Wiring two cameras

[dlavery]

Originally Posted by jgannon

Conceptually, it's not that difficult. Get them both operational (with a little help from Kevin), and mount them on opposite corners of the front your robot. You know the width of your robot (likely 28"), and you can get the two angles from the values of the pan servos. You have angle-side-angle, and there's only one solution to the triangle. Now you know how far away you are. Pivot until the two base angles are congruent, and charge forward until you're the proper distance from the vision tetra. Use the same theory with the yellow triangle to the goal. It's very doable, and I'd love to pull it off. I can't think of a more incredible nerd feat attainable at this age than to meet Dave Lavery's personal challenge.

Actually, conceptually it is not that simple. Remember, the challenge is to come up with a real stereo correlation solution, not a simple geometric one. The biggest problem is that there are multiple vision tetras out on the field. You have to compute the true correlation between the camera images to determine if both cameras are looking at the same target. Only then can you back-solve the geometry.

If I am going to give away six dozens tasty treats without allowing any for myself, I have to make sure that the challenge is really worth it! C'mon, are you all going to leave Kevin as the only contender?

-dave

[Nitroxextreme]

Ummmmmmmmmmm....
I like Donuts

if i cant get some more funds maybe i can convince our programmer to do it
....He likes donuts too

[jgannon]

Originally Posted by dlavery

Actually, conceptually it is not that simple. Remember, the challenge is to come up with a real stereo correlation solution, not a simple geometric one. The biggest problem is that there are multiple vision tetras out on the field. You have to compute the true correlation between the camera images to determine if both cameras are looking at the same target. Only then can you back-solve the geometry.

If I am going to give away six dozens tasty treats without allowing any for myself, I have to make sure that the challenge is really worth it! C'mon, are you all going to leave Kevin as the only contender?

-dave

I think it's doable, with enough if statements. But, once again, my inspiration is not limited by lack of programming prowess, or lack of time, but rather lack of resources. Pity we don't have a mob boss *coughbeattycough* doing our fundraising, or I'd be all over this. Enjoy the donuts, Kevin.

[Kevin Watson]

Originally Posted by dlavery

C'mon, are you all going to leave Kevin as the only contender?

Dave,

Well, I've hit a snag. After an all-night Mountain Dew and pizza-fueled programming session, I've got 19 robot controllers setup as a pipelined stereo vision processor running the 27,000 lines of MER stereo vision code that I ported to PIC C18. I have an additional 53 robot controllers running a pipelined version of the MER "GESTALT" local path planner. Everything seems to run pretty well, but my EX18-12 battery completely discharges in only 5 seconds. I asked the crack FIRST engineering staff to make an exception to rule <R46> and allow me to add the additional three batteries I need, but they wouldn't budge. I then asked if I could use a flux capacitor for power, but that didn't go over too well, either (they did say I could use it for decorative purposes, though). Anyway, can I get some partial credit for the effort? I'd be pretty happy with just two dozen.

-Kevin

[ttedrow]

Originally Posted by Kevin Watson

Dave,

I then asked if I could use a flux capacitor for power, but that didn't go over too well,
-Kevin

If you use the flux capacitor your robot will still have to drive 88mph.

[Joe Ross]

Dave, is IFI giving you a kickback? For $40 out of your own pocket you get multiple teams buying $200 cameras from IFI. That's not to mention the 60+ robot controllers that Kevin bought

[NoodleKnight]

I thought the flux capacitor let you travel through time, it was the plutonium that gave that 1.21 giga-watts of power (then came Mr. Fusion, which kinda makes me wonder, where did all that extra plutonium go anyways? If it never left the car, then when the train killed the delorean, wouldn't there be a big problem?).</offtopic>

Kevin, what about a wall outlet (with an adapter, of course)? 500ft extention cord? =D

[Al Skierkiewicz]

Originally Posted by Kevin Watson

Dave,
I then asked if I could use a flux capacitor for power, but that didn't go over too well, either (they did say I could use it for decorative purposes, though). Anyway, can I get some partial credit for the effort? I'd be pretty happy with just two dozen.

