def __init__(self):
# 加载模型
self.emotion_classifier = load_model('models/emotion_model.hdf5')
self.face_cascade = cv2.CascadeClassifier(
'models/haarcascade_frontalface_default.xml')
facial_recognition_model_path = 'models/new_face_recognition_hog.pickle'
self.faceutil = FaceUtil(facial_recognition_model_path)
class Stranger:
def __init__(self):
facial_recognition_model_path = 'models/new_face_recognition_hog.pickle'
self.faceutil = FaceUtil(facial_recognition_model_path)
def facedetect(self, picpath, picname, timestamp):
pic = cv2.imread(picpath, )
frame = cv2.flip(pic, 1)
frame = imutils.resize(frame, width=600)
face_location_list, names = self.faceutil.get_face_location_and_name(
frame)
for ((left, top, right, bottom), name) in zip(face_location_list,
names):
# display label and bounding box rectangle on the output frame
if 'Unknown' in names:
print('有陌生人出没!')
event = {
'event_type':
2,
'event_date':
time.strftime("%Y-%m-%d %H:%M:%S",
time.localtime(timestamp)),
'event_location':
'BJTU',
'event_desc':
'陌生人来了',
'oldperson_id':
0,
'pic_name':
picname
}
return event
def detectID(self, picpath):
pic = cv2.imread(picpath, )
frame = cv2.flip(pic, 1)
frame = imutils.resize(frame, width=600)
face_location_list, names = self.faceutil.get_face_location_and_name(
frame)
dict = None
if len(face_location_list) == 1:
if 'old_1' in names:
dict = 0
if 'old_2' in names:
dict = 1
return dict
class Track:
def __init__(self):
facial_recognition_model_path = 'models/new_face_recognition_hog.pickle'
self.faceutil = FaceUtil(facial_recognition_model_path)
self.turnArea = (140, 120, 500, 370)
def test_track(self, picpath, picname, timestamp):
pic = cv2.imread(picpath)
frame = cv2.flip(pic, 1)
frame = imutils.resize(frame, width=600)
face_location_list, names = self.faceutil.get_face_location_and_name(
frame)
if 'Unknown' in names:
(left, top, right, bottom) = face_location_list[0]
if left < self.turnArea[0]:
print("摄像头向右转")
event = {
'event_type':
4,
'event_date':
time.strftime("%Y-%m-%d %H:%M:%S",
time.localtime(timestamp)),
'event_location':
'在庭院',
'event_desc':
'摄像头向右转',
'oldperson_id':
0,
'pic_name':
picname
}
return event
elif right > self.turnArea[3]:
print("摄像头向左转")
event = {
'event_type':
4,
'event_date':
time.strftime("%Y-%m-%d %H:%M:%S",
time.localtime(timestamp)),
'event_location':
'在庭院',
'event_desc':
'摄像头向左转',
'oldperson_id':
0,
'pic_name':
picname
}
return event
else:
print("能正常检测到陌生人")
class Interact:
def __init__(self):
facial_recognition_model_path = 'models/new_face_recognition_hog.pickle'
self.faceutil = FaceUtil(facial_recognition_model_path)
def test_distance(self, picpath, picname, timestamp):
pic = cv2.imread(picpath)
frame = cv2.flip(pic, 1)
frame = imutils.resize(frame, width=600)
face_location_list, names = self.faceutil.get_face_location_and_name(
frame)
#if ('volunteer', 'old_1') in names or ('volunteer', 'old_2') in names:
if len(face_location_list) == 2:
i = 0
(x1, y1) = (0, 0)
(x2, y2) = (0, 0)
for (left, top, right, bottom) in face_location_list:
i = i + 1
if i == 1:
(x1, y1) = ((left + right) / 2, (top + bottom) / 2)
if i == 2:
(x2, y2) = ((left + right) / 2, (top + bottom) / 2)
pixel_distance = sqrt((x1 - x2)**2 + (y1 - y2)**2)
measure = 20 / abs(face_location_list[0][0] -
face_location_list[0][2])
actual_distance = pixel_distance * measure
print("actual_distance", actual_distance)
if actual_distance <= 50:
event = {
'event_type':
1,
'event_date':
time.strftime("%Y-%m-%d %H:%M:%S",
time.localtime(timestamp)),
'event_location':
'BJTU',
'event_desc':
'有义工和老人交互!',
'oldperson_id':
0,
'pic_name':
picname
}
print('他俩唠嗑了')
return event
class Forbidden:
def __init__(self):
facial_recognition_model_path = 'models/new_face_recognition_hog.pickle'
self.faceutil = FaceUtil(facial_recognition_model_path)
self.forbidden = (190, 140, 430, 340)
print(self.forbidden)
