Quality assessment and grading of sweet potato using VNIR hyperspectral imaging

Hyperspectral imaging has emerged as a promising green analytical tool for grading and accurately quantifying quality attributes in agricultural produces. Currently, hyperspectral imaging is the only analytical tool that answers the commonly asked analytical questions: what, how much, and where in the samples. The main goal of this study was to explore the potential of a visible near-infrared (VNIR) hyperspectral imaging system (400-1000 nm) for grading and predicting internal quality parameters such as the dry matter of sweet potatoes. Samples were collected from different varieties such as Bayou Belle, Murasaki, and Orleans for image acquisition and dry matter measurements. Their spectral data were extracted and analyzed using principal component analysis (PCA), uniform manifold approximation and projection (UMAP), partial least squares discriminant analysis (PLS-DA), and partial least-squares regression (PLSR). Some dominant spectral wavelengths were selected to design a low-cost multispectral imaging system for real-time implementation. The results indicated that the VNIR hyperspectral imaging technique has the potential for fast and noninvasive grading and predicting the internal quality attributes of sweet potatoes.