使用opencv库实现口罩识别、人脸检测数据库

<>使用opencv库实现口罩识别、人脸检测数据库

之前做电子设计赛省赛，需要实现口罩识别，人脸检测，我们使用opencv自带库实现。最近无聊，重新在linux系统上完善一波，记录一下代码（待完善）。
文件结构如下：

* database文件夹:
用于存放图像数据，文件夹命名为id name，图像命名为数字.jpg

* xml文件夹
cv模型，github查查应该能找到

* data.py
主代码data.py如下： import cv2 import os import numpy as np from time import sleep
import time import shutil videoIn = cv2.VideoCapture("/dev/video0") ret,
np_frame= videoIn.read() #利用cv2内置的函数，初步识别人脸 face_cascade = cv2.CascadeClassifier
('xml/haarcascade_frontalface_default.xml') nose_cascade = cv2.CascadeClassifier
('xml/haarcascade_nose.xml') mouth_cascade = cv2.CascadeClassifier(
'xml/haarcascade_mouth.xml') eye_cascade = cv2.CascadeClassifier(
'xml/haarcascade_eye.xml') #利用cv2内置的人脸识别模块，可进行预训练和预测 face_recognizer = cv2.face.
LBPHFaceRecognizer_create() #全局使用 datapath = 'database' database = {'name':[],
'id':[]} #读取数据库 def Read_Database(): dirs = os.listdir(datapath) for dir in dirs
: id,name = dir.split()#文件夹命名格式为id name database['name'].append(name) database[
'id'].append(id) #用cv2内置的人脸检测模块，判断图片中是否有人。如果有，则将人脸部分和人脸坐标返回 def detect_face(img)
: gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) faces = face_cascade.
detectMultiScale(gray, scaleFactor=1.2, minNeighbors=5) if (len(faces) == 0):
return None, None (x, y, w, h) = faces[0] return gray[y:y + w, x:x + h], faces[0
] #进度条打印 def progress(per,width=50): text = ('\r[%%-%ds]'%width)%('#'*int(per*
width)) text += '%3s%%' text = text%(round(per*100)) print('\t'+text,end='')
#准备训练数据集：faces放置人脸图像，labels放置下表标签0、1、2、3... def prepare_training_data(
data_folder_path): dirs = os.listdir(data_folder_path) faces = [] labels = []
print('<------------------- Train Mode ------------------->') for dir_name in
dirs: if not dir_name == '.ipynb_checkpoints': id,name = dir_name.split() for i
in range(len(database['id'])): if database['id'][i] == id: label = int(i)
#训练的标签是id name在database中的下标 break subject_dir_path = data_folder_path + "/" +
dir_name subject_images_names= os.listdir(subject_dir_path) print('\n[Name] {0}
[Id] {1}'.format(name.ljust(5,' '),id.ljust(3,' '))) now_num = 0#本数据已经训练好的图片数量
for image_name in subject_images_names: if not image_name ==
'.ipynb_checkpoints': image_path = subject_dir_path + "/" + image_name image =
cv2.imread(image_path) face, rect = detect_face(image) if face is not None:
faces.append(face) labels.append(label) now_num += 1 #打印训练进度： per = now_num /
len(subject_images_names) progress(per) progress(1) print(
'\n\n<------------------- Complete ------------------->\n') return faces, labels
#画方框 def draw_rectangle(img, rect): (x, y, w, h) = rect cv2.rectangle(img, (x, y
), (x + w, y + h), (128, 128, 0), 2) #写文字 def draw_text(img, text, x, y): cv2.
putText(img, text, (x, y), cv2.FONT_HERSHEY_COMPLEX, 1, (128, 128, 0), 2) #图片预测
def predict(test_img): img = test_img.copy() face, rect = detect_face(img)
#得到人脸部分 #如果图像中检测到人脸 if not face is None: label = face_recognizer.predict(face)
if label[1]<100:#loss越小，可信度越高 label_text = database['name'][label[0]] else:
label_text= 'Unknown' draw_rectangle(img, rect) draw_text(img, label_text, rect[
0], rect[1] - 5) # print("found") else: draw_text(img, 'nofound',0,20) #
print("nofound") return img #新图像保存 def imgSave(img,num,id,name): newdir =
datapath+ '/' + id + ' ' +name#文件夹格式id name if not os.path.exists(newdir): os.
mkdir(newdir) face, rect = detect_face(img) if not face is None:#识别到脸部才进行录入 cv2.
imwrite(newdir+"/"+str(num)+".jpg", img[:,:,:]) num=num+1 return img, num