def ifInFobidden(self, left, top, right, bottom, picname, timestamp):
# 不在安全区域内
if not (left > self.forbidden[0] and top > self.forbidden[1]
and right < self.forbidden[2] and bottom < self.forbidden[3]):
print("有人闯入安全区域内")
event = {
'event_type':
4,
'event_date':
time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(timestamp)),
'event_location':
'BJTU',
'event_desc':
'有人闯入安全区域内',
'oldperson_id':
0,
'pic_name':
picname
}
return event
def testForbidden(self, picpath, picname, timestamp):
pic = cv2.imread(picpath)
frame = cv2.flip(pic, 1)
frame = imutils.resize(frame, width=640)
frame = imutils.resize(frame, height=480)
face_location_list, names = self.faceutil.get_face_location_and_name(
frame)
for ((left, top, right, bottom), name) in zip(face_location_list,
names):
self.ifInFobidden(left, top, right, bottom, picname, timestamp)
# 全局常量
FACIAL_EXPRESSION_TARGET_WIDTH = 28
FACIAL_EXPRESSION_TARGET_HEIGHT = 28
# load the face detector cascade and smile detector CNN
model = load_model(model_path)
# if a video path was not supplied, grab the reference to the webcam
if not input_video:
camera = cv2.VideoCapture(0)
time.sleep(2)
else:
camera = cv2.VideoCapture(input_video)
faceutil = FaceUtil()
# keep looping
while True:
# grab the current frame
(grabbed, frame) = camera.read()
# if we are viewing a video and we did not grab a frame, then we
# have reached the end of the video
if input_video and not grabbed:
break
if not input_video:
frame = cv2.flip(frame, 1)
# resize the frame, convert it to grayscale, and then clone the
ap.add_argument("-f", "--filename", required=False, default='', help="")
args = vars(ap.parse_args())
# 全局变量
facial_recognition_model_path = 'models/new_face_recognition_hog.pickle'
input_video = args['filename']
# 初始化摄像头
if not input_video:
vs = cv2.VideoCapture(0)
time.sleep(2)
else:
vs = cv2.VideoCapture(input_video)
# 初始化人脸识别模型
faceutil = FaceUtil(facial_recognition_model_path)
# 不断循环
while True:
# grab the current frame
(grabbed, frame) = vs.read()
# if we are viewing a video and we did not grab a frame, then we
# have reached the end of the video
if input_video and not grabbed:
break
if not input_video:
frame = cv2.flip(frame, 1)
# resize the frame, convert it to grayscale, and then clone the
# original frame so we can draw on it later in the program
frame = imutils.resize(frame, width=600)
face_location_list, names = faceutil.get_face_location_and_name(frame)
]
action_map = {
'blink': '请眨眼',
'open_mouth': '请张嘴',
'smile': '请笑一笑',
'rise_head': '请抬头',
'bow_head': '请低头',
'look_left': '请看左边',
'look_right': '请看右边'
}
# 设置摄像头
cam = cv2.VideoCapture(0)
cam.set(3, 640) # set video widht
cam.set(4, 480) # set video height
faceutil = FaceUtil()
counter = 0
while True:
counter += 1
_, image = cam.read()
if counter <= 10: # 放弃前10帧
continue
image = cv2.flip(image, 1)
if error == 1:
end_time = time.time()
difference = end_time - start_time
print(difference)
if difference >= limit_time:
error = 0
'''
训练人脸识别模型
'''
from imutils import paths
from oldcare.facial import FaceUtil
# global variable
dataset_path = 'dataset/'
output_encoding_file_path = 'models/new_face_recognition_hog.pickle'
# grab the paths to the input images in our dataset
print("[INFO] quantifying faces...")
image_paths = list(paths.list_images(dataset_path))
if len(image_paths) == 0:
print('[ERROR] no images to train.')
else:
faceutil = FaceUtil()
print("[INFO] training face embeddings...")
faceutil.save_embeddings(image_paths, output_encoding_file_path)
ANGLE = 20
ACTUAL_DISTANCE_LIMIT = 100 # cm
# 得到 ID->姓名的map 、 ID->职位类型的map
id_card_to_name, id_card_to_type = fileassistant.get_people_info(
people_info_path)
# 初始化摄像头
if not input_video:
vs = cv2.VideoCapture(0)
time.sleep(2)
else:
vs = cv2.VideoCapture(input_video)
# 加载模型
faceutil = FaceUtil(model_path)
print('[INFO] 开始检测义工和老人是否有互动...')
