1. Home
  2. General Robotics Forum
  3. visual processing

visual processing

After reading the Mike Simpson's article on interfacing optical mouse chips to a microcontroller, I got it up and running on a DIOS chip and a PSOC chip. It works nicely, I can download pictures at about 4 frames a second and read all the registers. I'm planning to rig a lens to it so that I can see things a little further out than a few millimeters.

My problem is, how to process an image to give me useful information. The Agilant chip only has an 18x18 bit image field, so that's only 324 bytes of data and shouldn't be too much for a small MCU to process. However, I haven't got the faintest idea of what to do with it. Any ideas?

chris in napa

Reply to
chris
Loading thread data ...

I have been working on line following.

It's 16x16 pixels, 6 bits per. The minimal amount of data is still useful. That few pixels would be suited to artifical neural nets, for instance.

For lens ideas, check out my articlw in N&V.

Reply to
Blueeyedpop

chris wrote: [..]

Nice setting, not to much information. You could:

1: find the brightest spot and move in that direction. 2: the darkest spot....

You can calculate a difference frame to see where something is moving.

3: move in the direction with most movement, 4: move in the direction with less movement.

etc.

or a changing combination.

Good luck,

Peter

Reply to
Peter van der Vos

I read that article. I thought that was you - it was the isopod that gave you away. I'm going with a pinhole lens. I used to make pinhole lens microscopes for my 7th graders years ago.

The Agilent chip - I got a 2610 out of a GE mouse - has only 8 pins and does a fair amount of data processing itself. It can return X and Y motion, high pixel/low pixel levels, average levels as well as a picture dump. I was amazed at the picture quality.

I'm looking for information on how to process the image. The CMU CAM can track an object, can find the center of a mass. Can I do edge detection? Can I recognize an object? I can't seem to find the basics of how to approach this.

chris in napa

Blueeyedp> I have been working on line following.

Reply to
chris

Where do I find the N&V publications and what does N&V stand for?

-JC

Reply to
John Casey

Where do I find the N&V publications and what does N&V stand for?

-JC

Reply to
John Casey

There are two magazines which might interest you, SERVO, and Nuts N Volts.

formatting link

Reply to
Blueeyedpop

I posted my mousecam research stuff,

formatting link
is an undocument xls file there, but I am home sick with a deadline and a fever, and don't have time to filter through my work, nor do I remember exactly what I was doing. Sorry

it basicly involves the sum of pixel values times their position, divided by their sum to find the centeroid.

Mike

Reply to
Blueeyedpop

Of course Nuts & Volts. Thanks you for the response.

-- John

millimeters.

information.

Reply to
John Casey

Looks good. Could you tell us how you changed the focus so you could image your hand? With an extra lens, or replacing the mouse lens?

-Daniel

Reply to
Daniel Watman

I have done both. Using an ancillay lens, treating the mouse' lens as a relay, and using a pinhole.

More control with the pinhole, wider field of view, darker, and less sharp.

Ancillary 1.9mm lens, focusing on where the mouse lens is focused, brighter sharper.

I did an article in a recent issue of nuts n volts in my robotics column.

formatting link

deadline

Reply to
Blueeyedpop

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.