#显示当前数据库 def Dis_Database(): print('序号'.ljust(4,' ')+'id'.ljust(6,' ')+'name'.
ljust(6,' ')) for i in range(len(database['name'])): id = database['id'][i] name
= database['name'][i] print(str(i).ljust(6,' ')+id.ljust(6,' ')+name.ljust(6,' '
))
####################################################################################################
stage= 0 Read_Database() faces, labels = prepare_training_data(datapath)#初始数据集
if (len(database['name']) != 0): face_recognizer.train(faces, np.array(labels))
#训练 while True: # 检测按键，停止训练，保存当前模型和最佳模型 #0正常显示;1口罩识别;2人脸检测;3数据录入 k = cv2.waitKey
(1) & 0xFF if k == ord('0'): stage = 0; elif k == ord('1'): stage = 1; elif k ==
ord('2'): stage = 2; elif k == ord('3'): stage = 3; if stage == 0: ret, np_frame
= videoIn.read() cv2.imshow("capture", np_frame) elif stage == 1: ret, np_frame
= videoIn.read() gray = cv2.cvtColor(np_frame, cv2.COLOR_BGR2GRAY) noses =
nose_cascade.detectMultiScale(gray) if(len(noses)>=1): draw_text(np_frame,
'Nomask',0,20) else: eyes = eye_cascade.detectMultiScale(gray) if(len(eyes)>=1):
draw_text(np_frame, 'Mask',0,20) cv2.imshow("capture", np_frame) elif stage ==
2: ret, np_frame = videoIn.read() if (len(database['name'])!=0): np_frame =
predict(np_frame) else: draw_text(np_frame, 'Database empty',0,20) cv2.imshow(
"capture", np_frame) elif stage == 3: choice = 0 print(
'<------------------------------- 数据库修改 ------------------------------->')
choice= int(input('功能选项(返回--0 录入数据--1 删除数据--2 添加数据--3):')) #功能实现 #直接返回 if choice
== 0: pass #选择录入数据 elif choice == 1: name = input('请输入姓名：') id = input('请输入id号：'
) number = int(input('请输入录入图片数量：')) num = 1 database['name'].append(name)
database['id'].append(id) while True: ret, np_frame = videoIn.read() cv2.imshow(
"capture", np_frame) np_frame, num = imgSave(np_frame, num , id ,name) draw_text
(np_frame, 'num:'+str(num),0,20) if num > number: faces, labels =
prepare_training_data(datapath) face_recognizer.train(faces, np.array(labels))
break if cv2.waitKey(1) & 0xFF == ord('q'):#强制退出 break print('数据录入成功！') #选择删除数据
elif choice == 2: Dis_Database() index = int(input('请输入想删除的数据的序号：')) while(index
> len(database['id'])-1 or index < 0): index = int(input('输入序号不存在，请重新输入：'))
del_path= datapath + '/' + database['id'][index] + ' ' + database['name'][index]
if not os.path.exists(del_path): print('文件夹不存在') else: shutil.rmtree(del_path)
del database['name'][index] del database['id'][index] if len(database['name'])
!= 0:#数据库非空则训练 faces, labels = prepare_training_data(datapath) face_recognizer.
train(faces, np.array(labels)) print('数据删除成功！') #添加数据 elif choice == 3:
Dis_Database() index = int(input('请输入想添加的数据的序号：')) number = int(input(
'请输入想添加的图片的数量：')) while(index > len(database['id'])-1 or index < 0): index = int
(input('输入序号不存在，请重新输入：')) add_path = datapath + '/' + database['id'][index] + '
' + database['name'][index] old_len = len(os.listdir(add_path)) num = old_len +
1 id = database['id'][index] name = database['name'][index] while True: ret,
np_frame= videoIn.read() cv2.imshow("capture", np_frame) np_frame, num = imgSave
(np_frame, num , id ,name) draw_text(np_frame, 'num:'+str(num),0,20) if num >
old_len+ number: faces, labels = prepare_training_data(datapath) face_recognizer
.train(faces, np.array(labels)) break if cv2.waitKey(1) & 0xFF == ord('q'):#强制退出
break stage = 0 print('<------------------------------- 返回界面
------------------------------->') print('\n')
运行后，在非终端处：