# 不断循环
counter = 0
while True:
counter += 1
camera_turned = 0
# grab the current frame
(grabbed, frame) = vs.read()
# if we are viewing a video and we did not grab a frame, then we
# have reached the end of the video
if input_video and not grabbed:
break
def __init__(self):
facial_recognition_model_path = 'models/new_face_recognition_hog.pickle'
self.faceutil = FaceUtil(facial_recognition_model_path)
self.turnArea = (140, 120, 500, 370)
def __init__(self):
facial_recognition_model_path = 'models/new_face_recognition_hog.pickle'
self.faceutil = FaceUtil(facial_recognition_model_path)
class EmotionEstimate:
def __init__(self):
# 加载模型
self.emotion_classifier = load_model('models/emotion_model.hdf5')
self.face_cascade = cv2.CascadeClassifier(
'models/haarcascade_frontalface_default.xml')
facial_recognition_model_path = 'models/new_face_recognition_hog.pickle'
self.faceutil = FaceUtil(facial_recognition_model_path)
# 图片预处理
def preprocess_input(self, x, v2=True):
x = x.astype('float32')
x = x / 255.0
if v2:
x = x - 0.5
x = x * 2.0
return x
def apply_offsets(self, face_coordinates, offsets):
x, y, width, height = face_coordinates
x_off, y_off = offsets
return x - x_off, x + width + x_off, y - y_off, y + height + y_off
def predict_emotion(self, id, test_img_path, picname, timestamp):
pic = cv2.imread(test_img_path)
frame = cv2.flip(pic, 1)
frame = imutils.resize(frame, width=600)
face_location_list, names = self.faceutil.get_face_location_and_name(
frame)
if len(face_location_list) == 1:
# 情感类型字典
emotion_labels = {
0: 'angry',
1: 'disgust',
2: 'fear',
3: 'happy',
4: 'sad',
5: 'surprise',
6: 'neutral'
}
emotion_offsets = (20, 40)
emotion_target_size = self.emotion_classifier.input_shape[1:3]
# 读取原图片
bgr_image = cv2.imread(test_img_path)
# 转化为灰度图
gray_image = cv2.cvtColor(bgr_image, cv2.COLOR_BGR2GRAY)
# 获得人脸坐标列表
faces = self.face_cascade.detectMultiScale(
gray_image,
scaleFactor=1.1,
minNeighbors=5,
minSize=(30, 30),
flags=cv2.CASCADE_SCALE_IMAGE)
# 获得人脸坐标
for face_coordinates in faces:
x1, x2, y1, y2 = self.apply_offsets(face_coordinates,
emotion_offsets)
gray_face = gray_image[y1:y2, x1:x2]
try:
gray_face = cv2.resize(gray_face, emotion_target_size)
except:
continue
gray_face = self.preprocess_input(gray_face, True)
gray_face = np.expand_dims(gray_face, 0)
gray_face = np.expand_dims(gray_face, -1)
# 获取模型的预测结果
emotion_prediction = self.emotion_classifier.predict(gray_face)
emotion_label_arg = np.argmax(emotion_prediction)
emotion_text = emotion_labels[emotion_label_arg]
print(emotion_text)
# 编写事件描述和事件
desc = str(emotion_prediction[0][0]) + ',' + str(emotion_prediction[0][1]) + ',' + str(emotion_prediction[0][2]) + ',' + \
str(emotion_prediction[0][3]) + ',' + str(emotion_prediction[0][5]) + ',' + str(emotion_prediction[0][6])
event = {
'event_type':
0,
'event_date':
time.strftime("%Y-%m-%d %H:%M:%S",
time.localtime(timestamp)),
'event_location':
'BJTU',
'event_desc':
desc,
'oldperson_id':
id,
'pic_name':
picname
}
# 返回事件
return event
def __init__(self):
self.model = self.model_loader()
facial_recognition_model_path = 'models/face_recognition_hog.pickle'
self.faceutil = FaceUtil(facial_recognition_model_path)
def push_frame():
# 推流函数
facial_recognition_model_path = 'info/face_recognition_hog.pickle' # jian ce ren lian
facial_expression_model_path = 'models/face_expression_seven_class.hdf5' # fenxi qinggan
output_stranger_path = 'supervision/strangers'
output_smile_path = 'supervision/smile'
people_info_path = 'info/people_info.csv'
facial_expression_info_path = 'info/facial_expression_info_seven_class.csv'
faceutil = FaceUtil(facial_recognition_model_path)
facial_expression_model = load_model(facial_expression_model_path,
compile=False)
# 初始化人脸识别模型
# your python path
python_path = '/root/anaconda3/envs/tensorflow/bin/python'
# 全局常量
FACIAL_EXPRESSION_TARGET_WIDTH = 48
FACIAL_EXPRESSION_TARGET_HEIGHT = 48
ANGLE = 20
# 得到 ID->姓名的map 、 ID->职位类型的map、
# 摄像头ID->摄像头名字的map、表情ID->表情名字的map
id_card_to_name, id_card_to_type = fileassistant.get_people_info(
people_info_path)
facial_expression_id_to_name = fileassistant.get_facial_expression_info(
facial_expression_info_path)
# 控制陌生人检测
strangers_timing = 0 # 计时开始
strangers_start_time = 0 # 开始时间
strangers_limit_time = 2 # if >= 2 seconds, then he/she is a stranger.
