使用Kinect_V2相机+pykinect2_识别并保存人体skeleton运动时序坐标信息

运行之前，需要先去Microsoft官网下载并安装Kinect V2的SDK：

Kinect for Windows SDK 2.0 - Microsoft​​​​​​​

接着，开始从下面的网址下载Kinect V2相机的Python版本功能包pykinect2：

pykinect2-Kinect-V2-SDK-Python版本下载

再通过本地安装指令，将上面的功能包安装到自己的Python运行环境中：

pip install '+ path'  # ‘+ path’替换为自己下载的功能包所在的本地路径

将这些依赖安装好：

import csv
import pygame
from pykinect2 import PyKinectRuntime, PyKinectV2
from pykinect2.PyKinectV2 import *

之后即可开始运行程序：

import csv
import pygame
from pykinect2 import PyKinectRuntime, PyKinectV2
from pykinect2.PyKinectV2 import *

# 初始化Pygame
pygame.init()

# 设置窗口尺寸
width, height = 960, 540
size = (width, height)
screen = pygame.display.set_mode(size)
pygame.display.set_caption("PyKinect2 Body Tracking")

# 创建Kinect实例
kinect = PyKinectRuntime.PyKinectRuntime(PyKinectV2.FrameSourceTypes_Body)

# 创建CSV文件
with open('skeleton_data.csv', 'w', newline='') as csvfile:
    writer = csv.writer(csvfile)
    joint_ids = [joint_id for joint_id in range(0, PyKinectV2.JointType_Count)]
    header = ['Frame'] + [f'Joint {joint_id+1} X' for joint_id in joint_ids] + [f'Joint {joint_id+1} Y' for joint_id in joint_ids] + [f'Joint {joint_id+1} Z' for joint_id in joint_ids]
    writer.writerow(header)  # 写入表头

    # 游戏主循环
    running = True
    frame_count = 0
    frame_rate = 30  # 帧率设定为30帧/秒
    interval = 1.0 / frame_rate  # 计算每帧的时间间隔
    last_time = pygame.time.get_ticks() / 1000.0  # 上一帧的时间戳
    tracked_body_index = None  # 跟踪的人体索引

    while running:
        # 处理事件
        for event in pygame.event.get():
            if event.type == pygame.QUIT:
                running = False

        # 获取骨架数据
        if kinect.has_new_body_frame():
            bodies = kinect.get_last_body_frame().bodies

            if bodies is not None:
                # 寻找正在跟踪的人体
                for i in range(0, kinect.max_body_count):
                    body = bodies[i]
                    if body.is_tracked:
                        tracked_body_index = i
                        break

        # 跟踪到人体后才进行数据记录
        if tracked_body_index is not None:
            # 计算当前时间戳
            current_time = pygame.time.get_ticks() / 1000.0

            # 根据时间间隔记录数据
            if current_time - last_time >= interval:
                body = bodies[tracked_body_index]
                frame_data = []

                # 存储关节坐标数据
                for joint_id in range(0, PyKinectV2.JointType_Count):
                    joint = body.joints[joint_id]
                    frame_data.extend([joint.Position.x, joint.Position.y, joint.Position.z])

                # 调整关节坐标数据的排列顺序
                rearranged_data = []
                for joint_id in range(1, PyKinectV2.JointType_Count+1):
                    rearranged_data.extend(frame_data[(joint_id-1)*3 : joint_id*3])

                # 写入CSV文件
                writer.writerow([frame_count+1] + rearranged_data)

                # 打印关节数据
                print(f"Frame: {frame_count+1}, Joint Positions: {rearranged_data}")

                # 更新上一帧的时间戳
                last_time = current_time

                # 帧数加1
                frame_count += 1

        # 清空屏幕
        screen.fill((0, 0, 0))

        # 绘制骨架连线图
        if tracked_body_index is not None:
            joints = bodies[tracked_body_index].joints

            # 定义骨架连线关系
            bone_connections = [
                (JointType_Head, JointType_Neck),
                (JointType_Neck, JointType_SpineShoulder),
                (JointType_SpineShoulder, JointType_SpineMid),
                (JointType_SpineMid, JointType_SpineBase),
                (JointType_SpineShoulder, JointType_ShoulderRight),
                (JointType_SpineShoulder, JointType_ShoulderLeft),
                (JointType_ShoulderRight, JointType_ElbowRight),
                (JointType_ElbowRight, JointType_WristRight),
                (JointType_WristRight, JointType_HandRight),
                (JointType_ShoulderLeft, JointType_ElbowLeft),
                (JointType_ElbowLeft, JointType_WristLeft),
                (JointType_WristLeft, JointType_HandLeft),
                (JointType_SpineBase, JointType_HipRight),
                (JointType_SpineBase, JointType_HipLeft),
                (JointType_HipRight, JointType_KneeRight),
                (JointType_KneeRight, JointType_AnkleRight),
                (JointType_AnkleRight, JointType_FootRight),
                (JointType_HipLeft, JointType_KneeLeft),
                (JointType_KneeLeft, JointType_AnkleLeft),
                (JointType_AnkleLeft, JointType_FootLeft)
            ]

            # 绘制骨架连线
            for connection in bone_connections:
                joint1 = joints[connection[0]]
                joint2 = joints[connection[1]]
                if joint1.TrackingState == TrackingState_Tracked and joint2.TrackingState == TrackingState_Tracked:
                    joint1_pos = kinect.body_joint_to_color_space(joint1)
                    joint2_pos = kinect.body_joint_to_color_space(joint2)
                    pygame.draw.line(screen, (255, 0, 0), (joint1_pos.x, joint1_pos.y), (joint2_pos.x, joint2_pos.y), 5)

        # 更新屏幕显示
        pygame.display.flip()

    # 退出Pygame
    pygame.quit()

# 释放Kinect资源
kinect.close()

其中，设置获取数据的频率为30帧/秒，按照时序将25个骨架识别点的坐标，以“Joint1X、Joint2X、...、Joint25X、Joint1Y、Joint2Y.....Joint25Z”的顺序，按行逐帧存储在csv文件中。