Prev: imagesc, pcolor, and NaN
Next: augmented reality
From: Vikas Dhummi on 8 Jan 2010 19:27
I am working on a project, The aim is to count the number of people in the Region of interest.
1.I started with Blob analysis.The issue is that the blobs are sometimes separated and gives more occupants than expected.
2.I tried with pixel count the number of pixels a person can give when he is difference from the background. This is not accurate as well
3.I am now looking at an algorithm which has a virtual line and those who crosses should be counted depending on the direction.
Any idea or code similar would be highly appreciated.
Thanks
From: Rune Allnor on 9 Jan 2010 00:14
On 9 Jan, 01:27, "Vikas Dhummi" <vikas...(a)gmail.com> wrote:
> I am working on a project, The aim is to count the number of people in the Region of interest.
> 1.I started with Blob analysis.The issue is that the blobs are sometimes separated and gives more occupants than expected.

You need to realize that an 'image' consits of pixels.
The only thing you can derive from that image are
relations between those pixels. There is no reason
to expect that one 'blob' should be related in any
way to any physical object, let alone anything as
specific as a 'person'.

Don't try to count 'people'. Try to count 'blobs'
or whatever, possibly discriminated by shape, size
and/or color.

> Any idea or code similar would be highly appreciated.

So you want to copy what others already did and
pass it on as your own? Forget it.

Rune
From: ImageAnalyst on 9 Jan 2010 00:44
You'll find lots of ideas here:
http://iris.usc.edu/Vision-Notes/bibliography/contents.html
especially in sections 18, 19, and 21.
Good luck!
 | 
Pages: 1
Prev: imagesc, pcolor, and NaN
Next: augmented reality