Two deliveries can weigh the same yet contain different amounts of water and oil. Knowing that difference before milling could help organise reception and understand yield. Near-infrared spectroscopy, or NIR, offers a route: illuminate a sample, record the light it returns and relate its spectrum to a reference analysis.
But reading a spectrum is not the same as knowing a delivery’s composition. The instrument needs calibration, the sample must be representative, and the result must hold beyond development data. Our project is at TRL 2: computational work with real public data and simulation. We have neither our own spectrometer nor a mill pilot yet.
What the light sees
Water and fats contribute to different infrared bands. Surface condition, light scattering and acquisition conditions also change the signal. A model may learn useful associations with laboratory values without identifying every component independently and infallibly.
We must also distinguish fruit oil content from oil recovered at the mill: recovery depends on processing. Dry matter is not fat either; it includes other constituents. We keep those reference quantities separate.
The maturity camera studies the exterior. NIR studies another, potentially complementary signal. None of our experiments establishes internal pigmentation without cutting the fruit.
The rest is for key holders
The background is open. We share the results, the figures and the full exam with the cooperatives, olive mills and research teams we work with.
Don't have a key? Ask us for one
