On Fri, Dec 20, 2002 at 09:36:22PM -0500, Larry Sanders wrote:
> - then write routines that sample and process the data
> for recognition of a passing fish. It will be hard to code
> a solution before you know the data pattern.
<snip>
I am not an engineer, but I'd have to agree with Larry. You'll never
know how to code this until you see the pattern of the data in action.
It will have to be heuristic. You look at a graph of the data, code to
that, and run the test again, on a predetermined number of fish. Code
to that, until you get a reasonably accurate result. I can fully imagine
fish going down the sleuce, but I can't imagine how to count them unless
I either a) slow the whole thing down so that only one fish gets through
at a time, or b) examine and test various known data sets as I code.
I think you mentioned how "commercial" solutions for this problem were
expensive? This may be the reason why. If I had unlimited funds to do
something like this, I'd probably use a grid of lasers. Or maybe
gravitic mass detectors. Or possibly force field perturbation anomaly
encoders. Using positronic neural nets, of course. ;-}
Paul
This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 19:46:41 EDT