1dlib获取关键点
https://www.jianshu.com/p/848014d8dea9
https://www.pyimagesearch.com/2017/05/01/install-dlib-raspberry-pi/
库下载
https://github.com/davisking/dlib
表情识别教程
https://www.cnblogs.com/qsyll0916/p/8893790.html
识别代码
https://gitee.com/Andrew_Qian/face/blob/master/from_video.py
依赖权重
二人脸表情识别系统（含UI界面，python实现
https://blog.csdn.net/qq_32892383/article/details/91347164
dilb程序实现
面部表情跟踪的原理就是检测人脸特征点，根据特定的特征点可以对应到特定的器官，比如眼睛、鼻子、嘴巴、耳朵等等，以此来跟踪各个面部器官的动作。
https://blog.csdn.net/hongbin_xu/article/details/79926839
三、安装依赖库
dlib需要以下依赖：
- Boost
- Boost.Python
- CMake
- X11
安装方法：
$ sudo apt-get update
$ sudo apt-get install build-essential cmake libgtk-3-dev libboost-all-dev -y
四、用pip3安装其他dlib运行依赖的库
$ pip3 install numpy
$ pip3 install scipy
$ pip3 install scikit-image
五、正式安装
解压下载好的dlib，进入dlib目录后
$ sudo python3 setup.py install
这一步耗时是最长的了，耐心等待。
六、验证
$ python3
Python 3.4.2 (default, Oct 19 2014, 13:31:11)
[GCC 4.9.1] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import dlib
>>>
七、把虚拟内存和GPU使用内存改回原始值
修改方法见“二、修改前的准备工作”
收工。
python代码
#!Anaconda/anaconda/python
#coding: utf-8
"""
从视屏中识别人脸，并实时标出面部特征点
"""
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
class face_emotion():
def __init__(self):
# 使用特征提取器get_frontal_face_detector
self.detector = dlib.get_frontal_face_detector()
# dlib的68点模型，使用作者训练好的特征预测器
self.predictor = dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
#建cv2摄像头对象，这里使用电脑自带摄像头，如果接了外部摄像头，则自动切换到外部摄像头
self.cap = cv2.VideoCapture(0)
# 设置视频参数，propId设置的视频参数，value设置的参数值
self.cap.set(3, 480)
# 截图screenshoot的计数器
self.cnt = 0
def learning_face(self):
# 眉毛直线拟合数据缓冲
line_brow_x = []
line_brow_y = []
# cap.isOpened（） 返回true/false 检查初始化是否成功
while(self.cap.isOpened()):
# cap.read()
# 返回两个值：
# 一个布尔值true/false，用来判断读取视频是否成功/是否到视频末尾
# 图像对象，图像的三维矩阵
flag, im_rd = self.cap.read()
# 每帧数据延时1ms，延时为0读取的是静态帧
k = cv2.waitKey(1)
# 取灰度
img_gray = cv2.cvtColor(im_rd, cv2.COLOR_RGB2GRAY)
# 使用人脸检测器检测每一帧图像中的人脸。并返回人脸数rects
faces = self.detector(img_gray, 0)
# 待会要显示在屏幕上的字体
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)):
# enumerate方法同时返回数据对象的索引和数据，k为索引，d为faces中的对象
for k, d in enumerate(faces):
# 用红色矩形框出人脸
cv2.rectangle(im_rd, (d.left(), d.top()), (d.right(), d.bottom()), (0, 0, 255))
# 计算人脸热别框边长
self.face_width = d.right() - d.left()
# 使用预测器得到68点数据的坐标
shape = self.predictor(im_rd, d)
# 圆圈显示每个特征点
for i in range(68):
cv2.circle(im_rd, (shape.part(i).x, shape.part(i).y), 2, (0, 255, 0), -1, 8)
#cv2.putText(im_rd, str(i), (shape.part(i).x, shape.part(i).y), cv2.FONT_HERSHEY_SIMPLEX, 0.5,
# (255, 255, 255))
# 分析任意n点的位置关系来作为表情识别的依据
mouth_width = (shape.part(54).x - shape.part(48).x) / self.face_width # 嘴巴咧开程度
mouth_higth = (shape.part(66).y - shape.part(62).y) / self.face_width # 嘴巴张开程度
# print("嘴巴宽度与识别框宽度之比：",mouth_width_arv)
# print("嘴巴高度与识别框高度之比：",mouth_higth_arv)
# 通过两个眉毛上的10个特征点，分析挑眉程度和皱眉程度
brow_sum = 0 # 高度之和
frown_sum = 0 # 两边眉毛距离之和
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)
# self.brow_k, self.brow_d = self.fit_slr(line_brow_x, line_brow_y) # 计算眉毛的倾斜程度
tempx = np.array(line_brow_x)
tempy = np.array(line_brow_y)
z1 = np.polyfit(tempx, tempy, 1) # 拟合成一次直线
self.brow_k = -round(z1[0], 3) # 拟合出曲线的斜率和实际眉毛的倾斜方向是相反的
brow_hight = (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
# print("眉毛高度与识别框高度之比：",round(brow_arv/self.face_width,3))
# print("眉毛间距与识别框高度之比：",round(frown_arv/self.face_width,3))
# 眼睛睁开程度
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
