吾生也有涯，而知也无涯~欢迎优化补充、指正！
本文主要介绍基于百度的H5业务场景下人脸识别的创建与实现!
首先百度开放平台进入人脸实名认证控制台新建项目:

接着在方案清单页面点击「新建方案」按钮，方案类型选择「H5实名认证」
授权声明可以跳过

接着根据业务需求选择身份证信息录入等功能,官方提供多种认证过程,根据自身需求选用!
本例采用照片活体认证,效果图如下:

样本核验指标如下,分为严格、正常、宽松三个等级，越严格，图片对角度、模糊度、遮挡等信息参数把控越高，推荐使用宽松。

认证结果对比分数值官方推荐阀值80,可根据自身需求进行调整
完成项目配置提交后可扫码体验认证流程:

接下来进入验证实现步骤
先上代码

public function bd(){
        global $_W,$_GPC;
        if(empty($_W['openid'])){
            header('location: ' . mobileUrl('account/login'));
        }
        $insertid = $_GPC['insertid']; //自定义参数,与实际认证业务处理无关

        $url = 'https://aip.baidubce.com/oauth/2.0/token';
        $post_data = [
            'grant_type' => 'cli********ls',
            'client_id' => 'L45W********n3MQ',
            'client_secret' => '1Tq9********jGb6',
        ];

        $res = $this->request_post($url, http_build_query($post_data));
        $res = json_decode($res, true);

        $accessToken = $res['access_token'];

        /**
         *  获取Verify Token
         */
        $url = sprintf("https://aip.baidubce.com/rpc/2.0/brain/solution/faceprint/verifyToken/generate?access_token=%s", $accessToken);
        $post_data = [
            'plan_id' => "10671" // H5认证方案ID
        ];
        $res = $this->request_post($url, json_encode($post_data));
        $res = json_decode($res, true);
        $verifyToken = $res['result']['verify_token'];
        // 验证成功返回链接
        $successUrl =  mobileUrl("member/index/submit",['insertid' => $insertid,'access_token'=>$accessToken,'verify_token'=>$verifyToken],true);
        // 验证失败返回链接
        $failedUrl = mobileUrl("member/index/submit",['insertid' => $insertid,'access_token'=>$accessToken,'verify_token'=>$verifyToken],true);
        // 生成H5验验跳转链接
        $h5Url = sprintf("https://brain.baidu.com/face/print/?token=%s&successUrl=%s&failedUrl=%s", $verifyToken, urlencode($successUrl), urlencode($failedUrl));
        header(sprintf("Location: %s", $h5Url));

        //include($this->template());
    }
    public function request_post($url = '', $param = '')
    {
        if (empty($url) || empty($param)) {
            return false;
        }

        $postUrl = $url;
        $curlPost = $param;
        $curl = curl_init();//初始化curl
        curl_setopt($curl, CURLOPT_URL, $postUrl);//抓取指定网页
        curl_setopt($curl, CURLOPT_HEADER, 0);//设置header
        curl_setopt($curl, CURLOPT_RETURNTRANSFER, 1);//要求结果为字符串且输出到屏幕上
        curl_setopt($curl, CURLOPT_POST, 1);//post提交方式
        curl_setopt($curl, CURLOPT_POSTFIELDS, $curlPost);
        $data = curl_exec($curl);//运行curl
        curl_close($curl);

        return $data;
    }

如代码所示,先获取access_token(access_token的有效期为30天,建议每次请求重新获取),然后以此获取verify_token
业务H5网页通过获取token接口返回的verify_token信息请求认证H5页面，进行用户端流程操作
URL中的信息填写如下所示：
（1）token：填写verify_token
（2）successUrl：请求成功跳转的网址，网址需要加http/https前缀
（3）failedUrl：请求失败跳转的网址，网址需要加http/https前缀
用户端操作完成后回调上述业务地址,我们在此进行下一步处理

public function submit()
    {
        global $_W,$_GPC;
        $accessToken = $_GPC['access_token'];
        $verifyToken = $_GPC['verify_token'];
        $insertid = $_GPC['insertid']; //自定义参数,与实际认证业务处理无关
        /**
         * 获取认证结果
         */
        $apiUrl = sprintf('https://aip.baidubce.com/rpc/2.0/brain/solution/faceprint/result/detail?access_token=%s', $accessToken);
        $post_data = [
            'verify_token' => $verifyToken // H5认证方案ID
        ];
        $res = $this->request_post($apiUrl, json_encode($post_data));
        $res = json_decode($res, true);
        if(!$res['success'] || $res['result']['verify_result']['score'] < 0.7){
            $this->message("实名认证失败，请返回核对姓名和身份证号码",mobileUrl("member/dataform",[],true));
        }
        # other code
    }

如上所示,根据verify_token进行认证查询,verify_token的独立生成方法见上文!
返回示例如下:

{
    "success": true,
    "result": {
        "idcard_ocr_result": {
            "address": "江苏省淮安市淮阴区XXX",
            "birthday": "19691211",
            "name": "孙XX",
            "id_card_number": "320821XXXXXXX",
            "gender": "女",
            "nation": "汉",
            "expire_time": "XXX",
            "issue_authority": "XXX",
            "issue_time": "XXX"
        },
        "verify_result": {
            "liveness_score": 0.39,
            "score": 11.65885925
        },
        "idcard_confirm": {
            "name": "孙XX",
            "idcard_number": "320821XXXXXXX"
        }
    },
    "log_id": "1436111694"
}

示例根据score进行判断,数值越大则认证越严格!
以上就是人脸识别的全部流程,博文创建时间仓促,欢迎指正与讨论!
注:示例是在微擎下搭建的,其它框架请自行调整!