Python 运行时出错 opencv（闪烁检测）

Question

我使用Visual Studio 2019，（Python3.6 / 64bit）

这是一个使用 opencv 和 dlib 检测眨眼和张嘴的代码。

Traceback (most recent call last):
  File "C:\Users\이시우\source\repos\planners facedetect\test1.py", line 132, in <module>
    mouths = get_mouth_pen_ratio( mouth_points, landmarks)
NameError: name 'landmarks' is not defined
[ WARN:0] global D:\bld\libopencv_1632857399304\work\modules\videoio\src\cap_msmf.cpp (438) `anonymous-namespace'::SourceReaderCB::~SourceReaderCB terminating async callback

转码时出现上述错误。

起初，它运行良好，但突然出现此错误。 有什么问题？

 for face in faces:
        landmarks = predictor(gray, face)     

我一边执行语句一边找问题，发现上面的代码跳过了for语句中的语句。

以下是我编写的代码的全文。

在while语句中，我们通过if、elif语句测量了睁眼和闭眼时的眼睛比例，并通过中间值写入检测眨眼。

import cv2
import dlib
from math import hypot
import timeit
import winsound as sd

def beepsound():
    fr = 2000    
    du = 1000    
    sd.Beep(fr, du)


detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor("./shape_predictor_68_face_landmarks.dat")

font = cv2.FONT_HERSHEY_SIMPLEX

mouth_points = [48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67]
r_eye_points = [42, 43, 44, 45, 46, 47]
l_eye_points = [36, 37, 38, 39, 40, 41]

count_mouth_open = 0

def midpoint(p1, p2):
    return int((p1.x + p2.x)/2), int((p1.y + p2.y)/2)


def get_mouth_pen_ratio(mouth_points, facial_landmarks):
   
    left_point = (facial_landmarks.part( mouth_points[12]).x, facial_landmarks.part(mouth_points[12]).y)
    right_point = (facial_landmarks.part( mouth_points[16]).x, facial_landmarks.part(mouth_points[16]).y)
    center_top = midpoint(facial_landmarks.part( mouth_points[13]), facial_landmarks.part(mouth_points[14]))
    center_bottom = midpoint(facial_landmarks.part( mouth_points[19]), facial_landmarks.part(mouth_points[18]))

    
    hor_line = cv2.line(image, left_point, right_point, (0, 255, 0), 2)
    ver_line = cv2.line(image, center_top, center_bottom, (0, 255, 0), 2)
   
    hor_line_lenght = hypot( (left_point[0] - right_point[0]), (left_point[1] - right_point[1]))
    ver_line_lenght = hypot( (center_top[0] - center_bottom[0]), (center_top[1] - center_bottom[1]))

    
    if ver_line_lenght != 0:
        ratio = hor_line_lenght / ver_line_lenght
    else:
        ratio = 60
    
    return ratio



def get_blinking_ratio(eye_points, facial_landmarks):
    
    left_point = (facial_landmarks.part( eye_points[0]).x, facial_landmarks.part(eye_points[0]).y)
    right_point = (facial_landmarks.part( eye_points[3]).x, facial_landmarks.part(eye_points[3]).y)
    center_top = midpoint(facial_landmarks.part( eye_points[1]), facial_landmarks.part(eye_points[2]))
    center_bottom = midpoint(facial_landmarks.part( eye_points[5]), facial_landmarks.part(eye_points[4]))

    hor_line = cv2.line(image, left_point, right_point, (0, 255, 0), 2)
    ver_line = cv2.line(image, center_top, center_bottom, (0, 255, 0), 2)

    hor_line_lenght = hypot( (left_point[0] - right_point[0]), (left_point[1] - right_point[1]))
    ver_line_lenght = hypot( (center_top[0] - center_bottom[0]), (center_top[1] - center_bottom[1]))

    ratio = ver_line_lenght / hor_line_lenght
    
    return ratio



capture = cv2.VideoCapture(0)


capture.set(cv2.CAP_PROP_FRAME_WIDTH, 640)
capture.set(cv2.CAP_PROP_FRAME_HEIGHT, 480)


frame = 0
sum =0
open_average_ear = 0
close_average_ear = 0
step_count = 0
sleep_frame = 0


