Suggestion for robot vision camera


I'm here once more to gather some opinions for my project. My thesis advisor got a new grant for a new autonomous vehicle based on a small ATV. The prior project, the one I'm still working on, uses webcams for image gathering. The main problem I'm having is related to dynamic range and low fps. Because of vibration caused by navigation over rough terrains, several frames from the webcam are blurred. While I could certainly create an active gimbal ala Red Team, we don't have the resources to do that. Instead, a higher fps should be enough to freeze movement under certain vibration scenarios.

We have between $800 to $1500 to spend on a higher [than a webcam] camera solution (camera+lens+cables+frame grabber). While this money buys tens of webcams, I have already realized that it could be insufficient to buy a single frame grabber, depending on the frame grabber, so I seek your advice on what other options may be viable.

Currently we are contemplating this one:

formatting link
We also need pan/tilt mechanism that will withstand the forces generated by the shocks.

If you are a dealer or a commercial institution, feel free to offer me your product.



Reply to
Loading thread data ...


My mistake. I meant a faster shutter speed, not a higher fps.

Additionally, do you know the Kowa brand? Anything to say about it?



Reply to

Here's something you can look at:

formatting link
it will help.

Reply to

You might actually consider a small camcorder with image stabilization and a video output. Cost is maybe $450, and you can get realtime USB

2.0-based MPEG or AVI framegrabbers for under $100.

Some of these have IR remotes. Maybe you could connect the pushbuttons of the remote to outputs of your control computer, for things like backlighting control, zoom (these often have zooms over 24X), etc.

The thing is these aren't Web cams, and you get 30 fps out of them, always (well, camcorders sold in North America, that is). There's no setting a lower frame rate. So, if you can't handle that frame rate, you'll need to set up your incoming filter graph to toss out frames. Or, your program may do that naturally by ignoring incoming frames until the current frame has passed through your capture filter.

An advantage of using a camcorder is that you can record the scene even as you are passing the video through the video out. Say you have a two hour tape; after a run you can analyze the tape to determine better vision algorithms. Sort of simulation after the fact.

For the X/Y head, consider a couple of good DC servos. Not the R/C kind, but real servos with tachs or encoders on them. Try Murphy's in El Cajon; sometimes you get lucky and they have something. Will set you back $100 or so a piece, and that's surplus. Or, you can opt for size 23 or even size 34 steppers, and outfit them with a pot for feedback. Unless you get some honking 1/4-scale servos with metal gears they won't handle the stress.

FWIW, the Cohu's are nice, and they're a local outfit. Also a good choice.

-- Gordon

Padu wrote:

Reply to
Gordon McComb

In case anyone wonders, this is an ad for the EyeBot controller, which is about $1000 in the U.S.

But for less than half that, I'd rather have an XBC, which you can get from . This is a pretty high-powered controller featuring an FPGA, which can be configured for high-speed vision tasks, uses a GBA for the actual brains and display, etc. Here are some links that talk more about it:

I'm not affiliated with these guys, and I don't even have one, but it sure looks like a great controller to me.


- Joe

,------------------------------------------------------------------. | Joseph J. Strout Check out the Mac Web Directory: | |

formatting link
| `------------------------------------------------------------------'

Reply to
Joe Strout

PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.