RECOGNITION OF GEOGRAPHICAL INDICATIONS OF CANEPHORA COFFEES FROM BRAZIL USING A PORTABLE NIR SPECTROMETER IN TANDEM WITH SUPPORT VECTOR MACHINES

Resumo

The geographic origin is an important factor in determining coffee quality, as terroir influences all chemical and sensory characteristics and is commonly provided on labels of specialty coffees. When coffee exhibits its unique origin characteristics, it becomes a reason to add value to the product and protect its geographical indication, as seen in the cases of Brazilian robusta coffees: Robusta Amazônico and Conilon from Espírito Santo. In this regard, there is a demand for the development of efficient analytical methods to classify green coffee beans, enabling authentication of this product. However, the challenge lies in obtaining quick results without the need for sample preparation and without the use of potentially toxic reagents. In this work, a Portable Near-Infrared (NIR) Spectrometer was coupled with Support Vector Machines (SVM) to assess the classification capability of the geographical indications Robusta Amazônico (111 samples) and Conilon from Espírito Santo (105 samples). It is important to note that each sample came from a different producer. Spectra were collected directly from the beans using a MicroNIR spectrometer (Viavi Solutions Inc., USA). Full data were preprocessed with First derivative with Smoothing of Savitzky-Golay, Multiplicative Scatter Correction, and Mean Center. Using the Kennard-Stone algorithm, training (75% of samples) and testing (25% of samples) groups were separated, and a SVM model was developed with the optimization of the Kernel function parameter (γ) and penalty parameter (C) using the Genetic Algorithm. All data processing was performed in MatLab 2019.a software. The methodology was validated based on the determination of sensitivity and specificity values, which reached 100% for both classes. Thus, it can be confirmed that the application developed in this study is efficient, rapid, non-invasive, and clean. Its use can be for quality control of coffees, directly in the field, in industries, laboratories, and geographical indication certification procedures.