# 控制微笑检测
facial_expression_timing = 0 # 计时开始
facial_expression_start_time = 0 # 开始时间
facial_expression_limit_time = 2 # if >= 2 seconds, he/she is smiling
accum_time = 0
curr_fps = 0
fps = "FPS: ??"
# prev_time = time()
# 防止多线程时 command 未被设置
while True:
print('command lenth', len(command))
if len(command) > 0:
# 管道配置，其中用到管道
p = sp.Popen(command, stdin=sp.PIPE)
break
while True:
faceutil = FaceUtil(facial_recognition_model_path)
facial_expression_model = load_model(facial_expression_model_path,
compile=False)
if frame_queue.empty() != True:
counter = 0
while True:
counter += 1
image = frame_queue.get()
image = cv2.flip(image, 1)
image = imutils.resize(image,
width=VIDEO_WIDTH,
height=VIDEO_HEIGHT) # 压缩，加快识别速度
# if counter%10!=0:
# cv2.imshow("Checking Strangers and Ole People's Face Expression",
# image)
#
# # Press 'ESC' for exiting video
# k = cv2.waitKey(1) & 0xff
# continue
gray = cv2.cvtColor(image,
cv2.COLOR_BGR2GRAY) # grayscale，表情识别
# if True:
# cv2.imshow("Checking Strangers and Ole People's Face Expression",
# gray)
# continue
face_location_list, names = faceutil.get_face_location_and_name(
image)
# 得到画面的四分之一位置和四分之三位置，并垂直划线
one_fourth_image_center = (int(VIDEO_WIDTH / 4),
int(VIDEO_HEIGHT / 4))
three_fourth_image_center = (int(VIDEO_WIDTH / 4 * 3),
int(VIDEO_HEIGHT / 4 * 3))
cv2.line(image, (one_fourth_image_center[0], 0),
(one_fourth_image_center[0], VIDEO_HEIGHT),
(0, 255, 255), 1)
cv2.line(image, (three_fourth_image_center[0], 0),
(three_fourth_image_center[0], VIDEO_HEIGHT),
(0, 255, 255), 1)
# 处理每一张识别到的人脸
for ((left, top, right, bottom),
name) in zip(face_location_list, names):
# 将人脸框出来
rectangle_color = (0, 0, 255)
if id_card_to_type[name] == 'old_people':
rectangle_color = (0, 0, 128)
elif id_card_to_type[name] == 'employee':
rectangle_color = (255, 0, 0)
elif id_card_to_type[name] == 'volunteer':
rectangle_color = (0, 255, 0)
else:
pass
cv2.rectangle(image, (left, top), (right, bottom),
rectangle_color, 2)
# 陌生人检测逻辑
if 'Unknown' in names: # alert
if strangers_timing == 0: # just start timing
strangers_timing = 1
strangers_start_time = time.time()
else: # already started timing
strangers_end_time = time.time()
difference = strangers_end_time - strangers_start_time
current_time = time.strftime(
'%Y-%m-%d %H:%M:%S',
time.localtime(time.time()))
if difference < strangers_limit_time:
print('[INFO] %s, 房间, 陌生人仅出现 %.1f 秒. 忽略.' %
(current_time, difference))
else: # strangers appear
event_desc = '陌生人出现!!!'
