Abstract
Several thousand grapevine varieties exist, with even more naming identifiers. Adequate specialized labor is not available for proper classification or identification of grapevines, making the value of commercial vines uncertain. Traditional methods, such as genetic analysis or ampelometry, are time-consuming, expensive, and often require expert skills that are even rarer. New vision-based systems benefit from advanced and innovative technology and can be used by nonexperts in ampelometry. To this end, deep learning (DL) and machine learning (ML) approaches have been successfully applied for classification purposes. This work extends the state of the art by applying digital ampelometry techniques to larger grapevine varieties. We benchmarked MobileNet v2, ResNet-34, and VGG-11-BN DL classifiers to assess their ability for digital ampelography. In our experiment, all the models could identify the vines' varieties through the leaf with a weighted F1 score higher than 92%.

Abstract
Nutrient quantification in hydroponic systems is essential. Reagent-less spectral quantification of nitrogen, phosphate and potassium faces challenges in accessing information-rich spectral signals and unscrambling interference from each constituent. Herein, we introduce information equivalence between spectra and sample composition, enabling extraction of consistent covariance to isolate nutrient-specific spectral information (N, P or K) in Hoagland nutrient solutions using orthogonal covariance modes. Chemometrics methods quantify nitrogen and potassium, but not phosphate. Orthogonal covariance modes, however, enable quantification of all three nutrients: nitrogen (N) with R = 0.9926 and standard error of 17.22 ppm, phosphate (P) with R = 0.9196 and standard error of 63.62 ppm, and potassium (K) with R = 0.9975 and standard error of 9.51 ppm. Including pH information significantly improves phosphate quantification (R = 0.9638, standard error: 43.16 ppm). Results demonstrate a direct relationship between spectra and Hoagland nutrient solution information, preserving NPK orthogonality and supporting orthogonal covariance modes. These modes enhance detection sensitivity by maximizing information of the constituent being quantified, while minimizing interferences from others. Orthogonal covariance modes predicted nitrogen (R = 0.9474, standard error: 29.95 ppm) accurately. Phosphate and potassium showed strong interference from contaminants, but most extrapolation samples were correctly diagnosed above the reference interval (83.26%). Despite potassium features outside the knowledge base, a significant correlation was obtained (R = 0.6751). Orthogonal covariance modes use unique N, P or K information for quantification, not spurious correlations due to fertilizer composition. This approach minimizes interferences during extrapolation to complex samples, a crucial step towards resilient nutrient management in hydroponics using spectroscopy.