How to detect people from top view

Hello,

I would like to detect people from a top view like in the image. I am relatively new to image processing and don't know the best approach on how to takle this. (The image is just one from a bigger dataset I have and I can't shoot a clean plate for these)

I would need to detect people in a scenario like this.

Thanks in advance!

Answer 1

[Image Analyst]

I'd try snapping a photo with no people in there first, and then subtract your people photo from it. Then threshold it.

backgroundImage = imread('background.png');
peopleImage = imread('people.png');
diffImage = imabsdiff(backgroundImage, peopleImage);

[Florian Wagner]

Hello Image Analyst,

thanks for the fast reply!

The image I posted is just one from a bigger dataset I have and I can't shoot a clean plate for these (I should have clarified that more, sorry!).

Do you have more ideas, even if they are more work/complicated (this is one processing step for a group project of 5), on how to detect them?

Thanks again!

Kind Regards,

Florian

[Image Analyst]

If it's a video, the Computer Vision Toolbox has examples/demos for tracking people. Look it up.

You can get a background image by taking the mode of a whole bunch of frames.

[Florian Wagner]

I only have a set of images, different scenes/places altogether

[Image Analyst]

So how is your algorithm supposed to know that a dark round spot in the image is the top view of a person and not a stain on the floor? If you have no reference image, and the blob is not moving, I don't think there is any reliable way.