PDA

View Full Version : any idea what chip they use for MOTION TRACKING??



kenpo
- 29th January 2006, 23:48
I've become really interested in doing motion tracking for robots. CMUcam is cool, but I like trying to make things. it's cheaper and more rewarding I think.

I saw this http://www.makingthings.com/products/documentation/teleo_video_user_guide/
on the net, it's a motion trackign circuit based off a pic. just plug in 1 or 2 cameras and it makes a 16x12 array to track changes! pretty freaking cool becuase unlike CMUcam, you could plug any camera in you want, eg infrared, nightcam, underwatercam, tv signal...

but I can't see what the two chips are... one is a pic, but there's something between the video connector and the pic.

any ideas??
I'd love to build a device for motion tracking.

forgie
- 30th January 2006, 09:24
I can't see what the two chips are... one is a pic, but there's something between the video connector and the pic.

any ideas??
I'd love to build a device for motion tracking.

It would probably be a high-speed ADC (analog-to-digital converter). Check out Maxim, they have a lot of good ADCs and do free samples! :)

kenpo
- 30th January 2006, 18:29
you think? I think the chip near the data connector looks like maxim, but the blurry logo on the other 2 chips doesn't look to be maxim, I think it looks more like a texas instruments logo....

forgie
- 31st January 2006, 02:02
I wasn't saying it is a Maxim chip, I'm just saying that they have good ADCs in my experience!

Maybe there's a video pre-amp or something like that as well. I don't know anything about video specifically, but the system would probably consist of

Input -> Preamp (possibly) -> ADC -> uController

Maybe there'd be some other chip in there, but if the uC was fast enough you wouldn't need one. Hmmm maybe an Atmel would be more appropriate for the rigours of high-speed video processing...

harrisondp
- 31st January 2006, 06:35
They probably use a LM1881 ( http://www.national.com/pf/LM/LM1881.html ) composite sync seperator chip to track which video line they are on. Then they probably take timed ADC samples on a block of lines and average them. The values are probably then stored in a 2d array which is then used to be compared with the next frames.

This is not necessarily a difficult process or project other then the fact that you probably have to set a PIC ADC to the fastest conversion time, which might be fast enough for what you need.

You will most likely have to use ASM or maybe a very very very efficent c compiler. You will also have to use some really really tight loops to track video lines and read adc values.

If it was me, I would have prolly used some sort of faster processor like a SX. This would allow better timing for video processing.

I was actually going to try this once, but I never got around to placing an order for a LM1881 (I'm just a poor high school student).

Harrison

forgie
- 1st February 2006, 03:13
Harrison looks like he knows what he's talking about. Follow his lead.... I've never worked with video, so my generic data acquisition block diagram may or may not have been helpful to you.

Good luck with it.