I’m a software engineer student however I’m new to computer vision field and have little knowledge about it, but I’m eager to learn and came in hopes to get some direction / reference to what I should look into when approaching this problem. Thank you for everyone in advance.
I have a project with two features: Using live feed from camera (raspberry pi or IP camera):
1. Detect (bounding box), track and count people going in and out of the building / past the camera. (This part has a lot of information and resources of how to do it, so I should be fine)
2. Identify if the same person hasn’t passed more than 3 times. If the person is going in and out of the camera view - program should understand that it’s the same person and the count shouldn’t increase.
This second part is what’s difficult and my quick google search didn’t give me much - because I assume identification of a person should be done by cropping the detected box of a person, extracting and analysing every detected person’s features (clothing, color, height, etc.) (no face detection allowed - as it’s illegal now I think) and trying to match results with every person detected. (? classification problem ?) - However I’m lost.
I was wondering if anyone could give me some references and direction of how one should go about this problem ? - maybe basic… techniques / terms I should look into - which could help me to start on the right track and hopefully find the solution by the end of the university year. Or maybe basic step-by-step abstract idea ? Appreciate.