CMU and 2008

[King Duke]

Hey Guys,

I was wondering if anyone has any plans of using the CMU camera and if you guys do, would you mind giving us a vague idea of what you're going to use it for.

Thanks,

[Chris_Elston]

Try to see the difference of RED BALL and BLUE BALL on the ground.

Dunno yet...still testing it...

Originally Posted by King Duke

Hey Guys,

I was wondering if anyone has any plans of using the CMU camera and if you guys do, would you mind giving us a vague idea of what you're going to use it for.

Thanks,

[sloteera]

Try to see de white circles on the ball...
I guess that theses circles would be reflexives...

We gonna try on the tests, but we haven't the KOF. We'll got de KOF only in Frebruary... =/

[King Duke]

we were thinking similarly

[Dr Nick]

I think it would definitely be helpful to be able to have the robot find the trackballs on its own so we're going to play around with the CMU Cam and if that doesn't work, might look for an alternative without on-board image processing.

[King Duke]

As anyone tried doing a physical filter over the lens? For example if you were on the red alliance you'd put a blue filter over the lens, thus eliminating the ID of the blue trackballs, and give the red track balls a distinctive purple color when compared to the silver and gray of the field.

[Uberbots]

Thats a very good idea, Duke.

we are reprogramming our camera to recognize multiple objects... its going well.

[bear24rw]

Originally Posted by Uberbots

Thats a very good idea, Duke.

we are reprogramming our camera to recognize multiple objects... its going well.

With the cmucam?

[Bharat Nain]

Jeez, when this season is over you guys are going to have to teach the rest of us how you recognize objects with the CMUcam. I'd be greatly interested if you don't mind.

[sweep4er]

Interesting thoughts...Of course, this was our first consideration as well. We're planning on testing further, but were wary of the camera's ability to distinguish between the non-illuminated objects with ambient lighting above.

At the moment, we are investigating its use in hybrid mode as a way to keep the IR receiver in line with our remote. The RoboCoach will be able to hold a flashlight tinted with a color the cmu will be trained to track. IR receiver mounted on the camera...you get the idea.

[Uberbots]

Originally Posted by bear24rw

With the cmucam?

yup. it works... to an extent. i have some toying to do with the algorithm with regards to noise filtering and color balance... or using different color coordinates like HSV.

it doesnt 'recognize' objects per say... but it is able to pick different objects apart and return objects that fit into query parameters to the master controller using a tolerance-differentiating algorithm.

[Andrew Schreiber]

I know that 397 will be running some experiments on locating the balls. Im thinking looking for the big red or blue objects is much easier than tracking white dots.

Originally Posted by Uberbots

yup. it works... to an extent. i have some toying to do with the algorithm with regards to noise filtering and color balance... or using different color coordinates like HSV.

it doesnt 'recognize' objects per say... but it is able to pick different objects apart and return objects that fit into query parameters to the master controller using a tolerance-differentiating algorithm.

Any recomended reading for those of us who have an interest in this both in competition and outside of it?

[Nathan]

Originally Posted by King Duke

As anyone tried doing a physical filter over the lens? For example if you were on the red alliance you'd put a blue filter over the lens, thus eliminating the ID of the blue trackballs, and give the red track balls a distinctive purple color when compared to the silver and gray of the field.

We've been thinking about doing that. We're hoping to go to a camera store and try a bunch of different filters...

Have you actually tried this yet?

Nathan

[Andrew Schreiber]

For filters could you just use red/blue cellophane? I mean you would probly lose some quality but the color should still be there. And for testing purposes I'm sure this could prove the concept. Finding cellophane, or any tinted thing should work and be cheaper in some cases than buying a filter.

[Nathan]

Chris (my programming mentor) went to a camera shop and tried a bunch of filters. You'll have to ask him to get the details, but from what he said I don't think they worked very well.

Nathan