This week, I started to do some things with AI vision because the core of my project is face recognition. I connect the camera, and the holder with the Raspberry Pi, and this is the photo of my devices👇

Step 1: Connect the Camera

First, I wanted to connect the camera to my Pi, but I had some problems with this step: I installed a package in the virtual environment, and it did not work. Ultimately, I installed this package in the globule environment and it worked well. So as a result, I can see the real-time video that the camera takes on my laptop.

Here is the code👇

import cv2
from picamera2 import Picamera2
import libcamera

def main():
    picamera = Picamera2()
    
    # 配置相机设置
    config = picamera.create_preview_configuration(main={"format": 'RGB888', "size": (640, 480)},
                                                   raw={"format": "SRGGB12", "size": (1920, 1080)})
    config["transform"] = libcamera.Transform(hflip=1, vflip=1)  # 水平和垂直翻转
    picamera.configure(config)
    picamera.start()

    while True:
        # 捕获帧
        frame = picamera.capture_array()

       # 转换为 BGR 格式
        frame = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR)
        
        # 显示图像
        cv2.imshow("Camera Output", frame)

        

        listen_keyboard.listen_keyboard()

if __name__ == "__main__":
    main()

But quickly, I realized a serious issue that people’s faces in the camera are blue, and the Chinese flag is blue too! I spent plenty of time to solve the problem and ultimately I realized that the Blue channel and the Red channel are reversed. So I added the following code to solve this issue👇

frame[..., [0, 2]] = frame[..., [2, 0]]  # B 和 R 通道交换

As a result, I can see the video with the correct color👇

Step 1: Connect the Holder

Then, I started to connect the holder, a machine to let the camera move.

Here is the wiring diagram👇

And here is the code I wrote👇

main.py

import cv2
from picamera2 import Picamera2
import listen_keyboard
import libcamera

def main():
    picamera = Picamera2()
    
    # 配置相机设置
    config = picamera.create_preview_configuration(main={"format": 'RGB888', "size": (640, 480)},
                                                   raw={"format": "SRGGB12", "size": (1920, 1080)})
    config["transform"] = libcamera.Transform(hflip=1, vflip=1)  # 水平和垂直翻转
    picamera.configure(config)
    picamera.start()

    while True:
        # 捕获帧
        frame = picamera.capture_array()

        # 转换为 BGR 格式
        frame = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR)

        frame[..., [0, 2]] = frame[..., [2, 0]]  # B 和 R 通道交换

        # 显示图像
        cv2.imshow("Camera Output", frame)

        

        listen_keyboard.listen_keyboard()

if __name__ == "__main__":
    main()