您现在的位置是:首页 >学无止境 >Opencv项目实战:基于dlib的疲劳检测网站首页学无止境
Opencv项目实战:基于dlib的疲劳检测
简介Opencv项目实战:基于dlib的疲劳检测
文章目录
一、项目简介
本项目基于dlib库
提供的人脸检测器
、关键点定位工具
以及眼睛纵横比算法
完成。通过分析摄像头或视频流中的人脸,实时计算眼睛纵横比EAR(Eye Aspect Ratio),以判断眼睛是否闭合。通过统计眨眼次数,可以检测出眨眼的频率和时长,用于评估用户的注意力水平或疲劳状态。
二、算法原理
论文地址:https://vision.fe.uni-lj.si/cvww2016/proceedings/papers/05.pdf
- 由于眨眼动作是一个过程,而不是一个帧图像就能瞬间完成。故设置
连续帧数的阈值(=3)
,即连续三帧图像计算得到的EAR值都小于EAR的阈值(=3)
,则表示眨眼一次。- 眨眼检测与疲劳检测的区别就是连续帧数的阈值设置,原理相同!
三、环境配置
dlib库在计算机视觉和人工智能领域有广泛的应用,包括人脸识别、人脸表情分析、人脸关键点检测、物体检测和追踪等任务。它的简单易用性、高性能和丰富的功能使其成为研究人员和开发者的首选库之一。
- dlib是一个开源的C++机器学习和计算机视觉库,提供了人脸检测、关键点定位、人脸识别等功能,以及支持向量机和其他机器学习算法的实现,具有高性能和跨平台支持。
dlib工具的python API
下载地址:http://dlib.net/python/- dlib库包的介绍与使用:opencv+dlib人脸检测 + 人脸68关键点检测 + 人脸识别 + 人脸特征聚类 + 目标跟踪
3.1、dlib人脸检测器:dlib.get_frontal_face_detector()
dlib官方详细说明:dlib.get_frontal_face_detector()
3.2、dlib关键点定位工具:shape_predictor_68_face_landmarks.dat
dlib官方预训练工具的下载地址:http://dlib.net/files/
(1)5个关键点检测:shape_predictor_5_face_landmarks.dat
。五个点分别为:左右眼 + 鼻子 + 左右嘴角
(2)68个关键点检测:shape_predictor_68_face_landmarks.dat
脸部关键点注释详细请看:https://ibug.doc.ic.ac.uk/resources/facial-point-annotations/
四、项目实战(加载视频)
- 【参数配置】方式一:Pycharm + Terminal + 输入指令自动检测:
python detect_blinks.py --shape-predictor shape_predictor_68_face_landmarks.dat --video test.mp4
- 【参数配置】方式二:Pycharm + 点击Edit Configuration,输入配置参数
--shape-predictor shape_predictor_68_face_landmarks.dat --video test.mp4
,点击Run开始检测。
# 导入所需库
from scipy.spatial import distance as dist
from collections import OrderedDict
import numpy as np
import argparse
import time
import dlib
import cv2
# 定义脸部关键点索引
FACIAL_LANDMARKS_68_IDXS = OrderedDict([
("mouth", (48, 68)),
("right_eyebrow", (17, 22)),
("left_eyebrow", (22, 27)),
("right_eye", (36, 42)),
("left_eye", (42, 48)),
("nose", (27, 36)),
("jaw", (0, 17))
])
# 计算眼睛纵横比函数
def eye_aspect_ratio(eye):
# 计算垂直距离
A = dist.euclidean(eye[1], eye[5])
B = dist.euclidean(eye[2], eye[4])
# 计算水平距离
C = dist.euclidean(eye[0], eye[3])
# 计算眼睛纵横比EAR
ear = (A + B) / (2.0 * C)
return ear
# 解析命令行参数
ap = argparse.ArgumentParser()
ap.add_argument("-p", "--shape-predictor", required=True, help="面部地标预测器路径")
ap.add_argument("-v", "--video", type=str, default="", help="输入视频文件路径")
args = vars(ap.parse_args())
# 设置EAR阈值和连续帧数
EYE_AR_THRESH = 0.3
EYE_AR_CONSEC_FRAMES = 3
# 初始化计数器
COUNTER = 0 # 计算连续帧数3
TOTAL = 0 # 若连续帧数==3,则总眨眼次数+1
# 加载面部地标预测器
print("[INFO] 正在加载面部地标预测器...")
