基于人脸动态特征的活体检测系统实现

创作时间:

作者:

@小白创作中心

基于人脸动态特征的活体检测系统实现

引用

CSDN

https://blog.csdn.net/weixin_53742691/article/details/136610143

活体检测技术在人脸识别系统中扮演着至关重要的角色，它能够有效防止照片、视频等非活体介质的欺骗攻击。本文将介绍一种基于人脸动态特征（如眨眼、张嘴、摇头和点头）的活体检测方法，并提供完整的Python代码实现。

背景说明

在实际应用中，人脸识别系统常常面临以下挑战：

照片打卡问题：传统的打卡系统容易被照片欺骗
高昂的开发成本：现有的活体检测服务价格昂贵
低自主集成度：现有方案难以满足定制化需求

因此，开发一个成本低、可定制的活体检测系统具有重要意义。本文将介绍一种基于人脸动态特征的活体检测方法，通过分析视频流中的人脸动作来判断是否为活体。

技术原理

该系统主要通过分析视频流中的人脸动作来判断是否为活体。具体来说，系统会检测以下几种动作：

眨眼：通过计算眼睛的纵横比（EAR）来判断眼睛是否闭合
张嘴：通过计算嘴巴的纵横比（MAR）来判断嘴巴是否张开
摇头：通过分析鼻子到左右脸颊边界的距离变化来判断头部是否摇动
点头：通过分析眉毛到左右脸颊边界的距离变化来判断头部是否点头

代码实现

以下是完整的Python代码实现：

from scipy.spatial import distance as dist
from imutils.video import FileVideoStream
from imutils.video import VideoStream
from imutils import face_utils
import argparse
import imutils
import time
import dlib
import cv2
import numpy as np

def eye_aspect_ratio(eye):
    A = dist.euclidean(eye[1], eye[5])
    B = dist.euclidean(eye[2], eye[4])
    C = dist.euclidean(eye[0], eye[3])
    ear = (A + B) / (2.0 * C)
    return ear

def mouth_aspect_ratio(mouth):
    A = np.linalg.norm(mouth[2] - mouth[9])
    B = np.linalg.norm(mouth[4] - mouth[7])
    C = np.linalg.norm(mouth[0] - mouth[6])
    mar = (A + B) / (2.0 * C)
    return mar

def nose_jaw_distance(nose, jaw):
    face_left1 = dist.euclidean(nose[0], jaw[0])
    face_right1 = dist.euclidean(nose[0], jaw[16])
    face_left2 = dist.euclidean(nose[3], jaw[2])
    face_right2 = dist.euclidean(nose[3], jaw[14])
    face_distance = (face_left1, face_right1, face_left2, face_right2)
    return face_distance

def eyebrow_jaw_distance(leftEyebrow, jaw):
    eyebrow_left = dist.euclidean(leftEyebrow[2], jaw[0])
    eyebrow_right = dist.euclidean(leftEyebrow[2], jaw[16])
    left_right = dist.euclidean(jaw[0], jaw[16])
    eyebrow_distance = (eyebrow_left, eyebrow_right, left_right)
    return eyebrow_distance

def Face_Recognize(file_path):
    EYE_AR_THRESH = 0.27
    EYE_AR_CONSEC_FRAMES = 2
    MAR_THRESH = 0.5
    COUNTER_EYE = 0
    TOTAL_EYE = 0
    COUNTER_MOUTH = 0
    TOTAL_MOUTH = 0
    distance_left = 0
    distance_right = 0
    TOTAL_FACE = 0
    nod_flag = 0
    TOTAL_NOD = 0

    detector = dlib.get_frontal_face_detector()
    predictor = dlib.shape_predictor("./static/shape_predictor_68_face_landmarks.dat")
    (lStart, lEnd) = face_utils.FACIAL_LANDMARKS_IDXS["left_eye"]
    (rStart, rEnd) = face_utils.FACIAL_LANDMARKS_IDXS["right_eye"]
    (mStart, mEnd) = face_utils.FACIAL_LANDMARKS_IDXS["mouth"]
    (nStart, nEnd) = face_utils.FACIAL_LANDMARKS_IDXS["nose"]
    (jStart, jEnd) = face_utils.FACIAL_LANDMARKS_IDXS['jaw']
    (Eyebrow_Start, Eyebrow_End) = face_utils.FACIAL_LANDMARKS_IDXS['left_eyebrow']

