Abstract
Advancing spectroscopic sensing technologies for real world applications, such as robot mounted, in-field quality assessment of complex food products, requires improving instrument robustness and enabling measurements under challenging conditions. Maintaining sufficient spectrometer performance within practical constraints demands balancing sensor size and cost, tolerance to environmental influences, and sample complexity. A key to success lies in a holistic understanding of the entire measurement solution, from instrumentation to sample properties and data analysis.
SINTEF and Nofima have extensive experience in pushing sensor performance through the development of novel near-infrared (NIR) interactance-based spectroscopic instruments for non-contact, non-invasive, sub-surface measurements of complex food products. Interactance measurements penetrate deeper into the product than reflectance, providing more representative sample information and enabling more robust calibration models. This advantage, however, comes at the expense of signal strength, placing higher demands on instrument performance.
Here, we demonstrate that a first-principles, physics-based approach enables successful in-field sensing under highly demanding conditions. Using a portable instrument designed for robotic operation, this approach realised accurate prediction of total soluble solids (sugar content) in on-the-plant strawberries under daylight conditions, as illustrated in Figure 1. We present our methodology for quantifying the effects of challenging environments and non-optimal instrument properties, and for understanding how the physical and chemical characteristics of complex, highly scattering food products impact measurement quality. Finally, we describe how these insights inform instrument design and strengthen multivariate calibration models, highlighting the importance of physics-based design choices for achieving robust spectroscopic sensing in real-world environments.