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(args["shape_predictor"])
# 分别获取左眼和右眼坐标索引
(lStart, lEnd) = FACIAL_LANDMARKS_68_IDXS["left_eye"]
(rStart, rEnd) = FACIAL_LANDMARKS_68_IDXS["right_eye"]
# 读取视频
print("[INFO] 开始视频流...")
vs = cv2.VideoCapture(args["video"])
time.sleep(1.0)
# 将shape对象转换为numpy数组
def shape_to_np(shape, dtype="int"):
# 创建一个dtype类型的空ndarray用于存储68个关键点的坐标
coords = np.zeros((shape.num_parts, 2), dtype=dtype)
# 遍历每个关键点,提取坐标并存储到ndarray中
for i in range(0, shape.num_parts):
coords[i] = (shape.part(i).x, shape.part(i).y)
return coords
# 不断循环处理每一帧图像
while True:
# 预处理
frame = vs.read()[1]
if frame is None:
break
# 调整图像大小
(h, w) = frame.shape[:2]
width = 1200 # 脸部大小会影响检测器的识别,太小可能会识别不到
r = width / float(w)
dim = (width, int(h * r))
frame = cv2.resize(frame, dim, interpolation=cv2.INTER_AREA)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# 人脸检测
rects = detector(gray, 0)
# 遍历检测到的每个人脸
for rect in rects:
# 获取关键点坐标
shape = predictor(gray, rect)
shape = shape_to_np(shape)
# 提取左眼和右眼区域坐标
leftEye = shape[lStart:lEnd]
rightEye = shape[rStart:rEnd]
# 计算左右眼纵横比EAR
leftEAR = eye_aspect_ratio(leftEye)
rightEAR = eye_aspect_ratio(rightEye)
# 计算平均纵横比
ear = (leftEAR + rightEAR) / 2.0
# 绘制眼睛区域轮廓(凸包)
leftEyeHull = cv2.convexHull(leftEye)
rightEyeHull = cv2.convexHull(rightEye)
cv2.drawContours(frame, [leftEyeHull], -1, (0, 255, 0), 1)
cv2.drawContours(frame, [rightEyeHull], -1, (0, 255, 0), 1)
# 检查是否满足EAR阈值
if ear < EYE_AR_THRESH:
COUNTER += 1
else:
# 如果连续几帧都是闭眼的,增加总数
if COUNTER >= EYE_AR_CONSEC_FRAMES:
TOTAL += 1
# 重置计数器
COUNTER = 0
# 在图像中,显示眨眼次数和纵横比
cv2.putText(frame, "Blinks: {}".format(TOTAL), (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
cv2.putText(frame, "EAR: {:.2f}".format(ear), (300, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
# 显示当前帧图像
cv2.imshow("Frame", frame)
key = cv2.waitKey(10) & 0xFF
# 按下Esc键退出循环
if key == 27:
break
vs.release() # 释放视频流
cv2.destroyAllWindows() # 关闭所有窗口
五、项目实战(摄像头获取帧图像)
- 【参数配置】方式一:Pycharm + Terminal + 输入指令自动检测:
python detect_blinks.py --shape-predictor shape_predictor_68_face_landmarks.dat
- 【参数配置】方式二:Pycharm + 点击Edit Configuration,输入配置参数
--shape-predictor shape_predictor_68_face_landmarks.dat
,点击Run开始检测。
# 导入所需库
from scipy.spatial import distance as dist
from collections import OrderedDict
import numpy as np
import argparse
import time
import dlib
import cv2
# 定义脸部关键点索引
FACIAL_LANDMARKS_68_IDXS = OrderedDict([
("mouth", (48, 68)),
("right_eyebrow", (17, 22)),
("left_eyebrow", (22, 27)),
("right_eye", (36, 42)),
("left_eye", (42, 48)),
("nose", (27, 36)),
("jaw", (0, 17))
])
# 计算眼睛纵横比函数
def eye_aspect_ratio(eye):
# 计算垂直距离
A = dist.euclidean(eye[1], eye[5])
B = dist.euclidean(eye[2], eye[4])
# 计算水平距离
C = dist.euclidean(eye[0], eye[3])
# 计算眼睛纵横比EAR
ear = (A + B) / (2.0 * C)
return ear
# 解析命令行参数
ap = argparse.ArgumentParser()
ap.add_argument("-p", "--shape-predictor", required=True, help="面部地标预测器路径")
args = vars(ap.parse_args())
# 设置EAR阈值和连续帧数
EYE_AR_THRESH = 0.3
EYE_AR_CONSEC_FRAMES = 3
# 初始化计数器
COUNTER = 0 # 计算连续帧数3
TOTAL = 0 # 若连续帧数==3,则总眨眼次数+1
# 加载面部地标预测器
print("[INFO] 正在加载面部地标预测器...")
