[Raspberry Pi] aarch64 system uses CSI camera

If it is a native 32-bit system, you don’t need to read this article

• OpenFans x64 system are not raspistill，raspivid , it is not possible to operate the camera by this command
• Python operating class PiCamera need libmmal.so , but this so no x64 version , can not be used PiCamera

The error is as follows OSError: libmmal.so: cannot open shared object file: No such file or directory

But you can use opencv/ffmpeg directly operate /dev/video0 to get the video/picture

equipment

• Raspberry Pi 4B
• CSI camera 1.3 is worth about 13 yuan
• OpenFans x64 Raspberry Pi system

step

Insert the CSI camera

The blue side is on the USB port side

Change setting

Since this system does not have raspi-config ，rpi-update etc., you need to modify /boot/config.txt to increase the camera

start_x=1             # essential
gpu_mem=128           # at least, or maybe more if you wish
disable_camera_led=1  # optional, if you don't want the led to glow

Read bcm2835-v4l2 module

modprobe bcm2835-v4l2

Configure automatic loading at boot
echo 'bcm2835-v4l2' >> /etc/modules

Reboot

After the system restarts, a /dev/video0 device will appear, and the installation is successful without reporting an error when executing the following command

ls /dev/video0

Multiple access

At this time, the camera can only be accessed by one process. If you want multiple processes to access at the same time, you can perform the following steps. Skip if there is no such requirement

Reference source:
https://gist.github.com/brutella/a71e8d7aa90af2c53ab500c4125bc178
https://raspberrypi.stackexchange.com/questions/19630/take-picam-image-while-motion-is-running/19897

1. Install the loop drive

apt install v4l2loopback-dkms

Or compile and install

The openfans system may not be able to compile and install
The following applies to the 32-bit Raspberry Pi native system

apt install raspberrypi-kernel-headers
git clone https://github.com/umlaeute/v4l2loopback
cd v4l2loopback
make
sudo make install

2. Create /dev/video1

This device is a multi-access device

modprobe v4l2loopback video_nr=1

Configure automatic creation at startup
echo 'options v4l2loopback video_nr=1' > /etc/modprobe.d/v4l2loopback.conf

3. Flow diversion

Any one of the following operations

3.1. Use ffmpeg to video0 to multiple video1

apt install ffmpeg
ffmpeg -f video4linux2 -i /dev/video0 -vcodec copy -f v4l2 /dev/video1

3.2. Use gstreamer to video0 to multi-channel video1

apt install gstreamer1.0-tools gstreamer1.0-plugins-good
gst-launch-1.0 v4l2src device=/dev/video0 ! v4l2sink device=/dev/video1

3.3. Use v4l2tools to video0 to multiple video1

git clone https://github.com/mpromonet/v4l2tools
sudo apt install liblog4cpp5-dev
make v4l2copy
./v4l2copy /dev/video0 /dev/video1

opencv2 read camera

Install opencv2

You can install opencv2 as follows, but the version is not the latest

apt install python3-opencv

Or install through pip, the latest version, but the speed is very slow, it takes a long time to compile numpy

pip3 install opencv-python

Screenshot & video

image-0~9.jpg and one video.avi in the current directory

import cv2 as cv

# 读取设备
cap = cv.VideoCapture('/dev/video0', cv.CAP_V4L)
# 读取摄像头FPS
fps = cap.get(cv.CAP_PROP_FPS)

# set dimensions 设置分辨率
cap.set(cv.CAP_PROP_FRAME_WIDTH, 800)
cap.set(cv.CAP_PROP_FRAME_HEIGHT, 400)

video = cv.VideoWriter('video.avi', cv.VideoWriter_fourcc('I', '4', '2', '0'), fps, (800, 400)) # 初始化文件写入 文件名 编码解码器 帧率 文件大小

# 录制10帧
for i in range(10):
    # take frame 读取帧
    ret, frame = cap.read()
    if ret:
        # write frame to file
        cv.imwrite('image-{}.jpg'.format(i), frame) # 截图
        video.write(frame) # 录制视频

# release camera 必须要释放摄像头
cap.release()

The picture is the picture of the camera