/philipp_marion_SINTEF_senseinside.png)
Background
SenseInside is a novel subsurface NIR (near-infrared) sensing technology developed by SINTEF and Nofima, now in an active commercialisation track. It has already been tested on a number of food quality applications. We currently are building a second SenseInside instrument and are ready to conduct robustness trials related to scalability. For SenseInside to become a scalable, manufacturable platform, the underlying measurement models must be robust to process variations — including differences in optics, spectrometer drift, and environmental conditions such as temperature or humidity. Predicted values made on one food product (e.g. strawberries or tomatoes) must work reliably across a fleet of instruments, despite inevitable instrument-to-instrument variability. Achieving this robustness is a major unsolved challenge in applied spectroscopy: predictions are sensitive to very small spectral variations and low SNR. This master’s thesis can tackle this challenge head-on, combining multimodal sensing (NIR + camera + environmental sensors) with calibration transfer, drift detection, and correction methods. The work will directly support SenseInside’s path toward industrial impact within SFI DigiFoods and beyond.
Objectives
The student can, for example:
• Explore multimodal sensor configurations (NIR + camera + environmental sensors) to study robustness under different conditions.
• Benchmark identical SenseInside units under both controlled and real-world conditions to understand variability.
• Enable trials for developing calibration transfer methodologies across multiple SenseInsides.
• Investigate drift detection and correction strategies (e.g. wavelength referencing).
• Benchmark identical SenseInside units under both controlled and real-world conditions to understand variability.
Methods & Timeline
Semester 1 (Spring 2026): Literature review on calibration transfer and drift detection; hands-on familiarisation with SenseInside platform.
Semester 2 (Autumn 2026): Experimental work with multiple SenseInside units; collect multimodal datasets under varied conditions (temperature, humidity, vibration). Test calibration transfer and drift correction methods.
Semester 3 (Spring 2027): Develop proof-of-concept algorithms for drift correction and calibration transfer; validate on real food product datasets (e.g. strawberries, tomatoes). Prepare thesis.
Why this project?
The student will gain cutting-edge experience at the boundary between sensors and the real world: hardware, multimodal data fusion, and AI for industrial sensing. The work is embedded in active industrial R&D projects (SenseInside, NIRCore, SFI DigiFoods), giving access to unique data, hardware, and networks. The results will have direct impact on commercialisation, shaping how robust sensing platforms are deployed in food processing and beyond.