Camera to use for Vision Processing

[Eryl2000]

Hi,

We are trying to vision processing, but we're not sure which camera is best to use. Last year we implemented vision control, but we had a lot of problems getting consistent results. The images we got were fairly whitewashed, which might be because our ring light was not bright enough, but I don't know if a brighter ring light would have even helped. The images were not really good enough to properly edge find with. I heard that changing the exposure on the camera can help fix this problem, but I'm not sure which cameras can do so and which ones do it best.
Any help would be appreciated!

[NShep98]

What kind of camera and ring light were you using before? A pretty standard setup could be this camera and this ring light. This is not the only option, but it's what my team is going with.

[Eryl2000]

Originally Posted by NShep98

What kind of camera and ring light were you using before? A pretty standard setup could be this camera and this ring light. This is not the only option, but it's what my team is going with.

We used those exact parts last year with minimal success. We also used the Axis 206. We weren't sure how to set the exposure (if it is even possible on those).

[Dale]

I'm not sure about being "best" but a lot of teams use the Microsoft HD3000 camera probably because it's often times available in the KOP or FIRSTChoice. We've used the Axis cameras because that's what we had. It's a far more expensive camera with a built in web server so it connects via ethernet.

There is a lot of support for both of these on Chief Delphi. I'd be interested if someone has found a camera that's better for FRC applications.

Yes, it is vital that you be able to control the exposure. You want to turn it WAY down so that your camera only sees the tape and other really shiny things.

A few other tips. Don't try to image process while the robot is moving (at least in the beginning.) Many teams use a image co-processor but it may not be required depending on what you are trying to do. The Cheesy Poofs, 254, have a lot of resources on how their image processing works. You could do worse than studying their materials carefully.

[Hadi379]

I thought the pixy was easy to get working, in all honesty we probably over thought the entire setup from the get go, but it turned out pretty simple so far.

[euhlmann]

Last year we used a Logitech C930E. It has decent exposure and white balance control, and (the reason we picked it) a wide field of view.

[gerthworm]

Originally Posted by NShep98

and this ring light.

All the green smaller rings are out of stock till April.... I guess amber might be another good option? Anyone had luck with those?

[rudun]

They are available on black circuit boards also.https://www.superbrightleds.com/more...it-board/1117/

[beijing_strbow]

We're using a Microsoft LifeCam HD-3000 for vision processing this year and it seems to be working fairly well (that's my general impression - I'm mostly working on something else). They're very cheap and easy to work with, considering that the WPI libraries were built for them.

You can fix the whitewashing by turning the exposure and brightness way down (I think we have it set at 1% on both of them, and it works pretty well).

[Billfred]

Originally Posted by gerthworm

All the green smaller rings are out of stock till April.... I guess amber might be another good option? Anyone had luck with those?

Garnet Squadron ran an RGB ring light at Palmetto last year. If I remember correctly, we actually introduced a little pink to ours and were able to track the reflective tape at home. (At the event, we quickly realized our RasPi had a glass jaw.)

[Dale]

All the green smaller rings are out of stock till April.... I guess amber might be another good option? Anyone had luck with those?

We've used amber in the past with mixed results. We're switching to green this year. Amber seems to be a more common color produced by house lights, etc. We were getting more noise than we do with green.