The correct management of pastures in agricultural regions plays a fundamental role in the production itself, in the prevention of plant biomass waste and the release of greenhouse gases (GHG). In addition, it is necessary to avoid inappropriate movement of the herd between pastures, as this is a time-consuming process and can be stressful for the animal. The success of this management requires an efficient assessment of the plant resources. The studies developed for this purpose are directly related to the amount estimation of biomass in a specific region of the pasture, because, in practice, it is not carried out accurately, due to the difficulty of measurement throughout the field. This work aims to develop a low-cost biomass estimation methodology, based on regression models that correlate the most relevant input features for the application with the actual density of the plantation, measured in g/m2 . For the features, the height of the forage grass was measured and the vegetation indexes based on RGB were calculated from images of unmanned aerial vehicles (UAV). Linear, nonlinear regression (MNLR), artificial neural networks (ANN) based on multi-layer perceptron (MLP) and the combination of all models generated (stacking ensemble) were the methodologies tested in order to achieve the best correlation. Satisfactory results were achieved using models of neural networks with two layers and using stacking ensemble methodology, reaching a RMSE of 31.76 g/m2 , MAPE of 13.35 percent and R-Squared of 0.9. Therefore, the proposed methodology may become a promising and affordable solution for plant biomass estimation toward efficient and sustainable herd management.