Innovative Spectral Approaches for Wheat Analysis: Classification, Vitreousness, and Protein Assessment

The global importance of wheat as a food commodity is widely recognized, with a production level of around 785 million metric tonnes, making it a staple in agriculture. Wheat quality can be categorized into two broad areas: internal and external quality. Internal quality is primarily influenced by the genotype, influencing attributes like chemical composition, gluten and protein content. Conversely, external quality depends on weather-induced damage, the presence of foreign materials and colour. However, wheat quality is a multifaceted and inclusive concept. For example, farmers may prioritize yield, while manufacturers focus on milling efficiency. Thus, wheat quality is context-dependent, influenced by its suitability for specific end uses. To satisfy the diverse quality demands and optimise profitability, it is essential to have a comprehensive understanding of wheat types, as each cultivar possesses distinct quality attributes such as protein content, hardness, and gluten levels, which in turn influence market prices. This necessitates advanced technology to rapidly and accurately classify wheat types and assess their quality, minimizing losses in time, labour, and resources. This thesis addresses this need by exploring the determination of wheat varieties and their quality parameters, including vitreousness, protein, and wet (WG) and dry gluten contents (DG), using advanced spectral technologies such as Visible-Near Infrared (Vis-NIR) hyperspectral imaging (HSI), Short-Wave Infrared (SWIR) HSI and Fourier Transform Near-Infrared (FT-NIR) spectroscopy. The findings reveal that thirty-seven wheat varieties can be classified using HSI with a validation accuracy of 94.20% and a test accuracy of 94.93% based on single kernel data and an accuracy of 100% for both validation and test sets using bulk data. Furthermore, wheat grains were classified using HSI in the Vis range with five selected wavelengths, achieving an accuracy of 92.3% for vitreous grains and 78.8% for non-vitreous grains. Additionally, the study demonstrates that the protein, WG and DG content of wheat grains can be predicted using Vis-HSI with four selected wavelengths, achieving R2P values of 0.97 for both WG and DG and 0.98 for protein. The results also indicate that WG, DG and protein levels can be predicted using a portable HSI with LED lighting at 4400 K, employing five selected wavelengths, with R2P values of 0.91, 0.84 and 0.88, respectively.