# print("眼睛睁开距离与识别框高度之比：",round(eye_open/self.face_width,3))
# 分情况讨论
# 张嘴，可能是开心或者惊讶
if round(mouth_higth >= 0.03):
if eye_hight >= 0.056:
cv2.putText(im_rd, "amazing", (d.left(), d.bottom() + 20), cv2.FONT_HERSHEY_SIMPLEX, 0.8,
(0, 0, 255), 2, 4)
else:
cv2.putText(im_rd, "happy", (d.left(), d.bottom() + 20), cv2.FONT_HERSHEY_SIMPLEX, 0.8,
(0, 0, 255), 2, 4)
# 没有张嘴，可能是正常和生气
else:
if self.brow_k <= -0.3:
cv2.putText(im_rd, "angry", (d.left(), d.bottom() + 20), cv2.FONT_HERSHEY_SIMPLEX, 0.8,
(0, 0, 255), 2, 4)
else:
cv2.putText(im_rd, "nature", (d.left(), d.bottom() + 20), cv2.FONT_HERSHEY_SIMPLEX, 0.8,
(0, 0, 255), 2, 4)
# 标出人脸数
cv2.putText(im_rd, "Faces: "+str(len(faces)), (20,50), font, 1, (0, 0, 255), 1, cv2.LINE_AA)
else:
# 没有检测到人脸
cv2.putText(im_rd, "No Face", (20, 50), font, 1, (0, 0, 255), 1, cv2.LINE_AA)
# 添加说明
im_rd = cv2.putText(im_rd, "S: screenshot", (20, 400), font, 0.8, (0, 0, 255), 1, cv2.LINE_AA)
im_rd = cv2.putText(im_rd, "Q: quit", (20, 450), font, 0.8, (0, 0, 255), 1, cv2.LINE_AA)
# 按下s键截图保存
if (k == ord('s')):
self.cnt+=1
cv2.imwrite("screenshoot"+str(self.cnt)+".jpg", im_rd)
# 按下q键退出
if(k == ord('q')):
break
# 窗口显示
cv2.imshow("camera", im_rd)
# 释放摄像头
self.cap.release()
# 删除建立的窗口
cv2.destroyAllWindows()
if __name__ == "__main__":
my_face = face_emotion()
my_face.learning_face()
# *_*coding:utf-8 *_*
# author: 许鸿斌
import sys
import cv2
import dlib
import os
import logging
import datetime
import numpy as np
def cal_face_boundary(img, shape):
for index_, pt in enumerate(shape.parts()):
if index_ == 0:
x_min = pt.x
x_max = pt.x
y_min = pt.y
y_max = pt.y
else:
if pt.x < x_min:
x_min = pt.x
if pt.x > x_max:
x_max = pt.x
if pt.y < y_min:
y_min = pt.y
if pt.y > y_max:
y_max = pt.y
# print('x_min:{}'.format(x_min))
# print('x_max:{}'.format(x_max))
# print('y_min:{}'.format(y_min))
# print('y_max:{}'.format(y_max))
# 如果出现负值，即人脸位于图像框之外的情况，应当忽视图像外的部分，将负值置为0
if x_min < 0:
x_min = 0
if y_min < 0:
y_min = 0
if x_min == x_max or y_min == y_max:
return None
else:
return img[y_min:y_max, x_min:x_max]
def draw_left_eyebrow(img, shape):
# 17 - 21
pt_pos = []
for index, pt in enumerate(shape.parts()[17:21 + 1]):
pt_pos.append((pt.x, pt.y))
for num in range(len(pt_pos)-1):
cv2.line(img, pt_pos[num], pt_pos[num+1], 255, 2)
def draw_right_eyebrow(img, shape):
# 22 - 26
pt_pos = []
for index, pt in enumerate(shape.parts()[22:26 + 1]):
pt_pos.append((pt.x, pt.y))
for num in range(len(pt_pos) - 1):
cv2.line(img, pt_pos[num], pt_pos[num + 1], 255, 2)
def draw_left_eye(img, shape):
# 36 - 41
pt_pos = []
for index, pt in enumerate(shape.parts()[36:41 + 1]):
pt_pos.append((pt.x, pt.y))
for num in range(len(pt_pos) - 1):
cv2.line(img, pt_pos[num], pt_pos[num + 1], 255, 2)
cv2.line(img, pt_pos[0], pt_pos[-1], 255, 2)
def draw_right_eye(img, shape):
# 42 - 47
pt_pos = []
for index, pt in enumerate(shape.parts()[42:47 + 1]):
pt_pos.append((pt.x, pt.y))
for num in range(len(pt_pos) - 1):
cv2.line(img, pt_pos[num], pt_pos[num + 1], 255, 2)
cv2.line(img, pt_pos[0], pt_pos[-1], 255, 2)
def draw_nose(img, shape):
# 27 - 35
pt_pos = []
for index, pt in enumerate(shape.parts()[27:35 + 1]):
pt_pos.append((pt.x, pt.y))
for num in range(len(pt_pos) - 1):
cv2.line(img, pt_pos[num], pt_pos[num + 1], 255, 2)
cv2.line(img, pt_pos[0], pt_pos[4], 255, 2)
