faces_encoding = []       # 已知的图片编码
    
        # 加载图片
        files = file_name(known_path)
        for i in range(len(files)):
            picture_of_known.append(face_recognition.load_image_file(known_path+files[i]))
    
        unknown_picture = face_recognition.load_image_file(unknown_path)
    
        # 编码图片
        for i in range(len(picture_of_known)):
            known_face_encodings = face_recognition.face_encodings(picture_of_known[i])
            if len(known_face_encodings) > 0:
                known_faces_encoding.append(known_face_encodings[0])
        unknown_face_encoding = face_recognition.face_encodings(unknown_picture)
        if len(unknown_face_encoding) > 0:
            unknown_face_encoding = unknown_face_encoding[0]
            # 人脸验证的结果
            results = face_recognition.compare_faces(known_faces_encoding, unknown_face_encoding, tolerance=0.5)
            for i in range(len(results)):
                if results[i]:
                    file = files[i].split(".")[0]
                    print("file:"+file)
                    return file
        return None

#### 3.文件上传部分 ####

##### 图片模糊前端代码 #####

文件上传部分主要是通过[WebUploader][webuploader]组件来实现的，将图片/视频上传到后台，后台识别其中的人脸部分，对其进行高斯模糊。对于视频这种大文件，采取的策略是分块上传之后进行合并。

<!DOCTYPE html>
    <html lang="zh-CN">
    <head>
        <meta charset="UTF-8">
        <meta content="IE=edge" http-equiv="X-UA-Compatible">
        <meta content="width=device-width,initial-scale=1" name="viewport">
        <meta content="description" name="description">
        <meta name="google" content="notranslate"/>
        <meta content="Mashup templates have been developped by Orson.io team" name="author">
    
        
        <meta name="msapplication-tap-highlight" content="no">
    
        <link rel="apple-touch-icon" sizes="180x180" href="/static/assets/apple-icon-180x180.png">
        <link href="/static/assets/favicon.ico" rel="icon">
    
        <title>Title page</title>
        <link rel="stylesheet" type="text/css" href="/static/css/webuploader.css">
        <link rel="stylesheet" type="text/css" href="/static/css/demo.css">
        <link rel="stylesheet" type="text/css" href="/static/css/style.css">
        <link href="/static/css/main.3f6952e4.css" rel="stylesheet">
    </head>
    
    <body class="">
    <div id="site-border-left"></div>
    <div id="site-border-right"></div>
    <div id="site-border-top"></div>
    <div id="site-border-bottom"></div>
    
    <header>
        <nav class="navbar  navbar-fixed-top navbar-default">
            <div class="container">
                <div class="collapse navbar-collapse" id="navbar-collapse">
                    <ul class="nav navbar-nav ">
                        <li><a href="{% url 'facerecognition:index' %}">01 : 首页</a></li>
                        <li><a href="{% url 'facerecognition:face_recognition'%}">02 : 人脸模糊</a></li>
                        <li><a href="{% url 'facerecognition:about' %}">03 : 关于本站</a></li>
                    </ul>
                    <ul class="nav navbar-nav navbar-right">
                        {
        % if request.session.is_login %}
                        <li><a>你好,{
        {
         request.session.user_name }}</a></li>
                        <li><a href="{% url 'users:logout' %}">登出</a></li>
                        {
        % else %}
                        <li><a href="{% url 'users:login' %}">登录</a></li>
                        <li><a href="{%  url 'users:register' %}">注册</a></li>
                        {
        % endif %}
                    </ul>
                </div>
            </div>
        </nav>
    </header>
    
    <div class="section-container">
        <div class="container">
            <div class="row">
    
                <div class="col-sm-8 col-sm-offset-2 section-container-spacer">
                    <div class="text-center">
                        <h1 class="h2">02 : 人脸模糊</h1>
                        <p>
                        <h3>这是一个人脸模糊程序，你可以通过拖拽或者点击添加图片按钮,<br/>
                        上传照片或者视频，我们会将其中的人脸模糊后返回给你</h3></p>
                    </div>
                </div>
    
