2 CMU Cameras... AT ONCE!

Today our programmers achieved something that I have not seen anyone do before. Get 2 CMU cameras up and running off of one robot controller. We recorded a bit of footage of them in action and I’ve uploaded videos in both YouTube and Google Video flavors.


Mike C.

PS. We will be replacing the board on the 2006 CMU camera in order to be in compliance with the rules. We just haven’t gotten around to it yet.

Woah! Now that is some nice stereo vision. It’s a great idea to figure out your position relative to the rack. Autonomous mode would rock with this.

First of all congrats on a job well done. How will two cameras help you out doesnt the camera(s) send a signal to the motor to approach the target? If your on the field and the cameras seek out two different sources , then what? Sorry if the questions seem stupid, its just out of curiosity.

“Yeah I’m taking a video of it to put on Chief Delphi…” - priceless…

This seems pretty cool. I’m not sure how useful it will be, but I’d love for you to prove me wrong.

We plan on using the two cameras to give us automatic scoring on not only the 4 columns underneath the lights, but allow us to even ourselves out between the two lights so we can easily score on the 4 other columns. We figured that if we can give the programmers enough time, a smooth autonomous scorer will be more effective than a manually controlled scorer.

Mike C.

Theres alot of cool things to do with 2 cameras. Your cameras both run the same search algorithm. If you have one search the left and one search the right you will effectivly cover more area and they wount both be searching in the same area.

But for the half second you save, is it worth the extra weight, processor load, and development time?

Sounds like you’ll find out in March. :wink:

There is only one serial port on the controller (aside from the programming port). How did you get both cameras to interface?

I can’t speak for certain, but I’m almost certain they did it by using the programming port. When you’re not programming, it’s just a regular UART.

Very useful in winning a control system award :smiley:

Yup, that must be it.

At one point i thought it would be cool to make a scriptng API that used te programming port as a console of sorts. I abandoned the idea wen i saw how much of a pain it was to read/write to the EEPROM.

You might want to check out the “bells and whistles” version of Kevin Watson’s camera code. He does exactly what you describe (prints to the terminal, takes input, stores new parameters to EEPROM), and from the looks of it, it’s not all that complex.

holy moly that’s incredible! did you alter the default code to make the cameras pan/tilt so smoothly?

Weight? The pan tilt assembly is already there. If they have two cameras working now I think they have dedicated people to development. I could be wrong. The processor load wouldnt be to much. But it also depends on it they are running alot of other sensors where that might be an issue. I guess we will have to wait and see what they do.

Nice work guys!
I like!

It’d be sweet to get both cameras, one to track one target, and the other to track another, and place a tube in one of the rows where the light box isn’t. Can’t wait to see it in action!

While the dual cameras are sweet and everything, I can’t help but wonder what the music playing in the back of the video is…

Any Ideas :D?

How difficult would you say it was to write code that used two cameras at once?

I’m currently trying to get the RC to talk to a single camera via the program port using Kevin’s serial_ports code… no luck so far.

The port is specified as the first option in camera.h. Other than building a suitable cable, that’s all you have to do.


Did you define the correct serial port in camera.h? And comment out the other one?

And a very important question: Are you using a null-modem cable?