cv2.line(img, pt_pos[0], pt_pos[-1], 255, 2)
cv2.line(img, pt_pos[3], pt_pos[-1], 255, 2)
def draw_mouth(img, shape):
# 48 - 59
pt_pos = []
for index, pt in enumerate(shape.parts()[48:59 + 1]):
pt_pos.append((pt.x, pt.y))
for num in range(len(pt_pos) - 1):
cv2.line(img, pt_pos[num], pt_pos[num + 1], 255, 2)
cv2.line(img, pt_pos[0], pt_pos[-1], 255, 2)
# 60 - 67
pt_pos = []
for index, pt in enumerate(shape.parts()[60:]):
pt_pos.append((pt.x, pt.y))
for num in range(len(pt_pos) - 1):
cv2.line(img, pt_pos[num], pt_pos[num + 1], 255, 2)
cv2.line(img, pt_pos[0], pt_pos[-1], 255, 2)
def draw_jaw(img, shape):
# 0 - 16
pt_pos = []
for index, pt in enumerate(shape.parts()[0:16 + 1]):
pt_pos.append((pt.x, pt.y))
for num in range(len(pt_pos) - 1):
cv2.line(img, pt_pos[num], pt_pos[num + 1], 255, 2)
# 获取logger实例，如果参数为空则返回root logger
logger = logging.getLogger("PedestranDetect")
# 指定logger输出格式
formatter = logging.Formatter('%(asctime)s %(levelname)-8s: %(message)s')
# 文件日志
# file_handler = logging.FileHandler("test.log")
# file_handler.setFormatter(formatter) # 可以通过setFormatter指定输出格式
# 控制台日志
console_handler = logging.StreamHandler(sys.stdout)
console_handler.formatter = formatter # 也可以直接给formatter赋值
# 为logger添加的日志处理器
# logger.addHandler(file_handler)
logger.addHandler(console_handler)
# 指定日志的最低输出级别，默认为WARN级别
logger.setLevel(logging.INFO)
pwd = os.getcwd()
predictor_path = os.path.join(pwd, 'shape_predictor_68_face_landmarks.dat')
logger.info(u'导入人脸检测器')
detector = dlib.get_frontal_face_detector()
logger.info(u'导入人脸特征点检测器')
predictor = dlib.shape_predictor(predictor_path)
cap = cv2.VideoCapture(0)
cnt = 0
total_time = 0
start_time = 0
while(1):
ret, frame = cap.read()
# cv2.imshow("window", frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
img = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
dets = detector(img, 1)
if dets:
logger.info('Face detected')
else:
logger.info('No face detected')
for index, face in enumerate(dets):
# print('face {}; left {}; top {}; right {}; bottom {}'.format(index, face.left(), face.top(), face.right(),
# face.bottom()))
shape = predictor(img, face)
# for index_, pt in enumerate(shape.parts()):
# pt_pos = (pt.x, pt.y)
# cv2.circle(frame, pt_pos, 2, (255, 0, 0), 1)
features = np.zeros(img.shape[0:-1], dtype=np.uint8)
for index_, pt in enumerate(shape.parts()):
pt_pos = (pt.x, pt.y)
cv2.circle(features, pt_pos, 2, 255, 1)
draw_left_eyebrow(features, shape)
draw_right_eyebrow(features, shape)
draw_left_eye(features, shape)
draw_right_eye(features, shape)
draw_nose(features, shape)
draw_mouth(features, shape)
draw_jaw(features, shape)
logger.info('face shape: {} {}'.format(face.right()-face.left(), face.bottom()-face.top()))
faceROI = cal_face_boundary(features, shape)
logger.info('ROI shape: {}'.format(faceROI.shape))
# faceROI = features[face.top():face.bottom(), face.left():face.right()]
faceROI = cv2.resize(faceROI, (500, 500), interpolation=cv2.INTER_LINEAR)
# logger.info('face {}'.format(index))
cv2.imshow('face {}'.format(index), faceROI)
if cnt == 0:
start_time = datetime.datetime.now()
cnt += 1
elif cnt == 100:
end_time = datetime.datetime.now()
frame_rate = float(100) / (end_time-start_time).seconds
# logger.info(start_time)
# logger.info(end_time)
logger.info(u'帧率：{:.2f}fps'.format(frame_rate))
cnt = 0
else:
cnt += 1
# logger.info(cnt)