Estimation of combustible material in Cerrado grassland area from RGB sensor images

Abstract

The quantification of fuel material in the Cerrado area is limited by the difficulty in obtaining data, the high costs and the high time spent in the field. In search of alternatives that facilitate the data acquisition, the use of RGB sensors stands out being able to be a useful and effective tool in quantifying the combustible material. In this context, the objective of this work was to evaluate the feasibility of using images from an airborne RGB sensor by a multirotor to estimate the combustible material by means of regression analysis. The fuel material was sampled from the area that was weighed in the field and dried in an oven. With the digital images processing, the height (htMDA) and the vegetation index (NGRDI) of the pixels covering the sample units were obtained, followed by a correlation analysis between the digital processing data and the combustible material. Subsequently, three regression models were adjusted, in which adjusted coefficient of determination (R²_aj) was obtained from 0.39 to 0.80. The use of RGB sensors has potential for estimation of combustible material. When the htMDA and NGRDI variables are combined, values closer to the mid-range are obtained.