Summary: Publish frames from a connected camera or MJPEG/RTSP stream to an MQTT topic, and view the feed in a browser on another computer with Streamlit.
Long introduction: A typical task in IOT/science is that you have a camera connected to one computer and you want to view the camera feed on a second computer, and maybe preprocess the images before saving them to disk. I have always found this to be way more effort than expected. In particular, working with camera streams can get quite complicated and may lead you to experiment with tools like Gstreamer and ffmpeg that have a steep learning curve. In contrast, working with MQTT is very straightforward and is often familiar to anyone with an interest in IOT. This repo, mqtt-camera-streamer uses MQTT to send frames from a camera over a network at low frames-per-second (FPS). A viewer is provided for viewing the camera stream on any computer on the network. Frames can be saved to disk for further processing. Also it is possible to setup an image processing pipeline by linking MQTT topics together, using an on_message(topic) to do some processing and send the processed image downstream on another topic.
Note that this is not a high FPS solution, and in practice I achieve around 1 FPS which is practical for IOT experiments and tasks such as preprocessing (cropping, rotating) images prior to viewing them. This code is written for simplicity and ease of use, not high performance.
Install system wide on an RPi, or on other OS use a venv to isolate your environment, and install the required dependencies:
$ (base) python3 -m venv venv
$ (base) source venv/bin/activate
$ (venv) pip3 install -r requirements.txt
The check-opencv-cameras.py script assists in discovering which cameras OpenCV can connect to on your computer (does not work with RPi camera). If your laptop has a built-in webcam this will generally be listed as VIDEO_SOURCE = 0. If you plug in an external USB webcam this takes precedence over the built-in webcam, with the external camera becoming VIDEO_SOURCE = 0 and the built-in webcam becoming VIDEO_SOURCE = 1.
To check which OpenCV cameras are detected run:
$ (venv) python3 scripts/check-opencv-cameras.py
Use the config.yml file in the config directory to configure your system. If your desired camera is listed as source 0 you will configure video_source: 0. Alternatively you can configure the video source as an MJPEG or RTSP stream. For example in config.yml you may configure something like video_source: "rtsp://admin:password@192.168.1.94:554/11" for a commercial RTSP camera. To configure a RPi camera running the web_streaming.py example you configure video_source: http://pi_ip:8000/stream.mjpg
Validate the config can be loaded by running:
$ (venv) python3 scripts/validate-config.py
Note that this script does not check the accuracy of any of the values in config.yml, just that the file path is correct and the file structure is OK.
By default scripts/opencv-camera.py will look for the config file at ./config/config.yml but an alternative path can be specified using the environment variable MQTT_CAMERA_CONFIG. You can set this using export MQTT_CAMERA_CONFIG=/home/pi/github/mqtt-camera-streamer/config/config.yml
To publish camera frames with OpenCV over MQTT:
$ (venv) python3 scripts/opencv-camera-publish.py
To view the camera stream with Streamlit:
$ (venv) streamlit run scripts/viewer.py
Note: if Streamlit becomes unresponsive, ctrl-z to pause Streamlit then kill -9 %%. Also note that the viewer can be run on any machine on your network.
To save frames to disk:
$ (venv) python3 scripts/save-captures.py
As save-captures.py but in addition saving the frame thumbnail to a sqlite db:
$ (venv) python3 scripts/db-recorder.py
The images can be viewed using sqlite browser
If you wish to run a server with UI for browsing the images then datasette with the datasette-render-images plugin can be used.
$ (venv) pip install datasette
$ (venv) pip install datasette-render-images
$ (venv) datasette captures/records.db
To process a camera stream (the example rotates the image):
$ (venv) python3 scripts/processing.py
You can view the camera feed using Home Assistant and configuring an MQTT camera. Add to your configuration.yaml:
camera:
- platform: mqtt
topic: homie/mac_webcam/capture
name: mqtt_camera
- platform: mqtt
topic: homie/mac_webcam/capture/rotated
name: mqtt_camera_rotated
- platform: mjpeg # the raw mjpeg feed if using picamera
name: picamera
mjpeg_url: http://192.168.1.134:8000/stream.mjpg
Need an MQTT broker? If you have Docker installed I recommend eclipse-mosquitto. A basic broker can be run with:
docker run -p 1883:1883 -d eclipse-mosquitto
Note that I have structured the MQTT topics following the homie MQTT convention, linked in the references. This is not necessary but is best practice IMO.
OpenCV is used to read the images from a connected camera or MJPEG/RTSP stream. On a Raspberry pi (RPi) installing OpenCV can be troublesome, and I found it necessary to first sudo apt-get install libatlas-base-dev libjasper-dev libqtgui4 python3-pyqt5 libqt4-test libilmbase-dev libopenexr-dev libgstreamer1.0-dev libavcodec58 libavformat58 libswscale5 before installing opencv using the instructions below. Likewise Streamlit can be challenging to install on an RPi, and if you dont need it then remove it from requirements.txt. If you do wish to install Streamlit on the RPi see this thread for latest guidance. On 24/3/2021 I was able to install opencv-python==4.5.1.48 but not streamlit on an RPi4 32bit.
Use an official RPi camera and ensure picamera is installed with pip3 install picamera. If you use the RPi in desktop mode you can check the camera feed using raspistill -o image.jpg. Use the official web_streaming example which creates an mjpeg stream on http://pi_ip:8000/stream.mjpg. This mjpeg stream can be configured as a source with mqtt-camera-streamer to translate the mjepg stream to an mqtt stream.
You can run any of the scripts as a service, which means they will automatically start on RPi boot, and can be easily started & stopped. Create the service file in the appropriate location on the RPi using:
sudo nano /etc/systemd/system/my_script.service
Entering the following (adapted for your script.py file location and args, assumes you are using system python3):
[Unit]
Description=Service for mqtt-camera-publish
After=network.target
[Service]
ExecStart=/usr/bin/python3 -u opencv-camera-publish.py
WorkingDirectory=/home/pi/github/mqtt-camera-streamer/scripts
StandardOutput=inherit
StandardError=inherit
Restart=always
User=pi
[Install]
WantedBy=multi-user.target
Once this file has been created you can to start the service using:
sudo systemctl start my_script.service
View the status and logs with:
sudo systemctl status my_script.service
Stop the service with:
sudo systemctl stop my_script.service
Restart the service with:
sudo systemctl restart my_script.service
You can have the service auto-start on rpi boot by using:
sudo systemctl enable my_script.service
You can disable auto-start using:
sudo systemctl disable my_script.service
- imageZMQ -> inspired this project, but uses ZMQ. Discussion on ZMQ vs MQTT here
- homie MQTT convention -> convention for structuring MQTT topics
- yolocam_mqtt -> another source of ideas
- In-depth review and comparison of the Raspberry Pi High Quality Camera