Modeling Biometric Variables of Olive Trees Using Linear Regression and Spectral Indices from UAV RGB Imagery

Abstract

This study evaluated the use of unmanned aerial vehicles (UAVs) to estimate biometric variables of Olea europaea L. in a commercial orchard in S˜ao Gabriel, Rio Grande do Sul, Brazil. A total of 349 trees were sampled for total height, stem circumference at 60 cm above ground level, and canopy dimensions. RGB imagery acquired with a DJI Mavic 3M was processed in Agisoft Metashape to generate an orthomosaic, digital models, and vegetation indices. Six RGB-derived indices were used as predictors in linear regression models for stem diameter at 60 cm, total height, and canopy area. Models including vegetation indices showed better performance, with higher adjusted R² and lower errors. UAV-derived metrics were consistently correlated with field measurements, indicating strong potential for inventory, monitoring, and management of olive orchards.