* 按1开始口罩识别
* 按2开始人脸检测
* 按3开始人脸检测数据库修改：修改时要在终端输入，可实现数据录入、删除、添加
* 按0回到正常显示界面。
人脸检测效果

pynq实现如下
import cv2 import os import numpy as np from time import sleep import time
import shutil from pynq.overlays.base import BaseOverlay from pynq.lib.video
import * # 显示器 base = BaseOverlay("base.bit") # monitor configuration: 640*480
@ 60Hz Mode = VideoMode(640,480,24) hdmi_out = base.video.hdmi_out
hdmi_out.configure(Mode,PIXEL_BGR) hdmi_out.start() screen_w=320 screen_h=240
frame_out_w= 320 frame_out_h = 240 # 摄像头 frame_in_w = 320 frame_in_h = 240
videoIn= cv2.VideoCapture(0) videoIn.set(cv2.CAP_PROP_FRAME_WIDTH, frame_in_w);
videoIn.set(cv2.CAP_PROP_FRAME_HEIGHT, frame_in_h); print("Capture device is
open: " + str(videoIn.isOpened())) #利用cv2内置的函数，初步识别人脸 face_cascade =
cv2.CascadeClassifier('xml/haarcascade_frontalface_default.xml') nose_cascade =
cv2.CascadeClassifier('xml/haarcascade_nose.xml') mouth_cascade =
cv2.CascadeClassifier('xml/haarcascade_mouth.xml') eye_cascade =
cv2.CascadeClassifier('xml/haarcascade_eye.xml') #利用cv2内置的人脸识别模块，可进行预训练和预测
face_recognizer= cv2.face.createLBPHFaceRecognizer() #全局使用 datapath = 'database'
database= {'name':[],'id':[]} #读取数据库 def Read_Database(): dirs = os.listdir(
datapath) for dir in dirs: if not dir == '.ipynb_checkpoints': id,name =
dir.split()#文件夹命名格式为id name database['name'].append(name) database['id'].append(
id) #用cv2内置的人脸检测模块，判断图片中是否有人。如果有，则将人脸部分和人脸坐标返回 def detect_face(img): gray =
cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) faces = face_cascade.detectMultiScale(
gray,scaleFactor=1.2, minNeighbors=5) if (len(faces) == 0): return None, None (
x, y, w, h) = faces[0] return gray[y:y + w, x:x + h], faces[0] #进度条打印 def
progress(per,width=50): text = ('\r[%%-%ds]'%width)%('#'*int(per*width)) text +=
'%3s%%' text = text%(round(per*100)) print('\t'+text,end='')
#准备训练数据集：faces放置人脸图像，labels放置下表标签0、1、2、3... def prepare_training_data(
data_folder_path): dirs = os.listdir(data_folder_path) faces = [] labels = []
print('<------------------- Train Mode ------------------->') for dir_name in
dirs:if dir_name == '.ipynb_checkpoints': continue id,name = dir_name.split()
for i in range(len(database['id'])): if database['id'][i] == id: label = int(i)
#训练的标签是id name在database中的下标 break subject_dir_path = data_folder_path + "/" +
dir_name subject_images_names= os.listdir(subject_dir_path) print('\n[Name] {0}
[Id] {1}'.format(name.ljust(5,' '),id.ljust(3,' '))) now_num = 0#本数据已经训练好的图片数量
for image_name in subject_images_names: if not image_name ==
'.ipynb_checkpoints': image_path = subject_dir_path + "/" + image_name image =
cv2.imread(image_path) face, rect = detect_face(image) if face is not None:
faces.append(face) labels.append(label) now_num += 1 #打印训练进度： per = now_num /
len(subject_images_names) progress(per) progress(1) print('\n\n
<------------------- Complete ------------------->\n') return faces, labels #画方框
def draw_rectangle(img, rect): (x, y, w, h) = rect cv2.rectangle(img, (x, y), (
x + w, y + h), (128, 128, 0), 2) #写文字 def draw_text(img, text, x, y):
cv2.putText(img, text, (x, y), cv2.FONT_HERSHEY_COMPLEX, 1, (128, 128, 0), 2)
#图片预测 def predict(test_img): img = test_img.copy() face, rect = detect_face(img)
#得到人脸部分 #如果图像中检测到人脸 if not face is None: label = face_recognizer.predict(face)
if label[1]<100:#loss越小，可信度越高 label_text = database['name'][label[0]]
base.rgbleds[5].write(2) else: label_text = 'Unknown' base.rgbleds[5].write(4)
draw_rectangle(img, rect) draw_text(img, label_text, rect[0], rect[1] - 5)
else: draw_text(img, 'nofound',0,20) base.rgbleds[5].write(0) return img #新图像保存
def imgSave(img,num,id,name): newdir = datapath + '/' + id + ' ' +name#文件夹格式id
name if not os.path.exists(newdir): os.mkdir(newdir) face, rect = detect_face(
img) if not face is None:#识别到脸部才进行录入 cv2.imwrite(newdir+"/"+str(num)+".jpg", img
[:,:,:]) num=num+1 return img, num #显示当前数据库 def Dis_Database(): print('序号'.ljust
