【 DigiKey DIY原创大赛】基于树莓派5B和CSI摄像头的手势控制---项目实操

系统框图

第一部分：

系统选择

选择树莓派5的官方64位版本

配置虚拟环境

使用virtualenv或者venv

python3 -m venv --system-site-packages env1

安装OPENCV库

pip3 install opencv-python==4.7.0.72 
pip3 install opencv-contrib-python==4.7.0.72

安装opencv要注意与python版本相对应。目前树莓派5官方64位系统自带的是

python 3.11版本。

piwheels - opencv-python参考这个网站查看二者对应版本

安装 mediapipe库

pip3 install mediapipe

编写简单测试代码

from picamera2 import Picamera2
import cv2
import time
import mediapipe as mp
import numpy as np


if __name__ == "__main__":

    camera = Picamera2() 
    camera.start()
    mpHands = mp.solutions.hands
    hands = mpHands.Hands()
    mpDraw = mp.solutions.drawing_utils

    while True:
       camera.resolution =(640,480)
       im = camera.capture_array("main")
       image_height, image_width, _ =np.shape(im)
       #raw_capture = PiRGBArray(camera,size=(640,480))
       #time.sleep(2)
       #camera.capture(raw_capture, format='bgr')
       #image = raw_capture.array 
       imgRGB= cv2.cvtColor(im, cv2.COLOR_BGR2RGB)       
       results = hands.process(imgRGB)      
       if results.multi_hand_landmarks:
           for hand in results.multi_hand_landmarks:

                print("\r%.2f %.2f %.2f %.2f %.2f %.2f "%(hand.landmark[0].z,hand.landmark[4].z,hand.landmark[8].z,hand.landmark[12].z,hand.landmark[16].z,hand.landmark[20].z),end="")

                #mpDraw.draw_landmarks(im,hand,mpHands.HAND_CONNECTIONS)

                for i in range(21):
                    pos_x = hand.landmark[i].x*image_width
                    pos_y = hand.landmark[i].y*image_height

                    cv2.circle(im, (int(pos_x),int(pos_y)), 3, (0,255,255),-1)


       cv2.imshow("hands",im)

       key =  cv2.waitKey(1) & 0xFF   


       if key ==  ord('q'):
            break
    camera.release() 

第二部分：

手势识别现象：

第三部分：

以下视频演示：

手势识别