LabVIEW
Color Calibration across multiple cameras
It's a long time ago that I did an application for this and I did not develop the math behind it (smarter people than me from the customer did) but color calibration is a bit of magic.
But are you by any means trying to do the color calibration directly in RGB color space? That is not advisable as RGB space has some non-linear aspects as far as color representation is concerned. If I remember correctly we did some translation to CIELab, then applied the color calibration routines (which were some matrix calculations) that the customer had developed and then converted everything back to RGB for display. And I did part of the conversions by integrating C routines from some open source project. We also controlled the cameras directly to acquire raw images which then got processed by calling dcraw.
You also need to normalize the data at some point to avoid over-saturation of one or more of the color channels when you apply the color correction.
Also, all this trouble only has any value if you have a completely controlled environment in which you make both the calibration and product snapshots. Ambient light of any form that can reach the object being photographed is absolutely surely going to change your color values that the camera sees. The customer used special boxes in which the camera was mounted, that had a calibrated light source, specially treated inside walls and could be completely closed for the actual picture acquisition.
You need to account for the light source to change its spectrum both over short time due to warming up and over longer periods of time due to aging.
This means that you need to have a warm up time before doing calibration and measurements as well as need to redo the calibration in certain intervals, even if it is a LED source. Other light sources will require even more regular recalibration as they tend to age quicker.
Turns out I remembered something wrong and we did not use CIELab but rather XYZ color space for this. I can't post the code here really but basically we used the XYZ color values from an IT8 color card and scanning/photographing that card we extracted the mean of the RGB values of each color spot, This RGB3 was converted to RGB9 which is a combination of the original RGB pixels, the squares of them and the multiplication of each with one of the other. This was then matrix multiplicated with its transposed matrix and again and then matrix multiplicated with the XYZ reference values for the IT8 colors. The resulting array was again transposed and used as calibration matrix.
The RGB values from the real image were then again translated into RGB9, transposed and this was matrix multiplied with above calibration matrix and transposed again to get the XYZ values for the to measure spots in the picture. This was it, as the customer wanted to have XYZ values for the final measurements.
Basically everything is more or less turned into a linear algebra solution, which I have to admit wasn't my strongest point in mathematics. 
Without someone who understands the mathematics involved with this really well, I'm afraid it won't be possible to do this right.
And you most likely want to use RAW images from the camera so that you can convert them with dcraw or similar into color images with known white reference values and all that. Otherwise your are pretty much at the mercy of the color transformation and image correction algorithms that your camera manufacturer put into their devices.
