PDA

View Full Version : Human Path Detection



vasanth
10-20-2005, 02:24 PM
Hello Everyone,
Long time since I came here... I require your valuable suggestion/ideas for the below propsed system.


The purpose of the system is to detect/map the path of human/humans in a given space.. I belive that 2 camers will have to be used for this purpose as shown in the image.

The known vlues will be angle CAB,ABC and the distance X. I believe that the position of the human can be calculated by using the laws of sines. But the problem with this system in the farther the human is the less acurate his path is mapped.. Are there any better alternatives to this idea... And is there anyway I can calculate the actual height of the human based on the above data...

regards
vasanth

JaWiB
10-20-2005, 03:46 PM
Unless I'm overlooking something, it seems like you should be able to find their height if you can find the angle of elevation to the top of their head, and the distance to the top of their head. Like something like this, where everything except h is known (distance to top of head you can already find based on what you posted, right?):

vasanth
10-20-2005, 06:30 PM
Unless I'm overlooking something, it seems like you should be able to find their height if you can find the angle of elevation to the top of their head, and the distance to the top of their head. Like something like this, where everything except h is known (distance to top of head you can already find based on what you posted, right?):

thanx.. But i dont think I can acurately measure the distance from the head.. cant this be done by using the height of the person on image and and approx distance of the person from the camera and the focal length of the lense...

I will see if i can get the distance from the head reliably..

JaWiB
10-20-2005, 07:03 PM
cant this be done by using the height of the person on image and and approx distance of the person from the camera and the focal length of the lense...


Hm that didn't even occur to me. I suppose there might be a way to do that, I don't really know, sorry. It sounds like it would be more complicated to do that...

What are you trying to do this for anyways?

novacain
10-20-2005, 09:50 PM
When I did image processing for trains, we had the cameras take images at set intervals (without the train).

This allowed us to calculate a scale/ratio from an object at a known distance (ie railway track) or a known size.

For speed cameras we used the lines painted on the road to determine scale. Can be done once when the system is installed then recalibrated as required.

vasanth
10-21-2005, 12:22 PM
thank you both.. i will try the novacain's "reference" idea.




What are you trying to do this for anyways?

I am building a human path maping and profile matching system for my fianl year university project. Which involves plotting the path of human/humans in a given space along with identifying key aspects such as their hair colour, kind of jacket worn etc...