Ask Your Question

How to detect a fallen person with opencv using a webcam?

asked 2015-03-13 07:06:07 -0500

sam3891 gravatar image


In my project, I should be able to detect a person on the floor in an image. I am using a Festo Robotino 2010 model, which has Logitech C250, laser scanner and IR detectors. What is the best way to go about it in OpenCV?

edit retag flag offensive close merge delete



fallen people you can detect by rotating the image 90° and then apply the pedestrian detector

thdrksdfthmn gravatar imagethdrksdfthmn ( 2015-03-13 08:11:47 -0500 )edit

Very useless suggestion as he does not know in which direction he has to turn the image! BTW: If you train a detector at the southpole, would it also work on the northpole?

FooBar gravatar imageFooBar ( 2015-03-13 13:40:43 -0500 )edit

Ok, he can apply the detector on the both rotated images (+90° and -90°). But this is just for test, if it is slow, than he may also train a detector of fallen people. IMHO there is no reason to try to train the detectors if this approach is not going to work.

thdrksdfthmn gravatar imagethdrksdfthmn ( 2015-03-16 03:28:00 -0500 )edit

1 answer

Sort by » oldest newest most voted

answered 2015-03-14 19:11:04 -0500

Going with OpenCV using image processing. Go with "movement" detection code. Still further more if you want to add then you add constrain of skin color.

edit flag offensive delete link more



I need to detect only a person who has already fallen.

sam3891 gravatar imagesam3891 ( 2015-03-15 16:21:16 -0500 )edit

@sam3891: I think you shall try to verify if the fallen persons are similar to the standing ones (rotated with 90°). Other approach: BOW (for people) + some "filters" for finding if the person is fallen.

thdrksdfthmn gravatar imagethdrksdfthmn ( 2015-03-16 03:31:40 -0500 )edit

Question Tools

1 follower


Asked: 2015-03-13 07:06:07 -0500

Seen: 245 times

Last updated: Mar 13 '15