This experiment's goal is to detect the vehicles moving towards the camera and alert the user if it can be dangerous pass
python3 -m pip install -r requirements.txt
python3 main.py
The calculation process of the collision relies on the depth information provided by the OAK-D camera.
Whereas we can operate in 3D coordinates, in fact most of the calculations are being made in 2D, taking only x (horizontal) and z (depth) into account
The collsion may occur when:
- Car trajectory is pointing towards the camera
- Car speed, and therefore time to impact, is below a threshold
If those two conditions are met, we display a warning message on the preview window.
All of these calculations are being done inside crash_avoidance.py file.
We also need to persist information about the object between the frames, so that we actually track
the cars moving on the image, not only detect them, therefore in tracker.py there is a code which
assign detected cars on the image to previously detected car positions