A portable Raman sensor for the rapid discrimination of olives according to fruit quality

In the real marketplace, providing high-quality olive oil is important from the perspective of both consumers and producers. Quality control should meet all requirements in the production process, from farm to packaging. The quality of olive oil can be affected by several factors, including agricultural techniques, seasonal conditions, farming systems, maturity, method and duration of storage, and process technology. The quality of oil produced also depends largely on the quality of the olives. In an enterprise aimed at producing high-quality oils, olives with defects ('ground'; i.e., fallen to the ground) should be separated from healthy fruit ('sound'; i.e., collected directly from the tree), because a very small portion of low-quality fruit can ruin the whole batch. The fruit falls partly because of its maturation process, but also because of pest and disease attack or weather conditions (strong wind). Fruit that has fallen to the ground can suffer a rapid deterioration in quality. Currently, the separation of fruits is based mainly on visual inspection or information provided by the farmer. These are not very reliable procedures. Methods using analytical parameters to characterize the oil, such as acidity and peroxide value, can be applied, but they require a lot of time and materials. Alternative techniques are therefore needed for the rapid and inexpensive discrimination of olives as part of a quality control strategy. The work described here aims to determine the potential of low-resolution Raman spectroscopy for the discrimination of olives before the oil processing stage in order to detect whether they have been collected directly from the tree (i.e., healthy fruit) or not. Low-resolution Raman spectroscopy was applied together with multivariate procedures to achieve this aim. PCA was used to find natural clusters in the data. Supervised classification methods were then applied: Soft Independent Modeling of Class Analogy (SIMCA), PLS Discriminate Analysis (PLS-DA) and K-nearest neighbors (KNN). The best results were obtained using the KNN method, with prediction abilities of 100% for 'sound' and 97% for 'ground' in an independent validation set. These results demonstrated the potential of a portable Raman instrument for detecting good quality olives before the oil processing stage, by developing models that could be applied before this stage, thus contributing to an overall improvement in quality control.