The goal of carwatch is to capture a video stream and detect cars waiting at the parking garage gate for it to close before leaving.

When a car is detected, carwatch will take a screenshot and save it into a JPG.

Carwatch also provides ability to create a weekly report and send it by email once a week with graph and screenshots attached in a zip file.

Carwatch is designed to capture a RTSP video stream and run on a Raspberry Pi.

Requirement min Python 3.

OpenCV install on Ubuntu:

sudo apt update
sudo apt install libatlas-base-dev
sudo apt install python3-opencv
python3 -c "import cv2; print(cv2.__version__)"

OpenCV install on Mac (for testing):

brew install opencv
echo 'export PATH="/usr/local/opt/qt/bin:$PATH"' >> ~/.bash_profile
source ~/.bash_profile

Python libraries needed:

pip3 install opencv-python
pip3 install pytesseract
pip3 install matplotlib
pip3 install pandas
pip3 install Pillow

After you clone the git repository, rename the file config_app.sample.py to config_app.py and edit the mail and RTSP configuration. You may also tune other detection settings in config_detection.py which might depend on your camera.

cd /home/user/carwatch
nohup python3 carwatch.py 2>&1 &> carwatch.log &

Service Monitor script to make sure the script is restarted in case it dies (also used to start when the OS boots):

*/5 * * * * cd /home/user/carwatch ; ./service_monitor.sh >> carwatch.log