Colour sensor idea, advice needed please.

[Dan Messenger]

Hi,

I have come up with an idea for a colour sensor, however I have ran into a
slight problem, and was wondering if you could help.

The theory is this:
By independantly shining red, green and blue light at a near object, and
independantly measuring the amount of light reflected (by each led in turn),
it *should* be possible to detect the colour of the object.

Now, my first idea was to do this: Have 3 ultrabright LEDs (red, green and
blue), arranged in a triangle around a cds cell. Shield the cds cell from
direct light using black card / insulation tape or whatever.

Then I thought about it some more, and came upon this problem: If the object
is not central, then the detected colour will vary. Consider an object whose
colour has red and blue elements. When the object is nearer the red led, the
red signal shall be stronger - and the blue weaker. When it is nearer the
blue led, the blue signal shall be stronger - and the red weaker.

The solution I came up with was to have the light emitting from a single
source, e.g. a tri-colour led. However I cant find a supplier ! On Maplins,
Farnells and RS I have found a "tri-colour led", but these are 3 lead
packages, which do Red + Green and any colour inbetween (full red and green
being yellow, the third colour).

And so to my questions:
Is this a feasable idea? Or am I mad?
What about the problem I speculated? Is this actually going to be a problem?
Where can i get a single source led that will do red, green and blue ?
Could I use multiple leds to 'balance out' the light ? (e.g. make sure there
are equal number of leds on each side of the cds cell)
Or is there an 'off-the-shelf' colour sensor that I can use? Will it be
cheaper and / or easier to implement than my idea ?

Thanks in advance for any help / advice you can give.
-D@n

 

[Jerry G.]

You are not far off at all. Infact colour measuring equipment do something
like what you want to do. They use white light, and have a photosensor
system using the proper calibrated filters to pass the light. The
luminosity must be first calibrated for a reference to take an accurate
measurement. If you want to do this to a good degree of accuracy, it will be
complicated.

This can get very involved. If you go to a good university library, or do
some searching on the net, you should be able to come up with some good
information about this topic. In a university library you would be able to
get the involved text books on the subject.

Accurate colour measuring equipment can easily run in to the many thousands
of dollars, just to give you an idea of what is involved.

You can look at a site called Photovolt
http://www.photovolt.com/pages/ss577.html This instrument is a precision
laboratory standard unit, that is used in may industries of various types.
For example, if you were making a product where you wanted the colour to be
consistent this meter will help to do that. These are used in many
industries such as food manufacturing, printing, paint manufacturing, and
any product where colour consistency is important. The human eye never sees
colour exactly the same all the time. The perception of colour changes with
the ambient light, and how we feel at the time. These colour meters can see
colour shifts much smaller than any human or animal can see.

A colour meter for liquids from Photovolt
http://www.photovolt.com/pages/byklcs.html





--

Greetings,

Jerry Greenberg GLG Technologies GLG
=========================================
WebPage http://www.zoom-one.com
Electronics http://www.zoom-one.com/electron.htm
=========================================


Hi,

I have come up with an idea for a colour sensor, however I have ran into a
slight problem, and was wondering if you could help.

The theory is this:
By independantly shining red, green and blue light at a near object, and
independantly measuring the amount of light reflected (by each led in turn),
it *should* be possible to detect the colour of the object.

Now, my first idea was to do this: Have 3 ultrabright LEDs (red, green and
blue), arranged in a triangle around a cds cell. Shield the cds cell from
direct light using black card / insulation tape or whatever.

Then I thought about it some more, and came upon this problem: If the object
is not central, then the detected colour will vary. Consider an object whose
colour has red and blue elements. When the object is nearer the red led, the
red signal shall be stronger - and the blue weaker. When it is nearer the
blue led, the blue signal shall be stronger - and the red weaker.

The solution I came up with was to have the light emitting from a single
source, e.g. a tri-colour led. However I cant find a supplier ! On Maplins,
Farnells and RS I have found a "tri-colour led", but these are 3 lead
packages, which do Red + Green and any colour inbetween (full red and green
being yellow, the third colour).

And so to my questions:
Is this a feasable idea? Or am I mad?
What about the problem I speculated? Is this actually going to be a problem?
Where can i get a single source led that will do red, green and blue ?
Could I use multiple leds to 'balance out' the light ? (e.g. make sure there
are equal number of leds on each side of the cds cell)
Or is there an 'off-the-shelf' colour sensor that I can use? Will it be
cheaper and / or easier to implement than my idea ?

Thanks in advance for any help / advice you can give.
-D@n

 

[Ian Stirling]

Hi,

I have come up with an idea for a colour sensor, however I have ran into a
slight problem, and was wondering if you could help.

The theory is this:
By independantly shining red, green and blue light at a near object, and
independantly measuring the amount of light reflected (by each led in turn),
it *should* be possible to detect the colour of the object.