                <div id="post-container" class="container-upload">
                    
    
                    <div id="uploader" class="wu-example">
                        <div class="queueList">
                            <div id="dndArea" class="placeholder">
                                <div id="filePicker"></div>
                                <p>或将照片拖到这里，单次最多可选300张</p>
                            </div>
                        </div>
                        <div class="statusBar" style="display:none;">
                            <div class="progress">
                                <span class="text">0%</span>
                                <span class="percentage"></span>
                            </div>
                            <div class="info"></div>
                            <div class="btns">
                                <div id="filePicker2"></div>
                                <div class="uploadBtn">开始上传</div>
                            </div>
                        </div>
                    </div>
                </div>
                <script type="text/javascript">
                // 添加全局站点信息
                var BASE_URL = '/webuploader';
                </script>
                <script type="text/javascript" src="/static/js/jquery.js"></script>
                <script type="text/javascript" src="/static/js/bootstrap.min.js"></script>
                <script type="text/javascript" src="/static/js/global.js"></script>
                <script type="text/javascript" src="/static/js/webuploader.js"></script>
                <script type="text/javascript" src="/static/js/demo.js"></script>
            </div>
        </div>
    </div>
    <footer class="footer-container text-center">
        <div class="container">
            <div class="row">
                <div class="col-xs-12">
                    <p>© UNTITLED | Website created by Micah Yin from GUET - More information <a href="http://www.guet.edu.cn/" target="_blank" title="模板之家">GUET</a> </p>
                </div>
            </div>
        </div>
    </footer>
    
    <script>
      document.addEventListener("DOMContentLoaded", function (event) {
        
         navActivePage();
      });
    </script>
    
    <script type="text/javascript" src="/static/js/main.70a66962.js"></script>
    </body>
    
    </html>

##### 图片模糊后台代码： #####

file\_upload()和file\_merge()用于处理文件上传和分块文件的合并

@csrf_exempt
    def file_upload(request):
        """用于处理上传文件"""
        if request.method == 'POST':
            upload_file = request.FILES.get('file')
            chunk = request.POST.get('chunk', 0)  # 获取该分片在所有分片中的序号
            if chunk != 0:
                filename = upload_file.name
                request.session['filename'] = filename  # 分片上传的标志
                filename = '%s%s' % (filename, chunk)  # 构成该分片唯一标识符
            else:
                filename = upload_file.name
            username = request.session.get('user_name')
            default_storage.save('./upload/%s/%s' % (username, filename), ContentFile(upload_file.read()))  # 保存分片到本地
        return HttpResponse('face_recognition.html', locals())
    
    
    @csrf_exempt
    def file_merge(request):
        """整合分片上传的文件"""
        filename = request.session.get('filename')
        username = request.session.get('user_name')
        # 若有分块上传的文件，则对文件进行整合
        if filename:
            upload_file = filename.split(".")[0]
            join_file("./upload/%s/" % username, filename, "./upload/%s/" % username, upload_file)
            request.session['filename'] = None
        else:
            pass
        # 遍历所有文件，对视频和图片进行人脸模糊处理
        in_dir = "./upload/%s/" % username
        out_dir = "./download/%s/" % username
        if not os.path.exists(out_dir):
            os.mkdir(out_dir)
        in_files = os.listdir(in_dir)
        in_files.sort()
        for file in in_files:
            in_path = os.path.join(in_dir, file)
            out_path = os.path.join(out_dir, file)
            if file.split(".")[1] == 'mp4':
                # print("注释掉了mp4!!!!")
                blur_video(out_path, in_path)
            else:
                blur_image(out_path, in_path)
        # 对文件进行压缩
        zip_ya(out_dir, "./download/%s.zip" % username)
        return HttpResponse('to_download')

#### 4.人脸区域模糊功能 ####

人脸部分模糊的代码，调用了opencv的分类器，对人脸区域进行了高斯模糊

def blur_image(filename, file_path):
        """
        图片人脸部分模糊
        :param filename:人脸模糊后的输出路径
        :param file_path: 读取文件的路径
        :return:
        """
        classfier = cv2.CascadeClassifier("/usr/local/share/OpenCV/haarcascades/haarcascade_frontalface_alt2.xml")
        src = cv2.imread(file_path)
        grey = cv2.cvtColor(src, cv2.COLOR_BGR2GRAY)
    
        # 人脸检测，1.2和2分别为图片缩放比例和需要检测的有效点数
        faceRects = classfier.detectMultiScale(grey, scaleFactor=1.2, minNeighbors=3, minSize=(32, 32))
        if len(faceRects) > 0:  # 大于0则检测到人脸
            for faceRect in faceRects:  # 单独框出每一张人脸
                x, y, w, h = faceRect
                image = src[y - 10:y + h + 10, x - 10:x + w + 10]  # 截取人脸部分
                dst = np.empty(image.shape, dtype=np.uint8)        # 空白numpy数组
                cv2.blur(image, (25, 25), dst)                     # 均值滤波,输出微dst
                src[y - 10:y + h + 10, x - 10:x + w + 10] = dst
        # 保存图像
        cv2.imwrite(filename, src)

[Link 1]: https://www.bilibili.com/video/av96339931
[face_recognition]: https://github.com/ageitgey/face_recognition
[webuploader]: http://fex.baidu.com/webuploader/
[tracking.js]: https://github.com/eduardolundgren/tracking.js