Determination of application volume for coffee plantations using artificial neural networks and remote sensing

• Published in: Computers and electronics in agriculture Vol. 184; p. 106096
• Main Authors: Oliveira, Mailson Freire de, Santos, Adão Felipe dos, Kazama, Elizabeth Haruna, Rolim, Glauco de Souza, Silva, Rouverson Pereira da
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01-05-2021; Elsevier BV
• Subjects: Artificial intelligence; Artificial neural networks; Coffee; Coffee canopy; Digital agriculture; Machine learning; Morphology; Multilayer perceptrons; Network topologies; Neural networks; Normalized difference vegetative index; Pesticides; Plantations; Radial basis function; Remote sensing; Variable rate spraying; Vegetation; Vegetation index; Variable rate spraying; Vegetation index; Coffee canopy; Machine learning; Digital agriculture
• ISSN: 0168-1699; 1872-7107
• DOI: 10.1016/j.compag.2021.106096

[Display omitted]
•Spray volume determination for coffee plantation can be done remotely.•Proximal sensing associated with machine learning can access physical variables.•Multilayer perceptron outperformed radial basis function network.
Methods for optimizing the application of phytosanitary products can be an alternative for sustainable agriculture. Such methods can be achieved with the use of artificial intelligence and remote sensing techniques. Our experiments were carried out in a commercial coffee plantation, where morphological variables (height and diameter) and vegetation indexes (normalized difference vegetation index, NDVI and normalized difference red edge, NDRE) were collected in the upper, medium, and lower thirds of the coffee plant. From the remote sensing data, experiments were developed to determine the best neural network topology, in terms of accuracy (RMSE) and precision (R2) and type (Multilayer Perceptron “MLP” and Radial Basis Function “RBF”), to estimate morphological variables. From these results, we evaluated the possibility of applying pesticides at a variable rate, using the tree row volume principle. The results show that, using remote sensing and artificial neural networks (MLP), it is possible to estimate coffee tree volume with reasonable accuracy. This can be done using a multi-layer perceptron model to estimate coffee tree height and diameter using vegetation indexes of different parts of the plant as input.