(4,' ')+'id'.ljust(6,' ')+'name'.ljust(6,' ')) for i in range(len(database[
'name'])): id = database['id'][i] name = database['name'][i] print(str(i).ljust(
6,' ')+id.ljust(6,' ')+name.ljust(6,' '))
####################################################################################################
stage= 0 Read_Database() faces, labels = prepare_training_data(datapath)#初始数据集
if (len(database['name']) != 0): face_recognizer.train(faces, np.array(labels))
#训练 while True: # 检测按键，停止训练，保存当前模型和最佳模型 # 0正常显示;1口罩识别;2人脸检测;3数据录入 ret, np_frame
= videoIn.read() if base.buttons[0].read()==1: stage=0 if base.buttons[1].read()
==1: stage=1 if base.buttons[2].read()==1: stage=2 if base.buttons[3].read()==1:
stage=3 if stage == 0: base.rgbleds[4].write(0) base.rgbleds[5].write(0) elif
stage== 1: gray = cv2.cvtColor(np_frame, cv2.COLOR_BGR2GRAY) noses =
nose_cascade.detectMultiScale(gray) if(len(noses)>=1): draw_text(np_frame,
'Nomask',0,20) base.rgbleds[4].write(4) else: eyes =
eye_cascade.detectMultiScale(gray) if(len(eyes)>=1): draw_text(np_frame, 'Mask'
,0,20) base.rgbleds[4].write(2) else: base.rgbleds[4].write(0) elif stage == 2:
if (len(database['name'])!=0): np_frame = predict(np_frame) else: draw_text(
np_frame,'Database empty',0,20) elif stage == 3: choice = 0 print(
'<------------------------------- 数据库修改 ------------------------------->')
choice= int(input('功能选项(返回--0 录入数据--1 删除数据--2 添加数据--3):')) #功能实现 #直接返回 if choice
== 0: pass #选择录入数据 elif choice == 1: Dis_Database() name = input('请输入姓名：') id =
input('请输入id号：') number = int(input('请输入录入图片数量：')) num = 1 database['name']
.append(name) database['id'].append(id) while True: ret, np_frame = videoIn.read
() outframe = hdmi_out.newframe() outframe[0:screen_h,0:screen_w,:] = np_frame[0
:screen_h,0:screen_w,:] hdmi_out.writeframe(outframe) np_frame, num = imgSave(
np_frame, num ,id ,name) draw_text(np_frame, 'num:'+str(num),0,20) progress((num
-1)/number) if num > number: progress(1) print('\n') faces, labels =
prepare_training_data(datapath) face_recognizer.train(faces, np.array(labels))
break # if base.buttons[0].read()==0:#强制退出 # break print('数据录入成功！') #选择删除数据 elif
choice== 2: Dis_Database() index = int(input('请输入想删除的数据的序号：')) while(index >
len(database['id'])-1 or index < 0): index = int(input('输入序号不存在，请重新输入：'))
del_path= datapath + '/' + database['id'][index] + ' ' + database['name'][index]
if not os.path.exists(del_path): print('文件夹不存在') else: shutil.rmtree(del_path)
del database['name'][index] del database['id'][index] if len(database['name'])
!= 0:#数据库非空则训练 faces, labels = prepare_training_data(datapath)
face_recognizer.train(faces, np.array(labels)) print('数据删除成功！') #添加数据 elif
choice== 3: Dis_Database() index = int(input('请输入想添加的数据的序号：')) number = int(
input('请输入想添加的图片的数量：')) while(index > len(database['id'])-1 or index < 0): index
= int(input('输入序号不存在，请重新输入：')) add_path = datapath + '/' + database['id'][index]
+' ' + database['name'][index] old_len = len(os.listdir(add_path)) num =
old_len +1 id = database['id'][index] name = database['name'][index] print(
add_path,old_len,id,name) while True: ret, np_frame = videoIn.read() outframe =
hdmi_out.newframe() outframe[0:screen_h,0:screen_w,:] = np_frame[0
:screen_h,0:screen_w,:] hdmi_out.writeframe(outframe) np_frame, num = imgSave(
np_frame, num ,id ,name) draw_text(np_frame, 'num:'+str(num),0,20) progress((num
-old_len-1)/number) if num > old_len + number: progress(1) print('\n') faces,
labels= prepare_training_data(datapath) face_recognizer.train(faces, np.array(
labels)) break # if base.buttons[0].read()==0:#强制退出 # break stage = 0 print(
'<------------------------------- 返回界面 ------------------------------->') print(
'\n') outframe = hdmi_out.newframe() outframe[0:screen_h,0:screen_w,:] =
np_frame[0:screen_h,0:screen_w,:] hdmi_out.writeframe(outframe)

• « 上一篇：范式的归纳整理（1NF、2NF、3NF、BCNF）
• » 下一篇：第一章 无名小生