What if we want to use the camera to track orbit balls?

[Bryan Herbst]

I believe that the most obvious (and intended) use for the AXIS camera is tracking trailers. They do have bright pink and green posts, after all.

But, what if we want to track the orbit balls? You may say "well, go right ahead," but each team only got one ball in the KOP. We may not be able to get more of them, and we are not sure of colors those around us got.

This really would only be useful in autonomous of course, but with all those balls flying into the crater and the inevitable large number of missed ones, I was thinking it may be more productive to use the same manipulator in autonomous that the team was planning on using for teleop to pick up the balls.

Thoughts?

[Dracco1993]

I think you will have a had time telling the balls apart. The moon rocks and the empty cells are so close in coler. I think it would be cool to try though.

[Tanner]

If you use the NI Vision Assistant to start off the code to track the balls, you can just import pictures from either the camera or your computer - hence you can download someone's pictures that have all the balls in nice quality.

I think its a good idea and all, but I think where you will most likely have the most trouble is getting a good acquire of the balls via the camera, due to the small color area, the similar colors, and the variable lighting. It probably could work and will work, but I doubt it would be reliable enough to control the robot with.

-Tanner

[Dracco1993]

You could do it with the super cells though.

[Bryan Herbst]

I agree that it would be more difficult to acquire good images on the field. We may try it out then decide whether to continue with it.

As for color similarity- We would probably either just go for the moon rocks, or if we get an ally capable of getting a super ball, that. That way, we wouldn't have to worry about similarity between the empty balls and moon rocks.

[Dracco1993]

Sounds like it might work...but if the camera thinks there are two moon rocks on the playing field but they are realy both empty cells and you pick them up then you just got penelty.

[bayesianlogic]

Unless it's against the Rules (don't know), y'might consider adding additional, cheaper cameras. There are "pet and surveillance" netcams available for under US$100.

Also, y'might look at transition edges in scans of the Moon Rocks on each color channel and use those as discriminating features.

[Dracco1993]

We have not found a rule yet that says that you can not add more cameras but we will have to see when the fourms open. If you had 2 cameras tou coulld also use trianglelation to find the distance on an opposing trailer.

[qwertyuiop[]\]

you dont need two cameras to find the distance of a trailer because you know the height of the camera,the height of the trailer, the tilt, and the pan. that is all you need and then you do some trig.

[bayesianlogic]

"... but there's always more than one way to do it!" Need? Who said anything about need? Besides, a little study of the pixellation of those Moon Rocks suggests finding a stable visual feature will be a challenge, let alone tracking them. The target on the trailer looks pretty good visually, though. Wonder how much their gonna wobble?

[Dracco1993]

Originally Posted by qwertyuiop[]\

you dont need two cameras to find the distance of a trailer because you know the height of the camera,the height of the trailer, the tilt, and the pan. that is all you need and then you do some trig.

You are right, but as we have found from tests the camera some times has a has a hard time getting the disstance corect.

P.S Sorry about my spelling.