像 cv2.boundingrect 一样围绕从 mediapipe 手界标检测返回的所有点创建一个矩形

Question

我正在使用以下代码使用 mediapipe 检测手部地标

import cv2
import mediapipe as mp

mphands = mp.solutions.hands
hands = mphands.Hands()
mp_drawing = mp.solutions.drawing_utils
cap = cv2.VideoCapture(0)

while True:
    _, frame = cap.read()
    framergb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
    result = hands.process(framergb)
    hand_landmarks = result.multi_hand_landmarks
    if hand_landmarks:
        for handLMs in hand_landmarks:
            mp_drawing.draw_landmarks(frame, handLMs, mphands.HAND_CONNECTIONS)
            print("\n\n\n")
    cv2.imshow("Frame", frame)

    cv2.waitKey(1)

我只想在检测器返回的所有点周围画一个矩形请告诉是否有任何方法可以内置 mediapipe 或使用 opencv

Answer 1

在while循环之前，确定每帧的宽度和高度为：

_, frame = cap.read()

h, w, c = frame.shape

对于检测到的每个 landLM，定义最小 x 和 y 坐标以及最大 x 和 y 坐标的初始变量。前两个变量稍后将作为矩形的起点，最后两个变量稍后将作为矩形的最后一个点：

            x_max = 0
            y_max = 0
            x_min = w
            y_min = h

遍历handLM变量，找到手的每个点的x和y坐标：

            for lm in handLMs.landmark:
                x, y = int(lm.x * w), int(lm.y * h)

在检测到新坐标时更新最小和最大 x 和 y 变量：

                if x > x_max:
                    x_max = x
                if x < x_min:
                    x_min = x
                if y > y_max:
                    y_max = y
                if y < y_min:
                    y_min = y

最后，用点画出矩形：

            cv2.rectangle(frame, (x_min, y_min), (x_max, y_max), (0, 255, 0), 2)

一共：

import cv2
import mediapipe as mp

mphands = mp.solutions.hands
hands = mphands.Hands()
mp_drawing = mp.solutions.drawing_utils
cap = cv2.VideoCapture(0)

_, frame = cap.read()

h, w, c = frame.shape

while True:
    _, frame = cap.read()
    framergb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
    result = hands.process(framergb)
    hand_landmarks = result.multi_hand_landmarks
    if hand_landmarks:
        for handLMs in hand_landmarks:
            x_max = 0
            y_max = 0
            x_min = w
            y_min = h
            for lm in handLMs.landmark:
                x, y = int(lm.x * w), int(lm.y * h)
                if x > x_max:
                    x_max = x
                if x < x_min:
                    x_min = x
                if y > y_max:
                    y_max = y
                if y < y_min:
                    y_min = y
            cv2.rectangle(frame, (x_min, y_min), (x_max, y_max), (0, 255, 0), 2)
            mp_drawing.draw_landmarks(frame, handLMs, mphands.HAND_CONNECTIONS)
    cv2.imshow("Frame", frame)

    cv2.waitKey(1)

像 cv2.boundingrect 一样围绕从 mediapipe 手界标检测返回的所有点创建一个矩形

create a rectangle around all the points returned from mediapipe hand landmark detection just like cv2.boundingrect does

python

opencv

mediapipe