Maker Pro
Maker Pro

'green color' sensor

Q

Quack

Jan 1, 1970
0
Hi,

I need to detect the presence of 3 colors at a maximum distance of
7-10cm. (fixed distance, position, orientation etc).

The colors are very specific and cannot be altered - the light
conditions are also fixed - bright indoor lighting (think casino).

The colors fly by very fast, and they are BRIGHT RED, DULL GREEN and
BLACK.

Thats right - reading results from a roulette wheel.

I baught a "BALLUFF BKT" LASER sensor. (www.quack.cc/36970.pdf)
This detects the RED/Black perfectly but i cant make it distinguish
the green from the black :(. (it has a teach function, but it doesnt
see enough of a difference in contrast i guess).

I liked this sensor because it used a visible laser diode (aiming is
important) and it has fast reaction time and a long sensing range. the
documentation says upto 15cm - but in reality i had it upto 25.

So i am looking for a "green color" sensor - more specific than just a
'contrast' sensor. It needs a range of about 7-10 cm, and if possible,
a visible aiming mark.


I read somewhere that a red led/laser will not pickup greens well
because green absorbs most red light - and i have seen "green led"
contrast scanners around - but only for short-range :(.

does anyone knows of a green-led/laser contrast scanner with a better
than 2cm maximum range ?

Thanks :)

Alex.
 
R

Robert Baer

Jan 1, 1970
0
Quack said:
Hi,

I need to detect the presence of 3 colors at a maximum distance of
7-10cm. (fixed distance, position, orientation etc).

The colors are very specific and cannot be altered - the light
conditions are also fixed - bright indoor lighting (think casino).

The colors fly by very fast, and they are BRIGHT RED, DULL GREEN and
BLACK.

Thats right - reading results from a roulette wheel.

I baught a "BALLUFF BKT" LASER sensor. (www.quack.cc/36970.pdf)
This detects the RED/Black perfectly but i cant make it distinguish
the green from the black :(. (it has a teach function, but it doesnt
see enough of a difference in contrast i guess).

I liked this sensor because it used a visible laser diode (aiming is
important) and it has fast reaction time and a long sensing range. the
documentation says upto 15cm - but in reality i had it upto 25.

So i am looking for a "green color" sensor - more specific than just a
'contrast' sensor. It needs a range of about 7-10 cm, and if possible,
a visible aiming mark.

I read somewhere that a red led/laser will not pickup greens well
because green absorbs most red light - and i have seen "green led"
contrast scanners around - but only for short-range :(.

does anyone knows of a green-led/laser contrast scanner with a better
than 2cm maximum range ?

Thanks :)

Alex.

Let me get this straight.
You *have* excellent illumination.
Then pray tell me why the foolishment of added monochromatic light?
Use two photodiodes, one with a green filter and the other with an
orange filter.
From the relative (ratio) outputs one can determine most colors,
specifically those you wish to detect.
That solves the color detection problem, and does nothing for pattern
recognition.
Scanning light detectors, not unlike the tubes used in TV cameras
could be pressed into that service, and since color filters are built
in, one gets all the benefits.
 
Q

Quack

Jan 1, 1970
0
Let me get this straight.
You *have* excellent illumination.

Yes it will never be used in the dark - but there will be ambient
light alterations during use - and possible shadows cast this way and
that.
Then pray tell me why the foolishment of added monochromatic light?
Use two photodiodes, one with a green filter and the other with an
orange filter.
From the relative (ratio) outputs one can determine most colors,
specifically those you wish to detect.

that sounds good in a closed environment - however with the
fluctuations in ambient light and the shadows, i would think that i
would need to calculate some averages over the entire spread of colors
(regularly) to ensure accurate detection.

Also how would this work over a distance of 7-10 cm ?
Given that your comparing the outputs of the two phototransistors,
then they would need to be angled such that they both point at
near-enough to the same dot X cm away to have an accurate reading 'at
the same point' - since the object to detect is moving and you wouldnt
want one to read the pad NEXT to the target and compare with the other
which may be pointed at a different color.

In which case an aiming laser/led is nice.

There are products which already overcome all of this - i know i could
make my own sensor/assembly - but i'd rather just buy something that
exists if i can :).

That solves the color detection problem, and does nothing for pattern
recognition.

I already have a working version which uses a 'white dot' placed on
the green to detect its presence - however i want to make a version
which does not need the dot - which will find the green itself.
So as far as logic/recognition goes - thats all done and working
already.

Scanning light detectors, not unlike the tubes used in TV cameras
could be pressed into that service, and since color filters are built
in, one gets all the benefits.

I didnt get what your trying to say there at all ? "press into that
service" ?


Thanks for your help!
i may use such a technique in the end (a much cheaper solution in the
long run, this these pre-fab 'smart' sensors run up in the 'hundreds
of pounds' a piece)


when you say 'green/orange' filters - is that something placed infront
of the phototransistor or is it built-in as part of the package ?
Got any part#'s for such a 'filtered' phototransistor ?

Alex.
 
R

Robert Baer

Jan 1, 1970
0
Quack said:
Yes it will never be used in the dark - but there will be ambient
light alterations during use - and possible shadows cast this way and
that.


that sounds good in a closed environment - however with the
fluctuations in ambient light and the shadows, i would think that i
would need to calculate some averages over the entire spread of colors
(regularly) to ensure accurate detection.

