RF Position tracking?
October 9, 2008 10:59 AM Subscribe
Anyone know of a reliable way to quickly & accurately track position over a 200 foot range? I am trying to get a camera to follow a moving subject over medium range distances outdoors.I have seen several products on sparkfun that transmit/receive RF, but do not know if they will work for position tracking.
THANKS!
Look into robot sentry guns. There are quite a few homebrew projects to make paintball, BB, or watergun versions of the sentry guns from Aliens (and numerous sci-fi video games). Replace the gun with a camera, and you have something pretty close to what you're looking for.
Most of them work via a fixed webcam with software tracing the path of the moving object.
Random link
posted by -harlequin- at 11:19 AM on October 9, 2008
Most of them work via a fixed webcam with software tracing the path of the moving object.
Random link
posted by -harlequin- at 11:19 AM on October 9, 2008
Really depends on how complicated (vs. inaccurate) you want to make this. I'm just going to list a bunch of genres here:
1) Triangulate based on signal strength -- target emits infrared light or RF signal, you have 2 or more detectors that compare received signal strength to triangulate target position. Gets fouled by a number of real-world troubles (eg, RF transmitter is in coat pocket, has straight line to detector A but has to go through person to get to detector B, over-attenuating that signal).
2) Target has differential GPS receiver with base station as reference, and transmits accurate lat/lon back to base station.
3) Base station uses sonar/lidar/etc. to track targets
4) Base station analyzes incoming video stream, moves camera to keep moving pixels in center of field of view
There are certainly even more ways of doing it. Ultrasonic rangefinders would be lovely, but the ones Sparkfun sells have a max range of a few meters. Most quick 'n dirty robots use method 1 in one incarnation or another. Both methods 1 and 2 require that your target is a friend of yours and cooperatively straps on your tracking gear. In theory you could use something like triangulating RFID signal strength, but 200' is way too far for that to work.
Most techniques that you'd use for a DIY project are going to be inaccurate enough (especially if the target is moving quickly) that you'll probably want to compensate with a wide field of view for the camera.
Note that the robot sentry project is probably the best accuracy you can get (and uses method 4), and that you'll need one very-wide-angle camera (to see the entire "playing field" at once) to use to target a second, probably zoomed camera.
posted by range at 11:26 AM on October 9, 2008
1) Triangulate based on signal strength -- target emits infrared light or RF signal, you have 2 or more detectors that compare received signal strength to triangulate target position. Gets fouled by a number of real-world troubles (eg, RF transmitter is in coat pocket, has straight line to detector A but has to go through person to get to detector B, over-attenuating that signal).
2) Target has differential GPS receiver with base station as reference, and transmits accurate lat/lon back to base station.
3) Base station uses sonar/lidar/etc. to track targets
4) Base station analyzes incoming video stream, moves camera to keep moving pixels in center of field of view
There are certainly even more ways of doing it. Ultrasonic rangefinders would be lovely, but the ones Sparkfun sells have a max range of a few meters. Most quick 'n dirty robots use method 1 in one incarnation or another. Both methods 1 and 2 require that your target is a friend of yours and cooperatively straps on your tracking gear. In theory you could use something like triangulating RFID signal strength, but 200' is way too far for that to work.
Most techniques that you'd use for a DIY project are going to be inaccurate enough (especially if the target is moving quickly) that you'll probably want to compensate with a wide field of view for the camera.
Note that the robot sentry project is probably the best accuracy you can get (and uses method 4), and that you'll need one very-wide-angle camera (to see the entire "playing field" at once) to use to target a second, probably zoomed camera.
posted by range at 11:26 AM on October 9, 2008
Note that the robot sentry project is probably the best accuracy you can get (and uses method 4), and that you'll need one very-wide-angle camera (to see the entire "playing field" at once) to use to target a second, probably zoomed camera.
You could do it with a single camera under some conditions. Let's say your target has a red jacket on; the camera zooms out pans/tilts around until something red comes into view. Then you pan/tilt so it is in the middle of the picture. Then you zoom to whatever level you want, and you track the red thing so that if the red thing is at the right of the picture the camera pans right until the red thing is in the middle of the picture, if it's to the left you pan left, and so on.
On the other hand if you have a stationary camera you can detect motion instead of colour to track your target, which might work better if your application demands it.
Needless to say, having a person operate the camera is easier than setting up a homebrew machine vision system!
posted by Mike1024 at 1:26 PM on October 9, 2008
You could do it with a single camera under some conditions. Let's say your target has a red jacket on; the camera zooms out pans/tilts around until something red comes into view. Then you pan/tilt so it is in the middle of the picture. Then you zoom to whatever level you want, and you track the red thing so that if the red thing is at the right of the picture the camera pans right until the red thing is in the middle of the picture, if it's to the left you pan left, and so on.
On the other hand if you have a stationary camera you can detect motion instead of colour to track your target, which might work better if your application demands it.
Needless to say, having a person operate the camera is easier than setting up a homebrew machine vision system!
posted by Mike1024 at 1:26 PM on October 9, 2008
I agree that range's option (4) is probably the one for you. After that, (2) since you're outdoors.
posted by TheOnlyCoolTim at 1:53 PM on October 9, 2008
posted by TheOnlyCoolTim at 1:53 PM on October 9, 2008
google "rf ultrasonic cricket"
Too many links to mention.
Concept is a polled beacon... RF signal sent to beacon. Beacon emits Ultrasonic noise. Multiple receivers receive time of flight info from beacon. Triangulation resolves position.
It's limited in accuracy and not passive, but in all the positioning problems I have investigated, it is one of the more interesting ones.
Gotta run or I'd reply longer.... good luck.
posted by FauxScot at 2:10 PM on October 9, 2008
Too many links to mention.
Concept is a polled beacon... RF signal sent to beacon. Beacon emits Ultrasonic noise. Multiple receivers receive time of flight info from beacon. Triangulation resolves position.
It's limited in accuracy and not passive, but in all the positioning problems I have investigated, it is one of the more interesting ones.
Gotta run or I'd reply longer.... good luck.
posted by FauxScot at 2:10 PM on October 9, 2008
This thread is closed to new comments.
posted by o0dano0o at 11:17 AM on October 9, 2008