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(args["shape_predictor"])
# 分别获取左眼和右眼坐标索引
(lStart, lEnd) = FACIAL_LANDMARKS_68_IDXS["left_eye"]
(rStart, rEnd) = FACIAL_LANDMARKS_68_IDXS["right_eye"]
# 读取视频
print("[INFO] 开始视频流...")
vs = cv2.VideoCapture(0)
time.sleep(1.0)
# 将shape对象转换为numpy数组
def shape_to_np(shape, dtype="int"):
# 创建一个dtype类型的空ndarray用于存储68个关键点的坐标
coords = np.zeros((shape.num_parts, 2), dtype=dtype)
# 遍历每个关键点,提取坐标并存储到ndarray中
for i in range(0, shape.num_parts):
coords[i] = (shape.part(i).x, shape.part(i).y)
return coords
# 不断循环处理每一帧图像
while True:
# 预处理
ret, frame = vs.read()
if not ret:
break
# 调整图像大小
(h, w) = frame.shape[:2]
width = 1200 # 脸部大小会影响检测器的识别,太小可能会识别不到
r = width / float(w)
dim = (width, int(h * r))
frame = cv2.resize(frame, dim, interpolation=cv2.INTER_AREA)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# 人脸检测
rects = detector(gray, 0)
# 遍历检测到的每个人脸
for rect in rects:
# 获取关键点坐标
shape = predictor(gray, rect)
shape = shape_to_np(shape)
# 提取左眼和右眼区域坐标
leftEye = shape[lStart:lEnd]
rightEye = shape[rStart:rEnd]
# 计算左右眼纵横比EAR
leftEAR = eye_aspect_ratio(leftEye)
rightEAR = eye_aspect_ratio(rightEye)
# 计算平均纵横比
ear = (leftEAR + rightEAR) / 2.0
# 绘制眼睛区域轮廓(凸包)
leftEyeHull = cv2.convexHull(leftEye)
rightEyeHull = cv2.convexHull(rightEye)
cv2.drawContours(frame, [leftEyeHull], -1, (0, 255, 0), 1)
cv2.drawContours(frame, [rightEyeHull], -1, (0, 255, 0), 1)
# 检查是否满足EAR阈值
if ear < EYE_AR_THRESH:
COUNTER += 1
else:
# 如果连续几帧都是闭眼的,增加总数
if COUNTER >= EYE_AR_CONSEC_FRAMES:
TOTAL += 1
# 重置计数器
COUNTER = 0
# 在图像中,显示眨眼次数和纵横比
cv2.putText(frame, "Blinks: {}".format(TOTAL), (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
cv2.putText(frame, "EAR: {:.2f}".format(ear), (300, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
# 显示当前帧图像
cv2.imshow("Frame", frame)
key = cv2.waitKey(1) & 0xFF
# 按下Esc键退出循环
if key == 27:
break
vs.release() # 释放视频流
cv2.destroyAllWindows() # 关闭所有窗口
风语者!平时喜欢研究各种技术,目前在从事后端开发工作,热爱生活、热爱工作。