Portable Near-Infrared Spectroscopy and Support Vector Regression for Fast Quality Evaluation of Vanilla (Vanilla planifolia)

Vanilla (Vanilla planifolia) is a high-value agricultural product, with its quality influenced by essential factors such as moisture and vanillin content. Conventional techniques for evaluating these characteristics are inefficient, require sample destruction, and are impractical for swift assessments. This research explores the feasibility of using portable Near-Infrared (NIR) spectroscopy combined with Support Vector Regression (SVR) to enable quick and noninvasive property prediction. Spectral information was obtained from vanilla samples using two portable NIR instruments, SCiO (740–1070 nm) and Neospectra (1350 2550 nm). Preprocessing techniques such as normalization, SNV, MSC, first derivative, first derivative-SNV, and first derivative-MSC were applied. For moisture content prediction, SCiO achieved an R² of 0.768, an RMSE of 4.720%, an RPD of 2.075 and an RER 10.197 using Min-Max normalization, while Neospectra yielded an R² of 0.758, an RMSE of 5.161%, an RPD of 2.033 and an RER 9.325 with MSC preprocessing. In contrast, predicting vanillin concentration proved more challenging, with SCiO achieving moderate accuracy with an R² 0.406, an RMSE 0.379%, an RPD 1.297, an RER 5.039, and Neospectra demonstrating limited performance with an R² 0.172, an RMSE 0.576%, an RPD 1.098 and an RER 3.315. These findings highlight the potential of portable NIR spectroscopy as a practical tool for assessing vanilla quality, particularly for moisture content, in industrial and field applications. Keywords: Moisture content, Portable NIR spectroscopy, Support vector regression, Vanilla planifolia, Vanillin content.