python - 有没有一种方法可以通过使用opencv/dlib和直播视频来获取额头(边界框)的区域

Question

我一直在进行一个项目，以从直播视频中获取前额区域，而不仅仅是像本例How can i detect the forehead region using opencv and dlib?那样使用和成像并裁剪前额。

cap = cv2.VideoCapture(0)

detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(predict_path)


while True:
    _, frame = cap.read()
    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

    faces = detector(gray) #detects number of faces present

    for face in faces:
        x1 = face.left()
        y1 = face.top()
        x2 = face.right()
        y2 = face.bottom()
        
        cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 0, 255), 3)
        
        landmarks = predictor(gray, face)

        for n in range(68, 81):
            x = landmarks.part(n).x
            y = landmarks.part(n).y

            cv2.circle(frame, (x, y), 4, (0, 255, 0), -1) 
            

我使用https://github.com/codeniko/shape_predictor_81_face_landmarks/blob/master/shape_predictor_81_face_landmarks.dat的地标设法获得了额头区域
但是我需要的是矩形边界框，该矩形边界框位于检测前额区域时地标所在的位置。这可能得到吗？如果没有，我该怎么做才能获得额头区域。提前致谢。

Answer 1

您已经通过以下方式找到了所需的坐标:

for face in faces:
    x1 = face.left()
    y1 = face.top()
    x2 = face.right()
    y2 = face.bottom()

    cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 0, 255), 3)

But what I need is the rectangle bounding box onto where the landmark is at detecting the forehead region.

然后更改y坐标:

cv2.rectangle(frame, (x1, y1-100), (x2, y2-100), (0, 0, 255), 3)

更新
要坚持到额头，我们需要获得最小和最大landmark坐标，然后需要绘制矩形。
步骤1:获取坐标:

初始化x_pts和y_pts

将landmark(n)点存储到数组中。

for n in range(68, 81):
    x = landmarks.part(n).x
    y = landmarks.part(n).y

    x_pts.append(x)
    y_pts.append(y)

    cv2.circle(frame, (x, y), 4, (0, 255, 0), -1)

步骤2:在检测到的点周围绘制矩形

获取最低和最高分

x1 = min(x_pts)
x2 = max(x_pts)
y1 = min(y_pts)
y2 = max(y_pts)

cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 0, 255), 3)

结果:
当我放大网络摄像头时:

当我很远的时候:

码:

import cv2
import dlib

cap = cv2.VideoCapture(0)

detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor("shape_predictor_81_face_landmarks.dat")

while True:
    _, frame = cap.read()
    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

    faces = detector(gray)  # detects number of faces present

    for face in faces:
        x1 = face.left()
        y1 = face.top()
        x2 = face.right()
        y2 = face.bottom()

        landmarks = predictor(gray, face)

        x_pts = []
        y_pts = []

        for n in range(68, 81):
            x = landmarks.part(n).x
            y = landmarks.part(n).y

            x_pts.append(x)
            y_pts.append(y)

            cv2.circle(frame, (x, y), 4, (0, 255, 0), -1)

        x1 = min(x_pts)
        x2 = max(x_pts)
        y1 = min(y_pts)
        y2 = max(y_pts)

        cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 0, 255), 3)

    cv2.imshow("out", frame)
    key = cv2.waitKey(1) & 0xFF

    # if the `q` key was pressed, break from the loop
    if key == ord("q"):
        break