A simple method for Eucalyptus species discrimination: FTIR spectroscopy and machine learning

Abstract

Wood sample classification is crucial in forestry, woodworking, and conservation efforts. It involves categorizing wood species based on physical characteristics like grain patterns, color, texture, and density. However, there are scientific gaps that hinder its effectiveness. One challenge is the difficulty in differentiating wood species with similar characteristics. To address this, we propose using FTIR spectroscopy and machine learning algorithms. In our study, we focused on the accurate classification of Corymbia citriodora (CIT), Eucalyptus grandis (GRA), and the hybrid species GG100 (GG). We evaluated bark and sapwood samples to determine the best for such a purpose. Using FTIR spectroscopy and Quadratic Discriminant Analysis with only seven PCs, we achieved 100% accuracy in bark sample classification. This approach offers advantages because FTIR spectroscopy captures unique chemical fingerprints, providing rapid and non-destructive analysis of wood samples. Machine learning algorithms learn from these spectral patterns for accurate species prediction. Notably, bark samples, easily obtained without tree damage, showed excellent classification results. By incorporating FTIR spectroscopy and machine learning, we can overcome the challenges of inter-species similarities and intra-species variations. This improves reliability and accuracy, supporting and ensuring high-quality wood products. Integrating FTIR spectroscopy and machine learning presents a promising solution for enhancing wood sample classification. Further research can expand species classification and improve wood identification processes.