REPRESENTAÇÃO EM MALHAS 3D A PARTIR DE DADOS DE TEXTURA DO SOLO INTERPOLADOS MEDIANTE REDE NEURAL ARTIFICIAL: ESTUDO DE CASO FESCON - PONTA GROSSA – PR

Abstract

The physical properties of the soil have great impact on their behavior, and these characteristics result in classification of the profile and soil suitability. In the case of Precision Agriculture is important to identify in the soil particle size space distribution or texture. Thus, from granulometric georeferenced data (sand, silt and clay) soils collected from a farmland Farm School Capao-da-Onca (FESCON Ponta Grossa - PR), was performed interpolation using Radial Basis Function (RBF) and supervised training for Artificial Neural Network (ANN), comparing the results obtained on a 3D model in order to verify the performance of the RNA used. It was found that especially the representations of RNA with the sand attribute was smoothed when interpolated the granulometric data in realation to RBF. The attributes clain and silt had some variations between ANN and RBF, it doesn't always smoothed.