Also how would this work over a distance of 7-10 cm ?
Given that your comparing the outputs of the two phototransistors,
then they would need to be angled such that they both point at
near-enough to the same dot X cm away to have an accurate reading 'at
the same point' - since the object to detect is moving and you wouldnt
want one to read the pad NEXT to the target and compare with the other
which may be pointed at a different color.

In which case an aiming laser/led is nice.

There are products which already overcome all of this - i know i could
make my own sensor/assembly - but i'd rather just buy something that
exists if i can :).


I already have a working version which uses a 'white dot' placed on
the green to detect its presence - however i want to make a version
which does not need the dot - which will find the green itself.
So as far as logic/recognition goes - thats all done and working
already.


I didnt get what your trying to say there at all ? "press into that
service" ?

Thanks for your help!
i may use such a technique in the end (a much cheaper solution in the
long run, this these pre-fab 'smart' sensors run up in the 'hundreds
of pounds' a piece)

when you say 'green/orange' filters - is that something placed infront
of the phototransistor or is it built-in as part of the package ?
Got any part#'s for such a 'filtered' phototransistor ?

Alex.

1) If there are too many shadows or large brightness range, then you can
add light with a simple fluroescent light.
2) The photodiode scheme is good over a wide illumination range,
especially when ratios are used.
However, if you are worried about it, use a third un-filtered sensor
for detecting overall illumination level.
3) One could use a vidicon ot TV camera or similar units for scanning
and color seperation in one unit.
 
Q

Quack

Jan 1, 1970
0
1) If there are too many shadows or large brightness range, then you can
add light with a simple fluroescent light.

wouldnt led's do this, or is the distance too far ? perhaps an
'ultrabright' led.

BTW - i noticed a lot of sensors have a focused, and in some cases a
'shaped' pattern at their optimum range. I guess they have a special
focus/filer on the LED - are these standard led's you can buy ?
is there a special name for this, cause my searches cant find much :(.
2) The photodiode scheme is good over a wide illumination range,
especially when ratios are used.
However, if you are worried about it, use a third un-filtered sensor
for detecting overall illumination level.

I'll have to get some photodiodes and do some experimentation, i see
what your getting at - sounds simple enough.

I did some product searches for photodiodes on farnell and RS
components websites, and they stock quite a range. Should i order a
range of different types and see which gives the best results, or are
there specific ones that would work better for green/red detection.
Any pointers in choosing some ?
3) One could use a vidicon ot TV camera or similar units for scanning
and color seperation in one unit.

Ahh i considered pointing a CCD chip camera at the rim of the wheel,
and using a SX chip (for the speed) to process the data and read a
particular x,y bit (or group of bits) to determine the color. (i guess
somewhat similar to doing something with a vidicon?).

Although i passed up on that because i didnt think i would get the
frame-rate.

The wheel spins from 1 to 3 revolutions per-second - and is divided up
into 37 segments. each segment is colored differently to the one next
to it.
This leaves only 27 milliseconds per segment at the fastest speed of
rotation.
I was worried a camera wouldnt react fast enough for that.
Or perhaps im wrong ? i havnt worked directly with CCD chip cameras
before - perhaps they can.

Thanks again for your help,

I'll start with getting some photodiodes and see how i go.

Alex.
 
R

Robert Baer

Jan 1, 1970
0
Quack said:
wouldnt led's do this, or is the distance too far ? perhaps an
'ultrabright' led.

BTW - i noticed a lot of sensors have a focused, and in some cases a
'shaped' pattern at their optimum range. I guess they have a special
focus/filer on the LED - are these standard led's you can buy ?
is there a special name for this, cause my searches cant find much :(.


I'll have to get some photodiodes and do some experimentation, i see
what your getting at - sounds simple enough.

I did some product searches for photodiodes on farnell and RS
components websites, and they stock quite a range. Should i order a
range of different types and see which gives the best results, or are
there specific ones that would work better for green/red detection.
Any pointers in choosing some ?


Ahh i considered pointing a CCD chip camera at the rim of the wheel,
and using a SX chip (for the speed) to process the data and read a
particular x,y bit (or group of bits) to determine the color. (i guess
somewhat similar to doing something with a vidicon?).

Although i passed up on that because i didnt think i would get the
frame-rate.

The wheel spins from 1 to 3 revolutions per-second - and is divided up
into 37 segments. each segment is colored differently to the one next
to it.
This leaves only 27 milliseconds per segment at the fastest speed of
rotation.
I was worried a camera wouldnt react fast enough for that.
Or perhaps im wrong ? i havnt worked directly with CCD chip cameras
before - perhaps they can.

Thanks again for your help,

I'll start with getting some photodiodes and see how i go.

Alex.

Actually the color detection "problem" can be simplified greatly, due
to the fact that the pattern is fixed (i think red/black repeatedly).
So, you can "strobe" that; simplist way might be to use the average
intensity waveform to generate a synch by limiting. That way you have
something to use for other processing needs.
Also, if you do have a light, it can be flashed (using that synch) at
the appropiate times, to "freeze" the wheel for the CCD scan; since ther
is more illumination, there may be sufficent sensitivity to "grab" the
picture (so to speak).
 
C

Charles Jean

Jan 1, 1970
0
Check out:

http:\\www.intor.com

They have photodiode detectors with an integral optical interference
filter. Your choice of wavelengths between 300 and 1100 nm. They
also have narrow bandpass. Pick one that matches the "green"
wavelength your trying to detect. They're about $20. Their
application engineers are very helpful.

Charlie
_____________________
 
E

EIDDO

Jan 1, 1970
0
I found that photo resistors are extemely sensitive to green light.
 
Top