基于SVM算法的人脸表情识别【人工智能】

一、图片预处理

准备好我们所需要的数据

首先打开数据集，将人脸检测出来并对图片进行裁剪

import dlib  # 人脸识别的库dlib
import numpy as np  # 数据处理的库numpy
import cv2  # 图像处理的库OpenCv
import os

# dlib预测器
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor('shape_predictor_68_face_landmarks.dat')

# 读取图像的路径
path_read = "D:\\Firefox_download\\genki4k\\files"
num = 0
for file_name in os.listdir(path_read):
    # aa是图片的全路径
    aa = (path_read + "/" + file_name)
    # 读入的图片的路径中含非英文
    img = cv2.imdecode(np.fromfile(aa, dtype=np.uint8), cv2.IMREAD_UNCHANGED)
    # 获取图片的宽高
    img_shape = img.shape
    img_height = img_shape[0]
    img_width = img_shape[1]

    # 用来存储生成的单张人脸的路径
    path_save = "D:\\Firefox_download\\genki4k\\files1"
    # dlib检测
    dets = detector(img, 1)
    print("人脸数：", len(dets))
    for k, d in enumerate(dets):
        if len(dets) > 1:
            continue
        num = num + 1
        # 计算矩形大小
        # (x,y), (宽度width, 高度height)
        pos_start = tuple([d.left(), d.top()])
        pos_end = tuple([d.right(), d.bottom()])

        # 计算矩形框大小
        height = d.bottom() - d.top()
        width = d.right() - d.left()

        # 根据人脸大小生成空的图像
        img_blank = np.zeros((height, width, 3), np.uint8)
        for i in range(height):
            if d.top() + i >= img_height:  # 防止越界
                continue
            for j in range(width):
                if d.left() + j >= img_width:  # 防止越界
                    continue
                img_blank[i][j] = img[d.top() + i][d.left() + j]
        img_blank = cv2.resize(img_blank, (200, 200), interpolation=cv2.INTER_CUBIC)

        cv2.imencode('.jpg', img_blank)[1].tofile(path_save + "\\" + "file" + str(num) + ".jpg")  # 正确方法

运行完成之后可以看到程序识别出了3878张人脸

二、划分数据集

代码：

import os, shutil

# 原始数据集路径
original_dataset_dir = 'D:\\Firefox_download\\genki4k\\files1'

# 新的数据集
base_dir = 'D:\\Firefox_download\\genki4k\\files2'
os.mkdir(base_dir)

# 训练图像、验证图像、测试图像的目录
train_dir = os.path.join(base_dir, 'train')
os.mkdir(train_dir)
validation_dir = os.path.join(base_dir, 'validation')
os.mkdir(validation_dir)
test_dir = os.path.join(base_dir, 'test')
os.mkdir(test_dir)

train_cats_dir = os.path.join(train_dir, 'smile')
os.mkdir(train_cats_dir)

train_dogs_dir = os.path.join(train_dir, 'unsmile')
os.mkdir(train_dogs_dir)

validation_cats_dir = os.path.join(validation_dir, 'smile')
os.mkdir(validation_cats_dir)

validation_dogs_dir = os.path.join(validation_dir, 'unsmile')
os.mkdir(validation_dogs_dir)

test_cats_dir = os.path.join(test_dir, 'smile')
os.mkdir(test_cats_dir)

test_dogs_dir = os.path.join(test_dir, 'unsmile')
os.mkdir(test_dogs_dir)

# 复制1000张笑脸图片到train_c_dir
fnames = ['file{}.jpg'.format(i) for i in range(1, 900)]
for fname in fnames:
    src = os.path.join(original_dataset_dir, fname)
    dst = os.path.join(train_cats_dir, fname)
    shutil.copyfile(src, dst)

fnames = ['file{}.jpg'.format(i) for i in range(900, 1350)]
for fname in fnames:
    src = os.path.join(original_dataset_dir, fname)
    dst = os.path.join(validation_cats_dir, fname)
    shutil.copyfile(src, dst)

# Copy next 500 cat images to test_cats_dir
fnames = ['file{}.jpg'.format(i) for i in range(1350, 1800)]
for fname in fnames:
    src = os.path.join(original_dataset_dir, fname)
    dst = os.path.join(test_cats_dir, fname)
    shutil.copyfile(src, dst)

fnames = ['file{}.jpg'.format(i) for i in range(2127, 3000)]
for fname in fnames:
    src = os.path.join(original_dataset_dir, fname)
    dst = os.path.join(train_dogs_dir, fname)
    shutil.copyfile(src, dst)

# Copy next 500 dog images to validation_dogs_dir
fnames = ['file{}.jpg'.format(i) for i in range(3000, 3878)]
for fname in fnames:
    src = os.path.join(original_dataset_dir, fname)
    dst = os.path.join(validation_dogs_dir, fname)
    shutil.copyfile(src, dst)

