Study of the microstructure-property-processing relationship in five potato (Solanum tuberosum) varieties during the frying process based on an automatic classification system using convolutional neural networks

Our objective was to identify and analyze the microstructural features of five different Peruvian potato varieties in fresh material and a frying process, using a 32-factorial arrangement of temperature and time. Two types of characteristics were measured. The first ones were of microstructural type (i.e., area, perimeter, length of major axis, length of minor axis, roundness, elongation, and compactness) and the second ones were of physicochemical type (i.e., L*, a*, b*, ∆E, acrylamide concentration, fat percentage, moisture percentage, and texture). For this purpose, potato microstructural characterization software was implemented, developing algorithms for image processing and analysis, as well as the classification of structural characteristics. Potato variety was found to exert a significant effect on the microstructural parameters of area, perimeter, major axis length, minor axis length, roundness, and compactness, followed by time, with a significant effect on the microstructural parameters of area, perimeter, major axis length, minor axis length, and compactness. Temperature exerts a significant effect only on roundness and elongation parameters. To observe the relationship between the microstructural and physicochemical parameters, a Pearson correlation was used where it was observed that the correlations between the physicochemical and microstructural variables evaluated were medium to strong.