Computer vision-based wood identification: an approach with LPQ and GLCM descriptors integrated with Ensemble classification

Given the recognized relevance of botanical identification for forest management and biodiversity conservation, this paper proposes the identification of plant species wood anatomy using computer vision techniques. Microscopic images of 30 species were sourced from the Forest Species Database and captured using an Olympus CX40 microscope at 100x magnification, ensuring high detail. Pre-processing techniques, including normalization and feature extraction, were applied to enhance texture and color characteristics. Statistical, structural, and spectral descriptors such as LBPHF 24,3 and the combination of LPQ and GLCM (GHS) were utilized. These descriptors were analyzed with MATLAB and classified using robust methods, including Ensemble classifiers. Cross-validation ensured reliability, and results achieved assertiveness rates of 96.7% and 99.3% for LBPHF and LPQ-GLCM combinations, respectively. This study demonstrates the effectiveness of automated methods in enhancing botanical identification processes, offering precise and efficient tools for forest management and biodiversity conservation.