Thanks guys. Useful information!

So 2 possible solutions for using with "window" :
- Some kind of laser colour reading by colour track
- Some kind of tube connected to the window so that readings can be done under standard lighting conditions. This would have some sensor at the end of the tube which reads colour

And 1 more possible solution like the Probat colour sensor.
- Insert a probe into the trier/into the beans

Writing/talking to sensor companies to figure out what are the possibilities.

Does anyone have image processing/signal processing background here? Lets say we were able to get colour information. How would you take that and make sense of that data?

Cheers

why not aim a bit higher and go for mass spectometry so you can analyze the components created? (that is, once a correlation between taste and compounds captured has been established)

If we're having t rely on a surrogate parameter why not pick one that has a better chance of good correlation to flavor?