Research topics for students

Person detection safety feature development

(Added 10.09.2024)

Description:

Camera system is detecting and tracking a person or multiple persons. If person is closer than the threshold (5 meters) and person’s trajectory is matching with the tracking device trajectory, then the system should raise an alarm. It behaves like a car that detects potential collision with the person while driving.

Device is moving in an open area up to 25 km/h. Different persons may appear to the device field of view from different areas and different speeds, also may stand by at one place.

Explanation:

Camera should be stereo camera that is able to detect distance to the object. For testing purposes Intel RealSense D435 camera could be used. Nvidia Jetson Nano could be used for processing the camera feed and perform real-time detection and tracking.

Yolo v9 or v10 nano/small could be used for person detection model.

Tracking mechanism should be researched to select a appropriate one. It is important that tracker does not loose an object. 

Expected Outcomes:

• System implementation should be done in Python.

• System can detect person and its distance.

• System should be able to track one or more (up to 3) persons.

• If person is closer than 5 meters and device’s trajectory is about to hit the person (or is closer than 0.5 meters to the person) then warning is raised.
• There should always be a warning active if at least one person is closer than the threshold and their trajectory collides with the tracking device one.
• Optional: Create a visual interface to show device’s trajectory and detected object trajectory, for demonstration purposes.