How not to detect a face from others

asked 2017-04-11 00:31:56 -0600

I'm working on a project which basically detects faces from a video stream and do some processing on the detected region. I'm using HOG and detectMultiScale().

Consider the scenario, a person pushing a wheelchair. I don't want to detect the person on the wheelchair but the person who is pushing the wheelchair. Any way on masking the area of wheelchair or not detect the guy on wheelchair?

edit retag flag offensive close merge delete

add a comment

1 answer

Sort by » oldest newest most voted

answered 2017-04-14 07:25:51 -0600

StevenPuttemans

20029 ●16 ●82 ●207 http://stevenputtemans...

There are several approaches, but they all depend on what you are able to do.

You can build a wheelchair detector. Then the closest found face detection, that is also lateral to the wheelchair, will probably be related to the guy in the wheelchair. However I can imagine compositions where this does not work. Wheelchair detection is possible through OpenCV. I have done something similar with walking aids.
You can do some kind of person (re)identification, and let the user define which users to ignore. Have a look at authentication problems if you go for this approach.