The demand for high quality and safety in food production protocols has been exponentially increased in the last decades. Productivity improvements in agri-food industries are linked to obtain the optimal organoleptic properties from the products demanded by the consumers at the same time that the food waste is reduced . In particular, tomato industry requires compliance with high quality standards and precise process control, where the utilization of fast and accurate analytical tools are more than necessary .
Chemical analyses are sometimes time-consuming and expensive and therefore not suitable for continuous, real-time measurements of quality parameters in tomato sauce production lines. In contrast, visible and near-infrared (VIS-NIR) spectroscopy has been presented as a rapid and non-destructive technique that has gained wide acceptance for food analysis in the last decades  .
Spectra acquired by reflectance or transmission spectroscopy in the VIS (380-750 nm) and NIR (750-2500 nm) regions can provide information from the composition of the sample. Several studies in literature have been focused on the evaluation of the parameters of tomato sauce using VIS-NIR spectroscopy , such as soluble solids content (SSC) , sugar content (brix degrees or ºBx) or acidity (pH measurement) .
However, extracting relevant information from the spectral data can be sometimes a challenging process. Particularly, the VIS/NIR spectrum obtained from a food sample can be distorted due to various interfering factors, such as water content, which highly absorbs NIR radiation; low signal-to-noise ratio; light scattering; instrumental noise; and heterogeneities in the sample. Here, the utilization of advanced analysis techniques and classification methods, such as principal component analysis (PCA), linear discriminant analysis (LDA) and artificial neural networks (ANN) respectively, have been proven as useful tools in these cases .
Current research has been focused on the classification of tomato sauces as a function of fat content (oil). The first step consisted of obtaining a large dataset from tomato sauce samples with different oil content using a typical optical transmission setup comprising a TAKHI halogen light source, two spectrometers (VIS-NIR) and an optical fiber VIS-NIR reflection probe connected to a cuvette holder (see Figure 1).
Obtained dataset is pre-processed using PCA and LDA techniques. PCA is a dimension reduction technique used to simplify the information in a data set while preserving as much of the variability in the data as possible. PCAs are also used as highly efficient unsupervised algorithms for reducing the dimensionality of multidimensional data that exhibits a high level of correlation. PCA was employed here to visualize samples grouped as a function of oil content using only three variables (see Figure 2).
Unlike PCA, LDA is focused on maximizing the separation between known categories in the target variable, rather than finding new axes that maximize variation in the dataset. LDAs in Figure 3 reveal that the data can be separated into classes. Both PCD and LDA analysis reveal the good quality of the data obtained to carry out the ANN classification process.
ANN models were used with pre-processed datasets to predict the oil content in tomato sauce, both for VIS and NIR samples. Tomato sauce samples without oil content (0%) were classified with 100% accuracy. In particular, the classification accuracy of tomato sauce with lower content (0%, 1%, 2% and 3%) is better (higher than 70%) than the classification of high oil content tomato sauce.
One of the future challenges of this work is to apply this methodology to tomato sauces with diverse composition through the utilization of more complex ANNs as well as further research with other tomato sauce parameters, such as ºBx or pH.
Overall, the findings presented here suggest that the combination of optical spectroscopy in the VIS-NIR regions and ANN can be a valuable tool for the classification of tomato sauce or other fluids as a function of % oil content or other parameters.
 A. Gardeazabal, T. Lunt, M. M. Jahn, N. Verhulst, J. Hellin, B. Govaerts, “Knowledge management for innovation in agri-food systems: a conceptual framework”, Knowledge Management Research & Prac-tice, vol. 21(2), pp. 303-315, 2023.
 B. G. Osborne, ‘Near-Infrared Spectroscopy in Food Analysis’, in Encyclopedia of Ana-lytical Chemistry, R. A. Meyers, Ed., Chich-ester, UK: John Wiley & Sons, Ltd, 2000, p. a1018. doi: 10.1002/9780470027318.a1018.
 V. Cortés, J. Blasco, N. Aleixos, S. Cubero, and P. Talens, ‘Monitoring strategies for quality control of agricultural products using visible and near-infrared spectroscopy: A re-view’, Trends Food Sci. Technol., vol. 85, pp. 138–148, Mar. 2019, doi: 10.1016/j.tifs.2019.01.015.
 D. Sun, J. Cruz, M. Alcalà, R. Castillo, S. Sans, and J. Casals, ‘Near infrared spectros-copy determination of chemical and sensory properties in tomato’, J. Infrared Spectrosc., vol. 29, p. 096703352110187, Jul. 2021, doi: 10.1177/09670335211018759.
 Y. Huang, R. Lu, and K. Chen, ‘Assessment of tomato soluble solids content and pH by spatially-resolved and conventional Vis/NIR spectroscopy’, J. Food Eng., vol. 236, pp. 19–28, Nov. 2018, doi: 10.1016/j.jfoodeng.2018.05.008.
 L. Xie, Y. Ying, H. Lin, Y. Zhou, and X. Niu, ‘Nondestructive determination of soluble solids content and pH in tomato juice using NIR transmittance spectroscopy’, Sens. In-strum. Food Qual. Saf., vol. 2, no. 2, pp. 111–115, Jun. 2008, doi: 10.1007/s11694-008-9031-7.
 D. Aykas, K. Borba, and L. Rodriguez-Saona, ‘Non-Destructive Quality Assessment of Tomato Paste by Using Portable Mid-Infrared Spectroscopy and Multivariate Analysis’, Foods Basel Switz., vol. 9, Sep. 2020, doi: 10.3390/foods9091300.