event_location = '房间'
print('[EVENT] %s, 房间, 陌生人出现!!!' %
(current_time))
cv2.imwrite(
os.path.join(
output_stranger_path,
'snapshot_%s.jpg' %
(time.strftime('%Y%m%d_%H%M%S'))),
image) # snapshot
pic_path = os.path.join(
output_stranger_path, 'snapshot_%s.jpg' %
(time.strftime('%Y%m%d_%H%M%S')))
# insert into database
command1 = '%s inserting.py --event_desc %s --event_type 2 --event_location %s --pic_path %s' % (
python_path, event_desc, event_location,
pic_path)
p2 = subprocess.Popen(command1, shell=True)
# 开始陌生人追踪
unknown_face_center = (int(
(right + left) / 2), int(
(top + bottom) / 2))
cv2.circle(image, (unknown_face_center[0],
unknown_face_center[1]), 4,
(0, 255, 0), -1)
direction = ''
# face locates too left, servo need to turn right,
# so that face turn right as well
if unknown_face_center[
0] < one_fourth_image_center[0]:
direction = 'right'
elif unknown_face_center[
0] > three_fourth_image_center[0]:
direction = 'left'
# adjust to servo
if direction:
print('%d-摄像头需要 turn %s %d 度' %
(counter, direction, ANGLE))
else: # everything is ok
strangers_timing = 0
# 表情检测逻辑
# 如果不是陌生人，且对象是老人
if name != 'Unknown' and id_card_to_type[
name] == 'old_people':
# 表情检测逻辑
roi = gray[top:bottom, left:right]
roi = cv2.resize(roi,
(FACIAL_EXPRESSION_TARGET_WIDTH,
FACIAL_EXPRESSION_TARGET_HEIGHT))
roi = roi.astype("float") / 255.0
roi = img_to_array(roi)
roi = np.expand_dims(roi, axis=0)
# determine facial expression
emotions = [
'Angry', 'Disgust', 'Fear', 'Happy', 'Neutral',
'Sad', 'Surprise'
]
emotion_value_list = facial_expression_model.predict(
roi)[0]
facial_expression_label = emotions[np.argmax(
emotion_value_list)]
if facial_expression_label == 'Happy': # alert
if facial_expression_timing == 0: # just start timing
facial_expression_timing = 1
facial_expression_start_time = time.time()
else: # already started timing
facial_expression_end_time = time.time()
difference = facial_expression_end_time - facial_expression_start_time
current_time = time.strftime(
'%Y-%m-%d %H:%M:%S',
time.localtime(time.time()))
if difference < facial_expression_limit_time:
print('[INFO] %s, 房间, %s仅笑了 %.1f 秒. 忽略.' %
(current_time, id_card_to_name[name],
difference))
else: # he/she is really smiling
event_desc = '%s正在笑' % (
id_card_to_name[name])
event_location = '房间'
print(
'[EVENT] %s, 房间, %s正在笑.' %
(current_time, id_card_to_name[name]))
pic_path = os.path.join(
output_smile_path, 'snapshot_%s.jpg' %
(time.strftime('%Y%m%d_%H%M%S')))
cv2.imwrite(pic_path, image) # snapshot
# insert into database
command1 = '%s inserting.py --event_desc %s --event_type 0 --event_location %s --old_people_id %d --pic_path %s' % (
python_path, event_desc,
event_location, int(name), pic_path)
p2 = subprocess.Popen(command1, shell=True)
else: # everything is ok
facial_expression_timing = 0
else: # 如果是陌生人，则不检测表情
facial_expression_label = ''
# 人脸识别和表情识别都结束后，把表情和人名写上
# (同时处理中文显示问题)
img_PIL = Image.fromarray(
cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
draw = ImageDraw.Draw(img_PIL)
final_label = id_card_to_name[
name] + ': ' + facial_expression_id_to_name[
facial_expression_label] if facial_expression_label else id_card_to_name[
name]
draw.text((left, top - 30),
final_label,
font=ImageFont.truetype('NotoSansCJK-Black.ttc',
40),
fill=(255, 0, 0)) # linux
# 转换回OpenCV格式
image = cv2.cvtColor(np.asarray(img_PIL),
cv2.COLOR_RGB2BGR)
# show our detected faces along with smiling/not smiling labels
p.stdin.write(image.tostring())
cv2.imshow(
"Checking Strangers and Ole People's Face Expression",
image)
# Press 'ESC' for exiting video
k = cv2.waitKey(1) & 0xff
if k == 27:
break
def __init__(self):
facial_recognition_model_path = 'models/new_face_recognition_hog.pickle'
self.faceutil = FaceUtil(facial_recognition_model_path)
self.forbidden = (190, 140, 430, 340)
print(self.forbidden)
