active tracker development at PLF
Posted: Wed Jan 03, 2007 4:59 pm
I've been experimenting with my old trackers and some new techniques. I've never used them with natualpoint cameras before, so this was all new to me. I had been using them with standard DV camcorders previously.
I'm using 920nm LEDs that take about 1.5 volts and 50milliamps max. The camera seems to see them fine at this wavelength though I should probably go find some 850nm LEDs to compare.
Using the sanding technique, I've been able to get about 90 degrees of visibility at a distance of about 10 feet. At this distance, both movement speed, and angle of incidence have enough effect as to make the track unreliable outside those parameters.
I have tried an array of 4 LEDs (parallel) closely packed but have the same results. I had hoped that having a tracking area 4x as large, would result in better performance. However, as the array goes off angle, its larger surface area diminishes down to a mere 2x and the LEDs lose their throw. So, no better results for now.
I have tried using the array to light the interior of a particularly thin ping pong ball. However, it appears the LEDs are still too directional, and the ping pong ball does not seem to bounce light around all that well. I got almost identical results in fact.
There are two thing that could be tested at this point, with what I expect to have for the next few weeks. First, I can try an array of non-parallel LEDs where they are tilted out from the center say... 45degrees each. This should help with the incidence issue significantly.
I expect to receive a few samples of some wide throw LED's from Jim. I will need to see how they compare both due to their wider throw, and due to their better wavelength match.
I suspect, that I also need to play with the thresholding and exposure of the camera to find the optimal performance parameters and design. However, I think I'll hold off on that until I have a c:120 rather than these trackir cameras. If nothing else, the ability to see grayscale will help significantly in judging exposure, before thresholding is applied.
This is actually rather promising. If I'm getting 10' without tweaking, then I suspect I'll get much better once everything is sorted out.
-brad
I'm using 920nm LEDs that take about 1.5 volts and 50milliamps max. The camera seems to see them fine at this wavelength though I should probably go find some 850nm LEDs to compare.
Using the sanding technique, I've been able to get about 90 degrees of visibility at a distance of about 10 feet. At this distance, both movement speed, and angle of incidence have enough effect as to make the track unreliable outside those parameters.
I have tried an array of 4 LEDs (parallel) closely packed but have the same results. I had hoped that having a tracking area 4x as large, would result in better performance. However, as the array goes off angle, its larger surface area diminishes down to a mere 2x and the LEDs lose their throw. So, no better results for now.
I have tried using the array to light the interior of a particularly thin ping pong ball. However, it appears the LEDs are still too directional, and the ping pong ball does not seem to bounce light around all that well. I got almost identical results in fact.
There are two thing that could be tested at this point, with what I expect to have for the next few weeks. First, I can try an array of non-parallel LEDs where they are tilted out from the center say... 45degrees each. This should help with the incidence issue significantly.
I expect to receive a few samples of some wide throw LED's from Jim. I will need to see how they compare both due to their wider throw, and due to their better wavelength match.
I suspect, that I also need to play with the thresholding and exposure of the camera to find the optimal performance parameters and design. However, I think I'll hold off on that until I have a c:120 rather than these trackir cameras. If nothing else, the ability to see grayscale will help significantly in judging exposure, before thresholding is applied.
This is actually rather promising. If I'm getting 10' without tweaking, then I suspect I'll get much better once everything is sorted out.
-brad