Python Mediapipe Holistic仅绘制手臂姿态问题求助

要实现只绘制手臂、排除其他身体部位的需求，核心是自定义姿态关键点的连接列表，替代默认的全身连接。以下是具体实现步骤和代码：

1. 明确手臂相关的姿态关键点 #

MediaPipe Holistic的Pose模块中，手臂的核心关键点及对应关系：

• 右侧手臂：右肩 → 右肘 → 右手腕
• 左侧手臂：左肩 → 左肘 → 左手腕

对应的关键点枚举值（也可直接用索引数字）：

• mp_holistic.PoseLandmark.RIGHT_SHOULDER（索引12）
• mp_holistic.PoseLandmark.RIGHT_ELBOW（索引14）
• mp_holistic.PoseLandmark.RIGHT_WRIST（索引16）
• mp_holistic.PoseLandmark.LEFT_SHOULDER（索引11）
• mp_holistic.PoseLandmark.LEFT_ELBOW（索引13）
• mp_holistic.PoseLandmark.LEFT_WRIST（索引15）

2. 自定义手臂连接列表 #

创建仅包含手臂部位的连接关系，替代默认的mp_holistic.POSE_CONNECTIONS：

ARM_CONNECTIONS = [
    (mp_holistic.PoseLandmark.RIGHT_SHOULDER, mp_holistic.PoseLandmark.RIGHT_ELBOW),
    (mp_holistic.PoseLandmark.RIGHT_ELBOW, mp_holistic.PoseLandmark.RIGHT_WRIST),
    (mp_holistic.PoseLandmark.LEFT_SHOULDER, mp_holistic.PoseLandmark.LEFT_ELBOW),
    (mp_holistic.PoseLandmark.LEFT_ELBOW, mp_holistic.PoseLandmark.LEFT_WRIST)
]

3. 完整代码示例 #

import cv2
import mediapipe as mp

# 初始化MediaPipe模块
mp_holistic = mp.solutions.holistic
mp_drawing = mp.solutions.drawing_utils

# 自定义手臂连接
ARM_CONNECTIONS = [
    (mp_holistic.PoseLandmark.RIGHT_SHOULDER, mp_holistic.PoseLandmark.RIGHT_ELBOW),
    (mp_holistic.PoseLandmark.RIGHT_ELBOW, mp_holistic.PoseLandmark.RIGHT_WRIST),
    (mp_holistic.PoseLandmark.LEFT_SHOULDER, mp_holistic.PoseLandmark.LEFT_ELBOW),
    (mp_holistic.PoseLandmark.LEFT_ELBOW, mp_holistic.PoseLandmark.LEFT_WRIST)
]

# 处理视频流
cap = cv2.VideoCapture(0)
with mp_holistic.Holistic(min_detection_confidence=0.5, min_tracking_confidence=0.5) as holistic:
    while cap.isOpened():
        ret, frame = cap.read()
        if not ret:
            break
        
        # 图像格式转换（MediaPipe要求RGB输入）
        image = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
        results = holistic.process(image)
        image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
        
        # 仅绘制手臂的关键点和连接
        mp_drawing.draw_landmarks(
            image,
            results.pose_landmarks,
            connections=ARM_CONNECTIONS,
            # 自定义关键点样式
            landmark_drawing_spec=mp_drawing.DrawingSpec(color=(255, 0, 0), thickness=2, circle_radius=4),
            # 自定义连接线条样式
            connection_drawing_spec=mp_drawing.DrawingSpec(color=(0, 255, 0), thickness=2)
        )
        
        cv2.imshow('Arm Only Visualization', image)
        if cv2.waitKey(10) & 0xFF == ord('q'):
            break

cap.release()
cv2.destroyAllWindows()

可选：包含手部绘制 #

如果需要连着手部一起绘制，可在上述代码中添加手部的绘制逻辑：

# 绘制右手关键点及连接
if results.right_hand_landmarks:
    mp_drawing.draw_landmarks(
        image,
        results.right_hand_landmarks,
        mp_holistic.HAND_CONNECTIONS,
        landmark_drawing_spec=mp_drawing.DrawingSpec(color=(255, 0, 0), thickness=2, circle_radius=3),
        connection_drawing_spec=mp_drawing.DrawingSpec(color=(0, 255, 0), thickness=2)
    )

# 绘制左手关键点及连接
if results.left_hand_landmarks:
    mp_drawing.draw_landmarks(
        image,
        results.left_hand_landmarks,
        mp_holistic.HAND_CONNECTIONS,
        landmark_drawing_spec=mp_drawing.DrawingSpec(color=(255, 0, 0), thickness=2, circle_radius=3),
        connection_drawing_spec=mp_drawing.DrawingSpec(color=(0, 255, 0), thickness=2)
    )

内容的提问来源于stack exchange，提问作者Abu Bakar BSA