-Kevin

Kevin,
The flux capacitor should have been allowed under the flowchart as long as it met the single component maximum cost. I did check with Digikey and they no longer carry the line. Apparently there is a world wide shortage of material now that we are approaching a sunspot minimum. The plutonium is another problem as it does not pass the hazardous material test and the Mr. Fusion would not either for the smell of the garbage you put in. (Inspection list "Do you have anything that smells bad or can damage the carpet?") However, you could charge it up by having all the competition attendees hold hands and shuffle across the carpet. (I am not volunteering for the last person in the line, I am going to be busy that day!) See, only simple solutions, I don't drink Dew so my head is clear.

[Mike Betts]

Originally Posted by NoodleKnight

I thought the flux capacitor let you travel through time, it was the plutonium that gave that 1.21 giga-watts of power...

I believe the unit of measure was jiga-watts vice giga-watts...

[patTeam241]

Instead of coming up with a complex stereocamera system, could you come up with a somewhat mechanical one. If memory serves me correct, you can set up the CMUcams to drive servos to center a color on the screen. If you then read the rotation with a potentiometer, couldn't you just triangulate the distance?

[jgannon]

Originally Posted by patTeam241

Instead of coming up with a complex stereocamera system, could you come up with a somewhat mechanical one. If memory serves me correct, you can set up the CMUcams to drive servos to center a color on the screen. If you then read the rotation with a potentiometer, couldn't you just triangulate the distance?

You don't even need potentiometers... you can just read the values of the pan servos. If you scroll up a ways, I talked about how you could triangulate it, and then Dave explained why it's not quite that simple.

[dlavery]

Originally Posted by patTeam241

Instead of coming up with a complex stereocamera system, could you come up with a somewhat mechanical one. If memory serves me correct, you can set up the CMUcams to drive servos to center a color on the screen. If you then read the rotation with a potentiometer, couldn't you just triangulate the distance?

Thus we come to the nut of the problem. That would work if you could guarantee that only one object of the target color could ever be in the field of view of the cameras. But since there may be up to four vision tetras visible from any location, plus the potential for background items of the same color to be detected, you will need to deteremine the true correlation of the objects in the image to determine if the cameras are looking at the same target object.

-dave

[Karma]

So in the Team Update 3, it said we could use another vision camera on our robot. Does that include any other types of cameras, such as USB? Some teams will be using Co-Processors, and thus have the capabilities to use more than two cameras, is this use even allowed? Thanks for all your help, and 955 will be the ones who win the donuts.

[Mike AA]

Originally Posted by dlavery

To the first team that successfully enables two CMUcamIIs on their robot and uses the PIC microprocessor in the IFI robot controller to compute the stereo correlation from the two camera images to determine range to the designated target, within the time limits of the autonomous period, I will personally award six dozen Krispy Kreme donuts.

Yes, it is physically and electrically possible to connect multiple CMUcams to the RC. But once you do, what are you going to do with the data?

-dave

(OK, now that I have said this, just watch someone like Kevin Watson have a working solution by this Friday... )

And yet after... 3 days of trying I still havent gotten one camera to control anything but the servo on the camera board . Back to re-reading for a 20th time.

Originally Posted by dlavery

Actually, conceptually it is not that simple. Remember, the challenge is to come up with a real stereo correlation solution, not a simple geometric one. The biggest problem is that there are multiple vision tetras out on the field. You have to compute the true correlation between the camera images to determine if both cameras are looking at the same target. Only then can you back-solve the geometry.

If I am going to give away six dozens tasty treats without allowing any for myself, I have to make sure that the challenge is really worth it! C'mon, are you all going to leave Kevin as the only contender?

-dave

I think I know how to get two involved and finding the correct tetra between the cameras but again I need to get it working with one first.

-Mike

[kham]

Two cameras? Jeez, we're still trying to figure out how to implement one camera...