Many posters on CD have reported past woes of having their prized computer vision algorithms fail under the bright lights of competition fields. This is not surprising as even high-end vision systems are notoriously dependent on lighting conditions. We all know that testing under the right conditions is the key to success in any engineering endeavor, so has anyone figured out a good way to simulate the kinds of lighting commonly used at competitions?
-George
George,
There is no standard for lighting at any event. Some will have color corrected tungsten while others will just use the mercury vapor or other style lighting already in place. Those venues that have skylights or direct paths to outside will also have sunlight that varies throughout the day and over the days of the event.
Probably not the answer that you’re looking for, but wouldn’t the exact lighting depend on the venue your competition is held at? I imagine you could simulate the general lighting via spotlights and such…
Darn, Al beat me to it.
The retro-reflective tape is very good at only reflecting light back to it’s source. That means if you shine a colored light that is right next to the camera at the tape, your camera should only see that color coming from the tape. A ring light that is placed around the lens works best. We’ve experimented with a ring light and even in horrible lighting conditions (by horrible, I mean bright lights all around and behind the target), the camera was able to pick up the tape quite reliably.
Make sure to read through the Vision Whitepaper settings on camera settings, it describes some ways to set the white balance and exposure settings that should minimize the effects of changes in lighting.
Thanks all for the tips. We are using the light ring that came with FIRST choice this year, and it works extremely well in our not-well-lit practice field. I’ll get a random selection of brighter lights for further testing.
-George
Spotlights. Lots and lots of spotlights.
you sir for got to mention strobe lights.