RGB Sensor

[Klitzie]

I'm building a RGB based color sensor using LDR but my problem is the blue led has most luminance compare to the red and green. So my solution is to connect a potentiometer before the led anode so that i can control the current to limit the blue led brightness. Is there a way on how to fully calibrate the brightness of the red,green and blue led so that there brightness is even to each other...

 

[justDIY]

don't worry about the brightness of the source light, just calibrate your sensor in the software using some M&M candies

 

[Klitzie]

How do i do that? can you give me a hint!

 

[justDIY]

get your m&ms, put one in front of the sensor, taking readings as you alternate red, blue, green, record the readings for that color, then tell the software those values equal that color. repeat with the other colors. then you'll have a table of known colors and their corresponding values. from there you can evaluate other colors and use fuzzy logic or whatever to determine what color the unknown color is closest to.

 

[padu]

I like robotics, photography and electronics, so I was planning on making a device that would involve all of my passions.

I'd like to build a color analyzer device, very simple to use: put it on the surface you want to measure the color and it tells you what R,G and B components are reflected by that surface.

In my understanding, I need to have a light source (or 3, one for each color component) and their wavelenght needs to be pure (well, wrong language I guess, I guess I mean it needs to contain only one wavelenght), and a light sensor, that would measure the amount of light reflected by it, then I use a microcontroller to make the calculations and drive an LCD and/or feed a host PC with those values.


I guess my question is: how can this be built to be very accurate? For example, if a given color in HSB space, I want the error in Hue, Saturation and Brightness to be inferior to +-0.1 deg.

I know of some commercial products that do exactly this, but they are so expensive. Is it feasible to build one for hobby?

 

[justDIY]

accuracy is all in the calibration.

I wouldn't mess around with trying to cobble a sensor together with a simple ldr. get a pre-fab rgb photo diode array specifically designed for color sensing applications - they are actually quite cheap.

next you'll need to build yourself some sort of full spectrum light source, assuming you want to measure reflected color versus emitted color? canon color led scanners use a large rgb array of tiny smt leds to illuminate the media they are scanning, and it seems to work well, however, a cool white fluorescent bulb would be far easier.

now all you do is feed your sensor a calibration panel - they sell them for calibrating color copiers and such... each color is well defined in a industry standard color space, so you just program your sensor to recognize these colors. once you've got it calibrated, then it is a matter of writing software to interpolate values for colors that fall between those the sensor was calibrated for.

 

[dknguyen]

From PICAXE:
https://www.electro-tech-online.com/custompdfs/2008/02/axe045.pdf

THe colour sensing IC mentioned is probably more important than the module though (although the module has a neat lens).