while True :
    
    if step_count == 0: 
        print("3초간 평소 눈의 크기를 측정합니다\n")
        while True:
            ready = input("편안하게 눈을 떠주시고 준비가 되셨으면 r을 입력해주세요: ")
            if ready == 'r':
                step_count += 1
                break
            else:
                print("잘못된 입력 값입니다. 다시 입력해주세요\n")
    elif step_count == 2:
        print("3초간 감은 눈의 크기를 측정합니다\n")
        print("부저소리가 날때까지 눈을 편안하게 감아주세요\n")
        while True:
            ready = input("준비가 되셨으면 r을 입력해주세요: ")
            if ready == 'r':
                step_count += 1
                break
            else:
                print("잘못된 입력 값입니다. 다시 입력해주세요\n")
    elif step_count == 4:
        print("s를 입력하면 졸음 인식 시스템이 시작됩니다.")
        while True:
            ready = input("준비가 되셨으면 s을 입력해주세요: ")
            if ready == 's':
                break
            else:
                print("잘못된 입력 값입니다. 다시 입력해주세요\n")
        print("q를 누르면 시스템이 종료됩니다.")
        step_count += 1

    _, image = capture.read()

    # convert frame to gray
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) 
                                                    
    faces = detector(gray)                         

    for face in faces:
        landmarks = predictor(gray, face)          
        
    mouths = get_mouth_pen_ratio( mouth_points, landmarks)
    if mouths <= 5.0: 
        count_mouth_open += 1

    left_eye_ratio = get_blinking_ratio( l_eye_points, landmarks)
    right_eye_ratio = get_blinking_ratio( r_eye_points, landmarks)

    blinking_ratio = (left_eye_ratio + right_eye_ratio) / 2

    
    if frame < 100: 
        frame = frame + 1
        sum = sum + blinking_ratio
    elif frame == 100:
        open_average_ear = sum / frame 
        print("ear = %f" %open_average_ear)
        sum = 0
        step_count += 1
        frame +=1 
        continue
    elif frame > 100 and frame < 200: 
        frame = frame + 1
        sum = sum + blinking_ratio
    elif frame == 200:
        close_average_ear = sum / (frame-100) 
        print("ear = %f" %close_average_ear)
        step_count += 1
        frame += 1 
        beepsound() 
        continue

    
    if close_average_ear != 0: 
        cv2.putText(image, "Mouth open: " + str(count_mouth_open), (50, 50), font, 2, (255, 0, 0))
        print("Mouth open: " + str(count_mouth_open)) 

        if blinking_ratio <= (open_average_ear + close_average_ear)/2:  
            cv2.putText(image, "blinking", (50, 50), font, 2, (255, 0, 0))
            print("blink")
            sleep_frame += 1

    
    if sleep_frame > 2: 
        frame += 1
        if frame - 201 == 15:
            if sleep_frame > 13: 
                print("졸음운전 판단")
                beepsound()
                sleep_frame = 0
                frame = 201
                pass 
            else:
                sleep_frame = 0 
                frame = 201 
                 

    
    cv2.imshow("Frame", image)

    key = cv2.waitKey(1) & 0xFF

    
    if key == ord("q"): 
        break

Answer 1

想象一下，如果面部检测器没有找到任何东西会发生什么：

faces = detector(gray)

for face in faces:
    landmarks = predictor(gray, face)

mouths = get_mouth_pen_ratio( mouth_points, landmarks)

那么您的landmarks确实没有在这里定义，因此存在错误。

现在，这真的取决于当没有检测到人脸时你想要做什么。 例如跳到下一帧？ 像这样的东西？

faces = detector(gray)
if not faces:
    continue

for face in faces:
    landmarks = predictor(gray, face)

if not landmarks:
    continue

mouths = get_mouth_pen_ratio( mouth_points, landmarks)

我也相信你的代码中的逻辑是错误的。 如果你有更多的面孔，你只保留最后一张的地标，只检测最后一张脸上的嘴巴。 怎么样：

faces = detector(gray)
for face in faces:
    landmarks = predictor(gray, face)
    if landmarks:
        mouths = get_mouth_pen_ratio( mouth_points, landmarks)
        # do something with mouths...