face_recognition小实战：显示未知图片中已知人物的脸

实验要求：

给一张已知的人脸(杨幂)，让你在未知图片中找到杨幂，并且在那张未知图片中把杨幂的脸标记起来

思路：

1. load_image_file方法加载已知图像和待检测图像

2. face_encodings方法分别返回已知人脸的编码信息和待检测图像的编码信息

3. face_locations方法得到多人图中，这几个人的面部位置

4. 遍历待检测图片每个人的人脸，以及遍历每个人人脸的位置坐标，提取人脸特征，与已知人脸特征作比对，如果是True的话，返回匹配成功信息

5. 在未知图像上对匹配到的人脸进行标记

代码如下：

import face_recognition
from PIL import Image, ImageDraw
import cv2
# 加载一张合照
unknown_image = face_recognition.load_image_file('C:/Users/lenovo/Desktop/opencv/daima/banknum/stage1/test_data/test3.jpg')
# 加载一张单人照片
known_image = face_recognition.load_image_file('C:/Users/lenovo/Desktop/opencv/daima/banknum/stage1/test_data/test1.jpg')

results = []

known_face_encodings = face_recognition.face_encodings(known_image)[0]
# face_encodings返回的是列表类型，我们只需要拿到一个人脸编码即可
unknown_face_encodings = face_recognition.face_encodings(unknown_image)

# 得到多人图片两个人的面部位置
face_locations = face_recognition.face_locations(unknown_image)

for i in range(len(face_locations)):  # face_locations的长度就代表有多少张脸
    top, right, bottom, left = face_locations[i]  # 取出每张脸的四个位置坐标
    face_image = unknown_image[top:bottom, left:right]
    face_encoding = face_recognition.face_encodings(face_image)  # 把单个的人脸特征提取出来
    if face_encoding:
        result = {}
        matches = face_recognition.compare_faces([unknown_face_encodings[i]], known_face_encodings, tolerance=0.39)
        if True in matches:
            print('在位置的图片中找到了已知面孔')
            result['face_encoding'] = face_encoding
            result['is_view'] = True
            result['location'] = face_locations[i]
            result['face_id'] = i+1
            results.append(result)

            if result['is_view']:
                print('已知面孔匹配上照片上的第{}张脸！！'.format(result['face_id']))

view_face_locations = [i['location'] for i in results if i['is_view']]

if len(view_face_locations) > 0:  # 表示找到了至少大于0个的匹配特征
    for location in view_face_locations:
        top, right, bottom, left = location  # 取出每张脸的四个位置坐标
        start = (left, top)
        end = (right, bottom)

        cv2.rectangle(unknown_image, start, end, (0, 255, 255), thickness=2)
        font = cv2.FONT_HERSHEY_DUPLEX
        cv2.putText(unknown_image, 'yangmi', (left+6, bottom+16), font, 1.0, (0, 0, 255), thickness=1)

cv2.imshow('known', known_image)
cv2.imshow('window', unknown_image)
cv2.waitKey()

运行结果：