|
|
|
![]() |
|
|||||||
|
||||||||
Waiting for next years game...
02-10-2006 17:06
Greg Marra
As a programmer, the mere thought makes me cringe.
Have you considered interlacing the lights (RGBYRGBY instead of RRGGBBYY) so that the color will be more evenly spread out?
02-10-2006 17:27
Noah Kleinberg
I didn't give it too much thought, but I guess that I thought it might not be intense enough for the cmucam if they're spread out, and this one can still sorta be used as a just green light in case they don't actually do more than one color.
I got the wires all sorted out and the filter on, so all that's left is the code...
02-10-2006 18:38
Schnabel
Hey, now you are 1 pixel away from the biggest megatron made! 
02-10-2006 19:20
chris31Nice. I might be making one to test some off season code on.
03-10-2006 09:18
Qbranch
|
Originally Posted by Greg Marra
As a programmer, the mere thought makes me cringe.
Have you considered interlacing the lights (RGBYRGBY instead of RRGGBBYY) so that the color will be more evenly spread out? |
03-10-2006 18:56
Greg Marra
|
Originally Posted by Qbranch
Actually... changing color configurations isn't as hard as it might look... as a hint, i suggest you look at Kevin's non bells and whistles version of the camera code.
|
03-10-2006 23:23
Alan Anderson
|
Originally Posted by Greg Marra
Changing the camera configuration is easy: getting the camera to lock onto targets in different lighting conditions is a pain. My experience has been that the CMUCam is very difficult to get just right between false positives and false negatives and can be extremely finicky.
|
03-10-2006 23:33
GdeaverThe cold cathodes are OK but I'd like to see what the camera would do with some traffic light LED's. The intensity is such that the camera should have no problem latching on. In the off season, try putting some UV blocking eye glass lenses on the camera and see the difference they can have
04-10-2006 13:14
JamesBrown
|
Originally Posted by Alan Anderson
My experience is exactly the opposite. If you work in RGB color space, it will tend to match similar light levels more easily than specific colors, but if you change it to YCrCb space (as the code provided last year does) it detects and tracks glowing color targets very well.
The only "false positives" we got when looking for the green cold cathode box were from daylight shining through windows at the side of our practice field. As for "false negatives", we did have a couple of hours once where it would lose track of the target at distances it usually had no problem with, but that was because the camera had been misfocused. Tweaking the focus back where it needed to be cleared up the problem immediately. |