For vision processing, what would you folks recommend as a colour of LED light/ring?
Green or white. (violet or orange can be ok)
there is a lot of red and blue around the field, you don’t want to accidentally detect signage.
Honestly, with a properly calibrated camera and vision processing code, any color is “the right color”.
I would choose what you have easy access to, and/or what looks best on your robot.
Agreed. Although, choosing white light may be the easiest cause you can filter out all light intensity below pure white light which is what you would get from the bounceback beam of the retroreflective tape.
Emphasis mine. Just remember, functionality > aesthetics
I agree.
Those of you that know me personally, know I’m a huge freak when it comes to the visual appearance (nothing looks as awesome as fluorescent orange spiral wrap…), but I don’t mess with the vision processing. Knowing that green is one of the best colors, I don’t care how much it clashes with red/orange, it works significantly better than other colors, so I say let the green shine!
We purchase our LEDs from superbrightleds.com, specifically using Angel Eyes. These are also what’s included on FIRST Choice. We usually use two, 60 mm and 80 mm, but are looking into adding a 100mm this year.
we’re finding that the orange we had is not popping out as quite so orange all of a sudden…
Keep in mind that many/most digital cameras actually use a Bayer pattern on their CMOS imager. This means that the cameras have twice as many sensing elements for detecting green light than red or blue.
Does it make a significant difference for our application? I am honestly not sure, but that was our thinking behind using a green LED last season.
We used a ring of 96 IR leds and an IR filter modded onto the Axis Cam last year, it was legal under the custom circutry rules and worked like a charm… White squares against black background.
Personally, I would stick away from white, just because there’s already a lot of “close to white” light in the area, coming from venue lighting that might interfere. Yes, with proper filters you can get rid of it… but you won’t be sure how good it is until you’re on the field. Using a specific color, like green, should help to eliminate that problem.
The reason camera’s emphasize green is because the manufacturer is trying to mimic the human eye’s sensitivity. Knowing this means green takes less light to reproduce the same signal. Does that make it “best” or “better”? That is debatable.
Again, if you properly calibrate your camera and vision processing, any color is fine. So, just use what you are comfortable with.
Green works. Definitely not red or blue. I’ve heard good things about purple as well. We got our ring lights here: http://www.superbrightleds.com/moreinfo/led-headlight-accent-lights/led-angel-eye-headlight-accent-lights/49/
I’m not the vision guru on our team, but the vision team is using an Orange LED strip. Not sure why.
Blue works just fine. See 254’s robot last year. Their camera tracking works well.
-Nick
We used a green LED ring from superbrightleds.com last year. Many teams at the Greater Kansas City regional used green LEDs and some used red or blue LEDs.
I think how the programmers setup the vision system has more to do with how successful the vision system is.
Blue worked very well for us last year, as did green. We preferred to aim manually though i.e. use the camera as a viewfinder. In LabView we put crosshairs over the camera feed which was incredible helpful
The green ring light is working great for us!!!’
We used a custom created ring of blue LEDs (on a custom PCB) in 2011. It worked great.
We used green last year, two sets of rings, sort of in a Venn-Diagram fashion, with the lens of the camera in the middle, which helped to get the light as close as possible to the lens. It worked well too, but we ran into a lot of problems with it at our second regional (North Carolina) due to all the sunlight shining directly in to the camera.
YMMV.
Consider that vision processing is not just about color. In a number of Michigan districts, bright LCD signs were almost directly behind the goals cycling through different colors and shapes, and it confounded all except the best vision systems. Several Michigan teams who were deadly at districts struggled at Worlds because of the LCD signs around the arena.
Remember that geomtry is equally if not more important. Length, width, aspect ratios, min and max sizes, and angles between the centroids of the objects you detect will all help you throw out bad targets and keep the good ones.
I came here to say the same thing.
This was a problem for a lot of teams at champs. The sponsor boards around the arena would cause trouble when they were the color of team’s tracking lights.
I’m thinking white (or at least high intensity white) could fix that issue. Hard to say without prototyping, though!