Position Tracking Devices

[dhitchco]

Well,
to "track" a moving object you need to have at least two people or devices "watching" the moving object. In your case, you can track a person on a basketball court by having only two observers, since there really is no 3rd-dimension inthe game.

In such a relatively-small space as a basketball court, you can't track the player via a technology such as GPS. But, you do have a few choices for technologies. First, the most obvious is to use radio frequency "beacons" such as cell phone that can emit a "beacon" which can be triangulate the player's position on the field. Even easier would be to use a 900Mhz low-power belt-worn transmitter and two receivers to triangulate the player's position. You could also do the same thing via light (e.g. lasers) but they would require "aiming" the light at the player each time you wanted to triangulate.

One other technology that may work is to use pixel-movement detection on a ceiling-mounted video camera. This is a technology that is being used in the surveillance industry.

Now......put away all of the technology and use humans as the observers. Divide the basketball court up into squares (3'x3') and then simply "call-out" that player "A" moved from square 17 to square 22 at time 00:00:04 or something like that.

I've often wanted to use tracking technology to monitor where runners are within a marathon, and stock car drivers use their in-car radio system to pinpoint their location ont he track, but it's all relatively expensive.....Plus, once you get the data, how do you want to use/track the player's movements? If you intend to keep their "traces" then you need a multi-record database to draw a map of the player's movements.

[Dave Flowerday]

Originally Posted by dhitchco

First, the most obvious is to use radio frequency "beacons" such as cell phone that can emit a "beacon" which can be triangulate the player's position on the field. Even easier would be to use a 900Mhz low-power belt-worn transmitter and two receivers to triangulate the player's position. You could also do the same thing via light (e.g. lasers) but they would require "aiming" the light at the player each time you wanted to triangulate.

There's basically two ways to triangluate if you have some sort of emitter on the person and two receivers off-field. These receivers either have to be able to measure the angle (or bearing) to the player from each sensor position OR they each have to be able to measure the distance to the player. And of course the distance between the receivers and their exact location needs to be known in advance. When you talk about using RF (cell phone or 900MHz), are you thinking of measuring the distance or the angle? Measuring angle would be tricky but theoretically possible I suppose, perhaps using a highly directional antenna and sweeping it back and forth across the court. Distance would be nearly impossible. To measure distance you'd have to measure the time-of-flight of the radio signal and if you're talking about measuring things in the range of 10s to 100s of feet you'd need extremely accurate timing. RF travels at roughly the speed of light, meaning that it travels 1 foot every nanosecond. So, you'd have to be able to measure the time of flight of the radio transmission with accuracy down to the nanosecond to be able to measure distance with about a foot of accuracy.

I suppose you could try to use signal strength as a way of estimating distance from the sensor, but that has many problems - other players blocking the signal would make the player appear father away, the difference in signal strength between 10 feet away and 20 may be too small to measure, etc.

You'd also have to contend with reflections and interference and an assortment of other problems. The bottom line is that there's a lot of good theory to be explored in solving this problem, but a practical implementation is extremely difficult.

[dhitchco]

Hey Dave,
good dialog. I work for Harris Corporation's RF Communications Division in Rochester NY.

Most of us in the RF world talk about "time of flight" as a way in which to determine bearing and distance to a target. basically, send out a pulse (like a bat), wait for the return signal, and determine the "time of flight" to get distance. Hopefully, you use all kinds of tricks to ensure that you're receiving the actual signal from the player, not a reflection (multipath signal) off of something else. Each player on the court would have his/her own frequency ID, so that's not a problem, and all players would 'beacon' every 500ms or so to the same set of three or four courtside receivers. A PC would do the time-of-flight calculations for each player's signal and do the vector triangulation.

ps: I don't know why professional sports haven;t gone to such a system to "track" a play (e.g. football playbook instead of John Madden's chalkboard)


TIES TO FIRST ROBOTICS:
As an aside, I've often wondered if the autonomous period for robots could be improved by 'triangulating' the robot position off of TWO (2) light sources...but, we have enough trouble focusing-in on one source as it is..... Maybe I can scrounge-up a garage business of some cheap RF beacons for bots in the future!

[KenWittlief]

It seems like the easy approach to track one player on a basketball court would be a video camera looking down from the ceiling.

There are tons of video/camera inspection SW programs used in many industries, I think it should be possble to get something off the shelf that can analyze a video image stream, and track a moving object - esp since the rest of the field of view will be stationary (the floor of the basketball court).

The only place it might have difficulty is when the player shoots the ball. You might need the player to wear a unique color hat or shirt, to discriminate between the player and the moving ball.

[seanwitte]

I don't know if there are commercial versions yet, but Dave Lavery has mentioned research in Carrier Phase Differential GPS psuedolites for local positioning. A GPS unit uses the psuedolites (specialized GPS transmitters) to determine centimeter-level accuracy within a small localized environment. As the goog (http://www.google.com) for more.

[Kevin Watson]

Originally Posted by dhitchco

...As an aside, I've often wondered if the autonomous period for robots could be improved by 'triangulating' the robot position off of TWO (2) light sources...

This ability was included in the 2004 season. Software and some documentation can be found here: http://kevin.org/frc/2004/

Originally Posted by dhitchco

...but, we have enough trouble focusing-in on one source as it is.....

Are you having problems with the CMUcam?

-Kevin

[dhitchco]

Ahhhh..now I get it!

If you only want to "track" one player on the court and the function is to 'return' balls to that player...well, heck this is just like FRC's Aim High in reverse.

Just put a green shirt on the player, which is tracked by a camera mounted on the backboard. That camera can keep track of the player whether he is inside the paint, or outside in the 3-point zone.

Now, under the basket, just mount a chute that can be swiveled left-right and angled up-down and use the software to aim that chute back at the player.

Wow, you might be able to shoot almost 100 free throws per minute that way....I'm getting tired thinking about it.