Taking into account the defined parameters and described collision events (see 3. Collision Detection), our approach for detecting near misses consider the following scenarios:

1. Cars that are going through an intersection (in perpendicular directions) and suddenly decrease their velocity and change their direction to avoid a collision.

2. Cars approaching the crosswalk while pedestrians are crossing the street.

In order to optimize the Near Misses detection, we aim to include the following scenarios:

*Extend the Areas of Interest interaction along with OpenVINO’s detection model, (i.e) by including "No Street Area" (for train rails) as AoI, and being able to detect when a vehicle has suddenly changed street lanes to the opposite direction.

*Detect when pedestrians are crossing the street without using the crosswalk.