From: Chuck Simmons
Organization: You jest.
X-Mailer: Mozilla 4.61 [en] (X11; U; Linux 2.0.33 i586)
Subject: Re: Detecting colored objects
References: <3DBEF3A3.56D6A15@SPAMstarpower.netTHANKYOU> <3DBF71A1.F2DD68F@SPAMstarpower.netTHANKYOU>
Date: Wed, 30 Oct 2002 14:43:27 GMT
NNTP-Posting-Date: Wed, 30 Oct 2002 06:43:27 PST
> Harry Dellamano wrote:
> > Jim, thanks for replying.
> > The objects are not really eggs but similar objects painted solid colors
> > for easy visual identification. This must be done electronically on the fly,
> > lets say moving 0.5 feet / second.
> > How do I use the spectrophotometer or colorimeter to clarify the problem?
> Well, I am in agreement with others who say that you need colored lights and/or
> detectors with colored filters. What you would do with the instrumentation is to
> clarify which filters are best, and how easily the different colors can be
> discriminated. On the one hand, if your 11 "colors" were 11 shades of gray,
> electronics could surely make these discriminations, but you might want such
> features as a stabilized light source to make the gray measurement more
> Stark color variations would be something like red, green, blue, yellow,
> cyan, magenta, black, white, pink, orange, yellow-green. With these
> wide-ranging color differences, you could use filters that select for the
> human primaries of red (610 nm), green (540 nm), and blue (450 nm).
> The filters need not be extreme narrow-band, but could be centered
> on these wavelengths more or less. Or the red filter could pass all
> wavelengths above a cutoff and the blue filter could pass all
> wavelengths below a cutoff, so they need not be expensive filters.
> The idea of the instrumentation would be to reduce trial-and-error
> in choosing the filters and lights, so that you get well-separated
> signals for the 11 samples. Depending on the colored objects,
> it might be easiest to use more than 3 detectors.
> You have an obvious issue of amplitude variability if the objects are
> rolling past the detector without stopping. The more that the colors
> are truly different, the more that you want to compare red versus green,
> etc, and try to ignore absolute amplitude.
> One way to test the reflectance in a narrow wavelength band might be
> to use an LED in that band. LEDs are not lasers but emit in fairly
> narrow bands. There are lots of LEDs now at different wavelengths.
Narrow band light sources and filters are completely inappropriate to
color recognition. The special dichroic filters designed for the purpose
are broadband with pretty flat in band response. Three are needed for
the visible region. It is useful to look at the filter spectral
characteristics of filters designed for the purpose. Kodak Wratten
gelatin filters designed for color separation are a low cost alternative
to dichroic filters.
... The times have been,
That, when the brains were out,
the man would die. ... Macbeth
Chuck Simmons firstname.lastname@example.org