From fundamentals to performance: robust near-infrared spectroscopic systems for real-world applications

Summary

Near-infrared (NIR) spectroscopy is widely used for rapid and non-destructive quality assessment across a broad range of fields, including food production, pharmaceuticals, and materials science. However, robust measurements under realistic industrial and field conditions remain challenging, particularly when compact instruments are used for non-contact measurements of intact, heterogeneous samples. For such samples, the surface often differs from the interior, making subsurface sampling necessary to obtain a measure of the sample as a whole. This can be achieved using interaction measurements, where the illuminated and detected regions are spatially separated. At the same time, constraints related to instrument size, speed, and signal quality must be balanced, while the instrument must be sufficiently robust for operation under demanding industrial and field conditions. This thesis investigates fundamental robustness challenges in NIR interaction spectroscopy and how robust non-contact subsurface measurements can be realised using compact and application-feasible instrumentation in real-world environments. Rooted in first principles and a system-level perspective, the work focuses on understanding how instrumentation, sampling geometry, sample properties, and environmental conditions jointly govern signal formation, and how this, together with multivariate data analysis, determines overall measurement performance. This is explored through four papers addressing different aspects of robust real-world NIR measurements. In the first paper, a physics-based framework for the development of robust spectroscopic measurement systems is presented, emphasising understanding of the complete measurement chain and the requirements imposed by practical applications. In the second paper, the optical properties of dried salt-cured cod muscle are characterised during drying to understand how structural and compositional changes influence absorption and scattering behaviour. In the third paper, these optical properties are used as the basis for Monte Carlo simulations of NIR interaction measurements. The simulations provide insight into signal formation, photon transport, and how representative subsurface sampling can be achieved in layered and heterogeneous samples. In the fourth paper, the feasibility of on-the-plant assessment of strawberry sweetness using a compact non-contact NIR instrument is demonstrated under realistic daylight conditions, with particular focus on minimising the impact of environmental disturbances on weak interactance signals. The main contribution of the thesis is increased understanding of the physical mechanisms governing robust NIR interaction measurements under demanding real-world constraints. The work demonstrates how interactions between sample properties, sampling geometry, instrumentation, and environmental conditions influence measurement performance, and how understanding these interactions can help push challenging applications towards the limits imposed by physics. Although the work is centred on food production and agricultural applications, the underlying robustness challenges and proposed approaches are relevant to a broader range of applied spectroscopic sensing problems. The presented work supports the development of robust compact NIR measurement systems for demanding applications where implementation remains challenging.

Read publication here:

NVA : hdl.handle.net/11250/5579040