Subject:
|
Re: cybermaster
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Wed, 23 Aug 2000 23:13:57 GMT
|
Viewed:
|
768 times
|
| |
|
|
try a cellular automata approach. You can do very fast edge detection etc.
using these algorithms
"Andy Gombos" <lego-robotics@crynwr.com> wrote in message
news:3989E0FE.1F75A174@yahoo.com...
> You could write a routine that instead of caputing 1 pixel, captures that
pixel, and the other 8 around it at once. Then you can see if the value of
the pixels change
> from one to the other signifigantly.
>
> 011
> 011
> 011
>
> In this simple example, you would be able to detect that the 1's (white)
stopped, and then knew that the paper stopped on that line for at least 3
pixels.
>
> Andy
> Laurentino Martins wrote:
>
> > At 16:55 03-08-2000 Thursday, Marco C. wrote:
> > > Hi Laurentino :)
> > >
> > > As you know (if you remember) I also have a CyberMaster.
> > >
> > > I prefered the RF-link to the IR-link, primarily because of the • mobility
> > > and freedom it would give to the robot, and the processing power of the • PC.
> > >
> > > My current (looong) project is exactly that, of using Robotic Vision as • a
> > > component to the robot guidance system. Landmark recognition by vision,
> > > crude obstacle detection by analising the glare intensity and distance
> > > between a pair of small flash lights mounted on the robot and • everything
> > > else I might come up to, in the process.
> >
> > My original idea was to create a path of while sheets of A4 paper in the • floor and make the bot follow the path.
> > The floor is dark, so the white sheets can be easily seen.
> >
> > The main problem is what to do with all the flow of graphics the camera • captures.
> > Animals do parallel-processing when analising the contents of an image • but a PC program usually don't.
> > The result is that every time I'm analising a pixel I don't see the
pixels around it, so I can't draw conclusions about that pixel until I look
at the other pixels.
> > The result is a slooooow program that has almost no use. When it reaches
a conclusion, it's usually too late to be of any use. That's my main
problem.
> >
> > I'm trying to create some algorithm (fuzzy logic?) that can draw some • conclusions without much processing.
> > One of these days I have to sit down and do some brainstorming on the • subject ...
> >
> > > I don't know if you ever saw any of my postings about this, but, since
> > > you're using MS VC++ and dealing with realtime image processing, have • you
> > > ever used a very good image/video-processing lib by Pong Suvan ?
> > > http://www.ccs.neu.edu/home/psksvp/download.htm
> >
> > Nope. I have to check that out.
> >
> > I've already done some routines of my own to improve the image clarity: • auto-equalize color/brightness, blur, sharpen, edge enhance, etc.
> > Auto-equalize can even improve images that at first look like to be • totally black (like under my bed at night).
> > This is possible because the camera outputs 24bits color and captures • many nuances in the image.
> >
> > > The movingblob() was one I planed to use on a sub-project: "Target"
> > > following, with the robot simply controlling the up/down, left/right
> > > rotation of the cam.
> > >
> > > If you didn't already, check this list:
> > > http://www.cs.cmu.edu/afs/cs/project/cil/ftp/html/vision.html
> > >
> > > My main problem right know: the european (PAL) version of XCam I was
> > > planing to use for wireless color feedback to the PC (through my • video-in
> > > card), costs 249.00 EUR (www.edentronic.com) = $225.25 USD = 56000$00 • PTE
> > > (at www.xkt.pt) instead of the original NTSC version, $88 USD • (www.x10.com)
> > >
> > > :(
> > >
> > > Guess I'll have to go for a cheaper option instead... an USB webcam. :/
> >
> > I use a Logitech camera with mic and a 5m USB extension.
> > The cable is not very pleasing to see around in the floor, but the bot • doesn't seem to mind a lot :-]
> >
> > Check the attached foto
> > (sorry the quality)
>
>
------------------------------------------------------------------------
> >
> > Name: Rover auto-foto.jpg
> > Rover auto-foto.jpg Type: JPEG Image (image/jpeg)
> > Encoding: base64
>
>
------------------------------------------------------------------------
> >
> > Laurentino Martins
> >
> > [ mailto:lau@mail.telepac.pt ]
> > [ http://www.terravista.pt/Enseada/2808/ ]
>
>
>
>
>
|
|
Message is in Reply To:
 | | Re: cybermaster
|
| You could write a routine that instead of caputing 1 pixel, captures that pixel, and the other 8 around it at once. Then you can see if the value of the pixels change from one to the other signifigantly. 011 011 011 In this simple example, you would (...) (24 years ago, 3-Aug-00, to lugnet.robotics)
|
4 Messages in This Thread:
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
