Hi,
I'm looking for an alternative to camera based motion tracking. The system is going to be used to make a camera follow a person ( pan using a servo). Is it possible to use IR or something similar? The situation would allow for the person to wear a transmitter of some sort. I just need to track if the person ( transmitter) and determine if they are to the left or the right of the camera's position.
thanks
How fast, how accurate, how much?
What do you mean how fast, like how many updates? The accuracy doesn't need to be too much, it just needs to set the person in the frame, I would guess about 10 - 15 degrees would do it. Price isn't too much of an issue, obviously $10,000 is too much, but something reasonable would be nice.
You might do some searching for sun trackers and and similar setups. You probably make a fairly simple setup based on the type IR used for tv control and such.
The biggest problem that I see with such a tracker, zoomkat, is that when someone is walking, they would be facing in the direction they are walking, but when they turn to face the camera, they're in another direction. The person the camera is tracking would have to wear LEDs everywhere just so the tracker has something to pick up. I can't see such a device being very accurate or reliable (you might get something that does "jerky motion", but that might be the extent).
Maybe a wireless accelerometer system might work better (perhaps coupled with a compass module); basically, the person on-camera would wear such a device (for sake of argument, let's say on their belt, at the rear, in a small box), and it would transmit the readings from the accelerometer, compass, and one or two buttons (read by an Arduino/ATMega) to another Arduino using a wireless interface (xbee or rf, most likely). This Arduino would sense the motion and direction the person is moving, and move the camera accordingly.
Before starting, the user would have to have the camera positioned so that they are "in frame" (so they would use buttons and a monitor to set up the framing; or someone else could do that), then press another button to initiate tracking. The software on the camera panning Arduino would have to do the calculations and such to know what direction the user is facing, how much he has moved and when they've stopped, varying the speed of the pan to match the speed of motion, etc - to keep them "in frame".
I still don't see this as being very accurate, but it might be "good enough", and probably not jerky (probably...). Then again, there might be a good reason why camera-based motion tracking is used for this purpose...
The IR tracking that zoomkat suggested sounds like the best option, the people that will be tracked are going to be going in a predictable fashion so losing the signal wont be much of an issue, the jerky movement can likely be smoothed out. has anyone done anything similar that will give me some guidance?
The person wanting the tracker will have to explain in detail of what is expected of the person wearing the beacon. The "wireless accelerometer system might work better (perhaps coupled with a compass module..." is probably into complexity the person doesn't want to get involved in.
The beacon cant be large or expensive because of the likeliness for it breaking, getting in the way, etc. For the IR tracking, how much interference will i get from outside light?
I've seen on the internet someone using a wiimote camera surrounded by an array of IR leds. It worked somewhat with bare fingers, but with reflective tape it was very good.
Would reflective tape patches on different parts of the clothing be too much to ask?
The IR tracking that zoomkat suggested sounds like the best option, the people that will be tracked are going to be going in a predictable fashion so losing the signal wont be much of an issue, the jerky movement can likely be smoothed out.
It might work; but let me give you a scenario. Let's say for the sake of argument, the IR beacon the user is wearing is mounted on a tie clip. They stand there, the camera is on them, then they turn to walk left...
Which direction should the camera move? How does it know they turned and are walking left? Do you put another beacon on each arm? Do you put an omni-directional beacon on their head (or make them wear a hat)?
Look at it this way, too: The person is facing the camera, the beacon is too. They side-step to the right 10 feet. How does the camera know to track them in that direction? Sure, the signal from the IR beacon gets less, but it would do the same whether they moved left or right. Do you just take a guess? If you always know what direction they will move before they do, then you could use this knowledge to move the camera properly, but humans being human means that your well-laid plans will likely fail if using this model...
You could put a compass on the beacon, and based on the direction the person was facing, change the signal the beacon put out; however, you would still face the possible occlusion issue should they block the beacon with their body as the move/turn/etc.
These are issues faced every day by motion capture artists and professionals; if it was just as simple as sticking a simple beacon and easily following it, these systems would cost a heck of a lot less. This is why with inbound camera-based systems, there are always multiple cameras on the actor-in-motion, in order to catch occlusions (even so, there is a lot of data cleanup to do afterward).
If you can think of a way (that is acceptable to everyone involved) to put beacons everywhere on the person being tracked (to prevent occlusion issues), and also somehow be able to track direction of motion from these beacons using a simple system on the Arduino controlling the camera, you may have a lot of people beating down your door. I don't think it is going to be as easy, though, without using some kind of camera system (or at a minimum, a single-line CCD imager array, or maybe a linear array of IR detectors) and a lot of beacons "stuck" everywhere on the person's body.
While what I suggested before is a lot more complex than what you were probably wanting, it also probably wouldn't suffer from any interference effects, it would be small and easily concealable (as concealable as a wireless microphone), it could be used outdoors, and it would likely be fairly accurate.
I know there are DMX controllers where a person on a stage wears a little IR emitter and a system uses that to control X-Y rotation of a spotlight so that the spotlight is always centered on the person. Although I don't know much about these systems, I have seen them in use, they have some kind of "sensors" mounted on the walls on either side of the stage. My best guess is that they use an emitter on the person and multiple receivers on the walls around the stage where one receiver gets a stronger signal than another and thus figures out the location between the receivers.
Theoretically this system would use some kind of omni-directional emitter so that if the person turns around, the system still knows where they are. When I looked at one of the "sensors" it just looked like a big array of LEDs that weren't putting off any light, which leads me to believe that they were receivers. Then again it could be the other way around, the things on the walls could be emitters and the unit on the person could be determining where they are based on the emitters, this would make sense if there were multiple people being tracked.