    vs = FileVideoStream(file_path).start()
    fileStream = True
    time.sleep(1.0)

    while True:
        if fileStream and not vs.more():
            break

        frame = vs.read()
        frame = imutils.resize(frame, width=600)
        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
        rects = detector(gray, 0)

        for rect in rects:
            shape = predictor(gray, rect)
            shape = face_utils.shape_to_np(shape)
            leftEye = shape[lStart:lEnd]
            rightEye = shape[rStart:rEnd]
            leftEAR = eye_aspect_ratio(leftEye)
            rightEAR = eye_aspect_ratio(rightEye)
            Mouth = shape[mStart:mEnd]
            mouthMAR = mouth_aspect_ratio(Mouth)
            nose = shape[nStart:nEnd]
            jaw = shape[jStart:jEnd]
            NOSE_JAW_Distance = nose_jaw_distance(nose, jaw)
            leftEyebrow = shape[Eyebrow_Start:Eyebrow_End]
            Eyebrow_JAW_Distance = eyebrow_jaw_distance(leftEyebrow, jaw)

            ear = (leftEAR + rightEAR) / 2.0
            mar = mouthMAR
            face_left1 = NOSE_JAW_Distance[0]
            face_right1 = NOSE_JAW_Distance[1]
            face_left2 = NOSE_JAW_Distance[2]
            face_right2 = NOSE_JAW_Distance[3]
            eyebrow_left = Eyebrow_JAW_Distance[0]
            eyebrow_right = Eyebrow_JAW_Distance[1]
            left_right = Eyebrow_JAW_Distance[2]

            if ear < EYE_AR_THRESH:
                COUNTER_EYE += 1
            else:
                if COUNTER_EYE >= EYE_AR_CONSEC_FRAMES:
                    TOTAL_EYE += 1
                COUNTER_EYE = 0

            if mar > MAR_THRESH:
                COUNTER_MOUTH += 1
            else:
                if COUNTER_MOUTH != 0:
                    TOTAL_MOUTH += 1
                    COUNTER_MOUTH = 0

            if face_left1 >= face_right1 + 2 and face_left2 >= face_right2 + 2:
                distance_left += 1
            if face_right1 >= face_left1 + 2 and face_right2 >= face_left2 + 2:
                distance_right += 1
            if distance_left != 0 and distance_right != 0:
                TOTAL_FACE += 1
                distance_right = 0
                distance_left = 0

            if eyebrow_left + eyebrow_right <= left_right + 3:
                nod_flag += 1
            if nod_flag != 0 and eyebrow_left + eyebrow_right >= left_right + 3:
                TOTAL_NOD += 1
                nod_flag = 0

            cv2.putText(frame, "Blinks: {}".format(TOTAL_EYE), (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
            cv2.putText(frame, "Mouth is open: {}".format(TOTAL_MOUTH), (10, 60), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
            cv2.putText(frame, "shake one's head: {}".format(TOTAL_FACE), (10, 90), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
            cv2.putText(frame, "nod: {}".format(TOTAL_NOD), (10, 120), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)

            cv2.putText(frame, "Live detection: wink(5)", (300, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
            if TOTAL_EYE >= 5:
                cv2.putText(frame, "open your mouth(3)", (300, 60), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
            if TOTAL_MOUTH >= 3:
                cv2.putText(frame, "shake your head(2)", (300, 90), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
            if TOTAL_FACE >= 2:
                cv2.putText(frame, "nod(2)", (300, 120), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
            if TOTAL_NOD >= 2:
                cv2.putText(frame, "Live detection: done", (300, 150), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)

        cv2.imshow("Frame", frame)
        key = cv2.waitKey(1) & 0xFF
        if key == ord("q"):
            break

    cv2.destroyAllWindows()
    vs.stop()

if __name__ == '__main__':
    Face_Recognize('path_to_your_video_file')

API接口

为了将上述功能封装成API服务，可以使用Flask框架。以下是API接口的实现代码：

import base64
from flask import Flask, request
import numpy as np
import cv2
import imutils
import dlib
from imutils import face_utils
import DynamicRecognition

app = Flask(__name__)

@app.route('/process_video', methods=['GET'])
def process_video():
    file_path = request.args.get('file_path')
    DynamicRecognition.Face_Recognize(file_path)
    return {"result": "Processed successfully"}

if __name__ == '__main__':
    app.run(debug=True)