Counting people in image sequence
in [Matlab]
|
From: George on 3 May 2010 04:15 Hello!!! I am trying to make a code in matlab for counting people in image sequences.But i am not familiriar with the program and i have several problems.I have read much stuff but i got complicated with all the information.So here it is what i have made until now.I am calculating the abs of two consecutive frames and the image i take as a result is something like this http://img91.imageshack.us/i/94725571.jpg/ http://img99.imageshack.us/i/11146961.jpg/ http://img99.imageshack.us/i/92737961.jpg/ Is this a right approach or not?What should i do next?How will i count the "human shapes"?
From: Dave Robinson on 3 May 2010 09:22 George <one_furious_angel(a)hotmail.com> wrote in message <1682024047.62826.1272888977438.JavaMail.root(a)gallium.mathforum.org>... > Hello!!! > I am trying to make a code in matlab for counting people in image sequences.But i am not familiriar with the program and i have several problems.I have read much stuff but i got complicated with all the information.So here it is what i have made until now.I am calculating the abs of two consecutive frames and the image i take as a result is something like this > http://img91.imageshack.us/i/94725571.jpg/ > http://img99.imageshack.us/i/11146961.jpg/ > http://img99.imageshack.us/i/92737961.jpg/ > Is this a right approach or not?What should i do next?How will i count the "human shapes"? I am not sure that this is a very robust method of tackling the problem at all - suppose a person remains stationary - doesn't this mean they don't get counted. I think that if this was my project I would look for peoples faces, using some form of racial invariant colour space (REDNESS from a normalized RGB colour transform forms quite a nice clustering for a skin discriminator), then count 'heads'. Although this in itself has problems such as Muslim women who are completely covered, or I guess people not facing the camera will show no skin. How about using a thermal imaging camera, and select targets based on heat emissions, people usually are all at the same temperature It really depends on how much control you have over your imaging environment. For example can you obtain a template background image of the environment when there are no people present? Tell us more and we can probably come up with more detailed help Regards Dave Robinson
From: George on 3 May 2010 06:34 Hello Dave, Thank you for the interest and for your time.Basically i have to use a normal camera.So thermal method is not suitable.I havent tried others methods,basically i was looking for a code to take a look and understand the method,but i couldnt find anything.For the beggining i dont want to make a very robust method,its ok if i dont count some people.I have to start from counting the people and then disolve occlusions or stationary objects.Do you have any idea if there any code available for that kind of problem? Thank you in advance!
|
Pages: 1 Prev: Sorting Question Next: Help needed here!! Using a camera to move the mouse cursor |