This works, but there are a number of problems.
Your apparent level of knowledge seems to indicate this would be a major
task, as overcoming the problems is not quite trivial.

Is this a feasable idea? Or am I mad?

Yes, possibly.

Or is there an 'off-the-shelf' colour sensor that I can use? Will it be
cheaper and / or easier to implement than my idea ?

Depends what you want it for.
Colour sensors are available.
More details on what you'r trying to do would be handy.

 

[Sir Charles W. Shults III]

Hi,

I have come up with an idea for a colour sensor, however I have ran into a
slight problem, and was wondering if you could help.

The theory is this:
By independantly shining red, green and blue light at a near object, and
independantly measuring the amount of light reflected (by each led in turn),
it *should* be possible to detect the colour of the object.

Now, my first idea was to do this: Have 3 ultrabright LEDs (red, green and
blue), arranged in a triangle around a cds cell. Shield the cds cell from
direct light using black card / insulation tape or whatever.

<snip>

Your idea is good, but you must take into account a few things.
First, a CdS cell is a poor choice because it has terrible response to
some colors. What you want is something with a better spectral response,
like a silicon photodiode. Second, you need to take ambient light readings
to try to compensate for them. While shielding the object is a great idea,
it does not always work as you would like. Three sensors with three
different filters is a pretty good way to go, but as you note, the positions
of the sensors can greatly vary the signal you receive.
Also, proper filters are difficult to get without incurring a lot of
expense. I can give you a suggestion for the placement problem, though.
Use fiber optics.
By making a bundle of fine light fibers and using it as your "eye", you
can eliminate the parallax or target position problem. This is because you
take the fibers and route them to the different sensors (or light sources)
as you would like, thus breaking your target into many tiny sections of the
image that are evenly distributed over your target.
Try this and when you run into your next roadblock, post again.

 

[Dan Messenger]

Hi,

Firstly thanks for your responses, 2ndly a bit more about what I want to do:

Basically I haven't decided what I want the sensor to do yet. We have to do
a project at uni which involves 2 stages. Stage 1 is to build a robot from
design and parts given, the bot base is about 9cm by 16cm, and has
ultrasound sensors on it. We have to program this robot to be able to avoid
/ follow objects (using a PIC). All the PCB etc for this is provided. The
2nd stage is to "add additional features" to the robot. This is where my
colour sensor idea comes in.

I haven't decided what I want he colour sensor to do yet, I am thinking
something along the line of getting the robot to move around and maybe
locate a particular coloured block, or perhaps stay in a region bounded by
coloured floor tiles, or a combination.

The precision of the sensor doesn't need to be too great, although obviously
it will need to be able to distinguish 'base colours', and obviously
calibration will be needed.

Currently I'm just throwing ideas around my head, and was wondering how
feasable an idea it was. Or if you have another idea of how I could adapt
the bot, feel free to suggest it (we have 30hours assigned lab time, and
whatever else we want to put in over the next 10 weeks).

-Dan

 

[Mark Fergerson]

Hi,

Firstly thanks for your responses, 2ndly a bit more about what I want to do:

Basically I haven't decided what I want the sensor to do yet. We have to do
a project at uni which involves 2 stages. Stage 1 is to build a robot from
design and parts given, the bot base is about 9cm by 16cm, and has
ultrasound sensors on it. We have to program this robot to be able to avoid
/ follow objects (using a PIC). All the PCB etc for this is provided. The
2nd stage is to "add additional features" to the robot. This is where my
colour sensor idea comes in.

I haven't decided what I want he colour sensor to do yet, I am thinking
something along the line of getting the robot to move around and maybe
locate a particular coloured block, or perhaps stay in a region bounded by
coloured floor tiles, or a combination.

The precision of the sensor doesn't need to be too great, although obviously
it will need to be able to distinguish 'base colours', and obviously
calibration will be needed.

Currently I'm just throwing ideas around my head, and was wondering how
feasable an idea it was. Or if you have another idea of how I could adapt
the bot, feel free to suggest it (we have 30hours assigned lab time, and
whatever else we want to put in over the next 10 weeks).

How about having it follow different colored stripes of
tape on the floor/walls/ceiling? It would have to find
stripes, then determine its orientation WRT them
(directional data could be extracted from sonar data), which
breaks the "target acquisition" problem into chunks.

For the add-on to the usual sonar, use two optical
systems in tandem. One looks all around to rough-cut targets
(is there a red/green/blue object visible and in which
direction?) correlated as above with sonar data and the
second targets a selected color. Charles Shults' fiberoptic
"mosaic" trick would be perfect for the first task, then a
limited version with a smaller field of view (or switchable
optics in the main system) does the fine work.

I don't see why you want an _active_ color vision system.
Why not use filtered broadband sensors using ambient light?
That way you could sense and compensate for changes in
available spectrum (say going from daylight to fluorescents,
even candlelight). If you're gonna show off, do it right! ;>)

Mark L. Fergerson