How to detect human motion and direction from a still image(opencv python/ matlab)?

asked 2017-10-31 09:26:40 -0500

josari gravatar image

updated 2017-11-01 09:05:46 -0500

Hi, guys

When given a still image and after detecting human(s), does anyone know (of any models/code);

-how to detect if the person is moving (by the width of their stance? usually if a person's legs are far apart we assume they are in a motion) and if they are,

-then how to detect the direction of the movement(direction of their head)?

Is this possible to do in a still image (as shown below) or can this only be done on an image sequence(video)?


image description

image description

edit retag flag offensive close merge delete


Can you further elaborate on what you mean by given a still image...detect if the person is moving? Or at least explain the outline of your task as a whole because as it stands, your question makes no sense. How can you detect motion on something that is still/stationary/stagnant?

eshirima gravatar imageeshirima ( 2017-10-31 10:06:09 -0500 )edit

Yes, I am assuming from an image if the person's legs are far apart, then they are in motion.

Also if there is a misalignment for the upper and lower part of the body, this could also mean they are moving?

josari gravatar imagejosari ( 2017-10-31 11:32:11 -0500 )edit

Can you post a couple sample test images? Or add a link to them if possible.

eshirima gravatar imageeshirima ( 2017-10-31 12:33:56 -0500 )edit
sturkmen gravatar imagesturkmen ( 2017-10-31 14:35:33 -0500 )edit

Hi I added these pictures.. please let me know what you think

josari gravatar imagejosari ( 2017-11-01 09:08:10 -0500 )edit

@sturkmen, the GEI + HOG idea would still need several frames, right ?

berak gravatar imageberak ( 2017-11-01 09:26:39 -0500 )edit