Movement tracking - maximum distance / range?

Projects General Guidance
1

Dear All,

I am totally new, hoping to get some reality check feasibility analysis before diving into the coding world.

There are some face tracking projects using Arduino + Servo + openCV + webcam (here or here), but the motion seems to be limited in a room in general.

Do you think it's possible to track a moving object in an area as big as 200ft x 100ft (30m x 60m)? Bearing some IR receiver or bright color as needed? Assuming the camera is placed in the center of one of the long sides, the maximum distance between object and camera is 100ft (30m). In case you are interested, I'd like the video camera automatically follow a figure skater on a full size indoor ice rink...

The major question is maximum detection distance / range, and I was hoping to find out which part is the limiting factor: webcam, openCV, or arduino?

Thank you very much in advance and looking forward to hear from you! In case the project is feasible (fingers crossed), some useful links would be MUCH MUCH APPRECIATED. XD

2

Your basic question of having a camera track somebody is frequently asked. I have not seen anybody provide a realistic solution that would not be complex and expensive. You can do a google advanced search of Arduino Forum for key words like "trackibg" to see previous post.

https://www.google.com/search?hl=en&as_q=tracking&as_epq=&as_oq=&as_eq=&as_nlo=&as_nhi=&lr=&cr=&as_qdr=all&as_sitesearch=http%3A%2F%2Farduino.cc%2Fforum&as_occt=any&safe=off&tbs=&as_filetype=&as_rights=

3

Thank you zoomkat for your fast reply! I did search the forum first but focused mostly on maximum distance.

I thought that since face tracking seems resolved, tracking a receiver or specific color on a moving person would be essentially similar in difficulty. Or is it not even feasible?

Thank you again!

4

I've been able to track in IR laser spot in the dark. But that's completely different!

5

jjane45:
The major question is maximum detection distance / range, and I was hoping to find out which part is the limiting factor: webcam, openCV, or arduino?

I doubt the Arduino is a limiting factor for distance/range, since all it would be doing is taking input and moving the pan/tilt unit (if I am understanding the setup right). As far as OpenCV - well, if you were tracking an active IR marker, provided the PC the software was running on was fairly well built, fast, memory etc (ie, optimized for OpenCV), that wouldn't be an issue likely, either.

That leaves the camera. What you want is something with high resolution, and a fast frame rate. If you wanted to do it right, you would invest in a firewire HD camera, and use OpenCV with that (not a cheap webcamera). You would want to mount it on a solid pan/tilt platform; you could probably do it with the right servos (maybe something from Servo City would work), but better would be a commercial unit employing steppers or other DC motors (and drop the Arduino, as such units typically are controlled by serial/usb/ethernet).

Finally, to make the tracking easiest, you would use either an active or passive IR marker(s) on the subject; multiple ones would be needed to deal with occlusion, then the markers subtracted out of the frame; to the software they would "appear" as gaussian blobs, so you would model the blob to find the approximate center, then use the multiple marker positions found to average together to find the average "center", and position the camera to that. Something like that would be your algorithm; likely OpenCV already has something built in for this that would work (or someone has already done something similar).