Analysis of biodegradable films made from cassava starch and oregano essential oil using hyperspectral imaging and portable NIR spectroscopy

The plastics industry has been focusing on the substitution of natural raw materials into biodegradable plastics. Cassava starch-based (CS) films, incorporated with cellulose (CL) and oregano essential oil (OEO) are edible, biocompatible, odorless, tasteless, transparent and colorless. However, the study of the properties and characterization of these bioplastics requires extensive scientific and technical knowledge. In this context, the vibrational techniques in the NIR region stands out as one of the most promising due to its effective interaction with the chemical compounds present in polymer matrices. This work aims to compare the performance of near-infrared hyperspectral imaging (NIR-HSI) and near-infrared spectroscopy (NIRS), in combination with machine learning, to estimate the percentage and distribution of oregano essential oil on the surface of cassava starch-based bioplastics. The films structured with OEO exhibited higher relative humidity (RH) values than films without OEO, and film thickness was uniform across all formulations. Lightness values indicate that films with OEO and CL are darker, thus the opacity of the films increased with the addition the OEO. Partial Least Squares Regression (PLSR) and Support Vector Machine Regression (SVMR) were used to predict film composition, and Partial Least Squares-Discriminant Analysis (PLSDA) and Support Vector Machine Classification (SVMDA) were used for classification models. It was demonstrated that the SVM provides the best results for prediction and classification. NIR-HSI was successfully used to predict the amount of OEO on the surface of the films. However, for classification of film based on its composition, NIRS is a promising alternative as screening technique, as both NIR-HSI and NIRS can identify patterns responsible for the variations in the biodegradable polymers studied.