Índices de vegetação para Mauritia flexuosa utilizando imagens VANT-RGB

Abstract

The differentiation of Mauritia flexuosa tree crowns in vereda environments within the Cerrado biome poses significant challenges due to the high structural heterogeneity of the landscape and the spectral limitations inherent to RGB sensors. This study evaluated the capacity of vegetation indices derived from Unmanned Aerial Vehicle (UAV) RGB imagery to distinguish species crowns from surrounding vegetation. Individual crowns and their respective surrounding areas were delineated from orthomosaics, followed by pixellevel analysis of the Excess Green Index ExG, TGI, and VDVI. The multivariate structure of the dataset was examined using Principal Component Analysis (PCA), with subsequent detection and removal of spectral outliers through the DBSCAN algorithm. Index distributions were assessed using histograms at multiple class resolutions, and uncertainty was estimated via bootstrap resampling. Statistical comparisons between crowns and surrounding areas were performed using non-parametric tests (Kolmogorov–Smirnov and Mann–Whitney U), complemented by effect size estimation through Cliff’s Delta. The results demonstrated statistically significant separation between classes (p < 0.001), with greater discriminatory power observed for VDVI (Δ≈0.62) and ExG (Δ≈0.61) compared to TGI (Δ≈0.49). PCA indicated that more than 64% of total variance was explained by the first two principal components, supporting the presence of a consistent spectral pattern. It is concluded that indices emphasizing the green channel provide greater efficiency in distinguish Mauritia flexuosa from the environment in RGB imagery, highlighting their potential for rapid inventories and environmental monitoring of veredas in the Cerrado.