# Copy next 500 dog images to test_dogs_dir
fnames = ['file{}.jpg'.format(i) for i in range(3000, 3878)]
for fname in fnames:
    src = os.path.join(original_dataset_dir, fname)
    dst = os.path.join(test_dogs_dir, fname)
    shutil.copyfile(src, dst)

这里是创建一个新文件夹，如果已有同名的，程序运行会报错

base_dir = 'D:\\Firefox_download\\genki4k\\files2'
os.mkdir(base_dir)

运行效果：

三、Dlib提取人脸特征识别笑脸和非笑脸

代码：

import cv2                     #  图像处理的库 OpenCv
import dlib                    # 人脸识别的库 dlib
import numpy as np             # 数据处理的库 numpy
class face_emotion():
    def __init__(self):
        self.detector = dlib.get_frontal_face_detector()
        self.predictor = dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
        self.cap = cv2.VideoCapture(0)
        self.cap.set(3, 480)
        self.cnt = 0  
    def learning_face(self):
        line_brow_x = []
        line_brow_y = []
        while(self.cap.isOpened()):

            flag, im_rd = self.cap.read()
            k = cv2.waitKey(1)
            # 取灰度
            img_gray = cv2.cvtColor(im_rd, cv2.COLOR_RGB2GRAY)  
            faces = self.detector(img_gray, 0)

            font = cv2.FONT_HERSHEY_SIMPLEX
     
            # 如果检测到人脸
            if(len(faces) != 0):
                
                # 对每个人脸都标出68个特征点
                for i in range(len(faces)):
                    for k, d in enumerate(faces):
                        cv2.rectangle(im_rd, (d.left(), d.top()), (d.right(), d.bottom()), (0,0,255))
                        self.face_width = d.right() - d.left()
                        shape = self.predictor(im_rd, d)
                        mouth_width = (shape.part(54).x - shape.part(48).x) / self.face_width 
                        mouth_height = (shape.part(66).y - shape.part(62).y) / self.face_width
                        brow_sum = 0 
                        frown_sum = 0 
                        for j in range(17, 21):
                            brow_sum += (shape.part(j).y - d.top()) + (shape.part(j + 5).y - d.top())
                            frown_sum += shape.part(j + 5).x - shape.part(j).x
                            line_brow_x.append(shape.part(j).x)
                            line_brow_y.append(shape.part(j).y)

                        tempx = np.array(line_brow_x)
                        tempy = np.array(line_brow_y)
                        z1 = np.polyfit(tempx, tempy, 1)  
                        self.brow_k = -round(z1[0], 3) 
                        
                        brow_height = (brow_sum / 10) / self.face_width # 眉毛高度占比
                        brow_width = (frown_sum / 5) / self.face_width  # 眉毛距离占比

                        eye_sum = (shape.part(41).y - shape.part(37).y + shape.part(40).y - shape.part(38).y + 
                                   shape.part(47).y - shape.part(43).y + shape.part(46).y - shape.part(44).y)
                        eye_hight = (eye_sum / 4) / self.face_width
                        if round(mouth_height >= 0.03) and eye_hight<0.56:
                            cv2.putText(im_rd, "smile", (d.left(), d.bottom() + 20), cv2.FONT_HERSHEY_SIMPLEX, 2,
                                            (0,255,0), 2, 4)

                        if round(mouth_height<0.03) and self.brow_k>-0.3:
                            cv2.putText(im_rd, "unsmile", (d.left(), d.bottom() + 20), cv2.FONT_HERSHEY_SIMPLEX, 2,
                                        (0,255,0), 2, 4)
                cv2.putText(im_rd, "Face-" + str(len(faces)), (20,50), font, 0.6, (0,0,255), 1, cv2.LINE_AA)
            else:
                cv2.putText(im_rd, "No Face", (20,50), font, 0.6, (0,0,255), 1, cv2.LINE_AA)
            im_rd = cv2.putText(im_rd, "S: screenshot", (20,450), font, 0.6, (255,0,255), 1, cv2.LINE_AA)
            im_rd = cv2.putText(im_rd, "Q: quit", (20,470), font, 0.6, (255,0,255), 1, cv2.LINE_AA)
            if (cv2.waitKey(1) & 0xFF) == ord('s'):
                self.cnt += 1
                cv2.imwrite("screenshoot" + str(self.cnt) + ".jpg", im_rd)
            # 按下 q 键退出
            if (cv2.waitKey(1)) == ord('q'):
                break
            # 窗口显示
            cv2.imshow("Face Recognition", im_rd)
        self.cap.release()
        cv2.destroyAllWindows()
if __name__ == "__main__":
    my_face = face_emotion()
    my_face.learning_face()

四、参考文献

Python人脸识别微笑检测