Newsgroups: sci.image.processing
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!uhog.mit.edu!bloom-beacon.mit.edu!gatech!swrinde!sdd.hp.com!hp-pcd!hpcvsnz!briand
From: briand@cv.hp.com (Brian Dixon)
Subject: HP SINGAPORE LED INSPECTION!
Sender: news@hpcvsnz.cv.hp.com (News )
Message-ID: <D82v1E.401@hpcvsnz.cv.hp.com>
Date: Thu, 4 May 1995 23:39:14 GMT
Nntp-Posting-Host: hpcvsgen.cv.hp.com
Organization: Hewlett-Packard
X-Newsreader: TIN [version 1.1 PL9.4]
Lines: 38

Unfortunately, I lost the email address of the Hewlett Packard engineer
that contacted me about inspection of LED arrays for color and intensity.
One of the topics discussed was calibration of the inspection system...and
I suggested the possible use of known sources of monochromatic light such
as a laser.  I have another idea (which I had forgotten about when
responding the first time): There are a number of spectral analyzers 
available in the world which can measure light wavelength quite accurately.
In addition, many of these systems can measure intensity variation quite
accurately.  Some are designed to measure spectral response of materials
such as reflected or absorbed light, others are designed for laser or
fiber optic system calibration and verification.  My suggestion is this,
why not use one of these systems together with a set of LEDs in a 
calibration procedure which would consist of A) measuring the LED output
with one of these analysis tools, then B) measuring the LED output
with the vision inspection system, and finally, C) calibrating the 
vision system accordingly.  The best procedure would be to find a set of
LEDs that range in color and intensity outside (and inside) the specification
limits for the LED product.  You could then use regression to fit a
calibration function to the measured values over a range of color and
brightness (intensity).  Also, as I mentioned before, conversion from the
RGB-space representation to the HSI-space representation would first be
necessary.  The advantage to this approach would also be that you'd be
able to calibrate around a set of wavelenths and intensities that you
actually expect to see in production.  The use of alternative light sources
may not allow calibration this close to what you'll actually be measuring.

If you (the HP engineer *specifically*) would like more information
such as contact information for a variety of vendors, please respond
to me with email (again!).  My apologies to those other non-HP
engineers who would also like such information...my time is pretty
pressed here at work and I only get brownie points for assisting other
HP people...

Brian

--
Brian Dixon, Machine Vision Engineer, Hewlett Packard (Corvallis, Oregon)
503-715-3143 (wk), briand@cv.hp.com (email). "Opinions & attitudes are mine!"
