Build your own color detector.

In my university lab last Summer I built a colour detector and I thought I'd let you all in on how. If you have a bit of programming knowhow and have the right electronic components (which are really easy to pick up either from the internet or from an electronics store) then this blog should familiarise you with the core principles involved when building a colour detector.

The image to the left is the finished product. 3 tinted LDRs the PIC16F819 micro controller make up the colour detector (as well as an optical to digital converter if you want a PC readout).

Understanding the nature of the colours present in light is a vital part of this project. Light has an additive nature, unlike pigments, which have a subtractive nature. This means that increasing proportions of red, green, and blue light result in

the formation of white light, and various combinations of those colours can form every colour in the visible spectrum.

Due to their ability to produce such a vast array of hues they also form the basis for colour formation in standard computer monitors, televisions and other liquid crystal displays. Technically the choice of red, green and blue are not exact, they are chosen due in large part to the sensitivities of the cone cells in the rear of the retina. Differences in the hues of the red, green and blue constituents chosen may produce different absolute colour spaces, for example sRGB and Apple RGB are two absolute colour spaces used in computer monitors. In computing terms the R, G, B value of a colour represents its red, green and blue components proportionally, normally each on a scale of 0-255 (This requires three 8-bit values). For various other purposes in computer graphics a further two bits per unit may be added, for example RGBA includes information about the alpha channel (transparency) of a pixel, and contains three 10-bit units of colour per pixel. In the case of visible light however only the RGB values, and their associated intensities need to be taken into account.

Since the LDRs that will be used in the colour detector have little or no wavelength dependence it is necessary to use colour filters to determine which wavelengths of light are hitting which detector and in what proportions. Colour filters work by transmitting certain wavelengths of light and reflecting or destroying through interference, the other wavelengths. Manufacturers often produce filters that allow not only the target colours wavelengths through. The effect of this is is often to brighten up the filters, by letting through more light.

A problem with this is that it is practically impossible to determine the exact chromaticities of the filter. A solution to this involves measuring intensities of pure red, green and blue light after they pass through the filter when compared to the intensities before transmission. By creating a transformation matrix from one set of colours to the other and inverting that matrix, a transform will be created to convert measured intensities (post filter) to the original colour of light (pre filter).

Since an optical interface is being used which can only transfer data digitally there is no way of transferring the 3 analogue voltage values directly to the computer. The PIC is used to sample the 3 values, store them, and then convert them into a form which is analysable by the PC. The way chosen to do this was to take a voltage value (in 8-bit so between 0 and 255), and a time delay proportional to this which turned the output port either on or off after each delay. The code snippet below shows how this was achieved although full code for all programs used in the project can be found in the appendices.

//liner delay subroutine (0xFF times input variable);

a equ 0x45

b equ 0x46

delay movwf a

movlw 0xFF

movwf b

ca movwf b

cb decfszb

goto cb

decfsz a

goto ca

return

The delay loop assumes that the voltage value has previously been stored in a. When the loop is called the embedded loop repeats 256a times, creating a time delay proportional to a. By turning the output port on and off, before and after the loop is called, a square wave output of voltage can be created with a period proportional to the input voltage.

movlw b'00000001'

movwf PORTB

// then call the delay loop

movlw b'00000000'

movwf PORTB

This can be repeated 3 times for each of the measured values, resulting in three flashes. The timed pulses will then be sent optically to the computer.

For documentation on how to program your PIC or other microchip I recommend visiting the manufacturers website at http://www.microchip.com/stellent/idcplg?IdcService=SS_GET_PAGE&nodeId=64

So the image to the right is pretty much a summary of the whole system from

LDR to PIC to LED. After this stage the 3 flashes and their corresponding delays (how long they are on/off for) is sent to the computer via an optical to digital converter.

The final step was to write some software in C (as well as a codec for the optical to digital converter), to output RGB values and display something nice on the screen, in this case a blue LED is being shone onto the LDRs.

The image below shows a schematic of the colour detector.

If you have any more questions about the project, let me know,

Good luck!