The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLII-3/W12-2020
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-3/W12-2020, 177–182, 2020
https://doi.org/10.5194/isprs-archives-XLII-3-W12-2020-177-2020
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-3/W12-2020, 177–182, 2020
https://doi.org/10.5194/isprs-archives-XLII-3-W12-2020-177-2020

  06 Nov 2020

06 Nov 2020

SUGARCANE PRODUCTIVITY ESTIMATION THROUGH PROCESSING HYPERSPECTRAL SIGNATURES USING ARTIFICIAL NEURAL NETWORKS

C. E. Espinosa, S. Velásquez, and F. L. Hernández C. E. Espinosa et al.
  • Remote Sensing Research Group, Universidad del Valle, Santiago de Cali, Colombia

Keywords: Neural Networks, Net Primary Productivity, Deep Learning, Backpropagation, Hyperspectral Signatures, Sugarcane

Abstract. This project uses an artificial neural network to calculate the net primary productivity of an organic sugarcane crop in Hatico’s farm, in Cerrito, Valle del Cauca. The pilot scheme used in this project is composed by 6 treatments of nitrogen fertilization based on green manures (poultry manure and cowpea). During the last two crops’ phenological phases, the artificial neural network was provided with hyperspectral data collected in the field. In addition, an exploratory data study was implemented in order to identify anomalous signs related to the light saturation and the curvature geometry. The first network applied was Autoencoder, in order to reduce the dimensionality of the radiometric resolution of the data. The second network applied was Multilayer Perceptron (MLP), to calculate the productivity values of the patches. After having compared the actual productivity values provided by Cenicaña, this project obtained an accuracy of 91.23% in the productivity predictions.