Society of Robots - Robot Forum
Software => Software => Topic started by: silo_xtreme on May 27, 2008, 12:57:51 PM
-
Hi Everyone,
I'm considering building my new robot and eventually upgrade it with vision processing. Is Linux the way to go or can I stick with my AVR? What is the minimum RAM and Processor Speeds to do edge / blob detection?
I'm just wondering what your experiences have been....
Mike
-
Depends on what image resolution you want to support. We're using a 500MHz Blackfin processor with 32Mbytes of SDRAM, and that seems to be a pretty good combination for real-time image processing on 320x256 to 640x512 resolution.
-
Excellent! What kind of image processing do you do on the black fin?
-
There's a large collection of assembly-language optimized image processing functions for Blackfin here -
http://www.analog.com/processors/blackfin/technicalLibrary/manuals/codeExamples.html
We've written some simple blob functions for our standard firmware ( http://www.surveyor.com/SRV_protocol.html ), but this only scratches the surface of what is possible.
-
i have some notes on my efforts in this direction:
http://mrdunk.googlepages.com/processor (http://mrdunk.googlepages.com/processor) and http://mrdunk.googlepages.com/sensors (http://mrdunk.googlepages.com/sensors).
although i have not documented my efforts on that page
i've recently been working with the Hammer board from http://www.tincantools.com/ (http://www.tincantools.com/)
which looks like it's a good alternative.
in my experience when trying to do image processing on limited embedded linux processors memory is more important than processor speed.
dunk.