Determinación del índice de electrificación mediante el procesamiento de imágenes satelitales para la zona rural del área de concesión de la Empresa Eléctrica Regional Centro Sur. C.A.

Abstract

The importance of addressing the use of technological tools to solve problems facing society is becoming more and more evident. Among these problems is the determination of the electrification index, an indicator that allows measuring the level of access to electricity. In Ecuador, finding this metric is limited to census applications that are expensive and require a large number of staff. Thus, in the present work, Deep Learning was applied for the analysis of satellite images in the ArcGIS Pro software. The objective is to extract the traces of the houses in 18 parishes of the Cuenca canton as a representative sample of the rural concession area of the “Empresa Eléctrica Regional Centro Sur” (Centrosur). For this, it is proposed to work with Convolutional Neural Networks, using the Mask-RCNN model, obtaining results with an average precision of approximately 94% and an F1 Score of 84%. Finally, the metric corresponding to the electrification index is calculated and a comparison is made with the statistics provided by the “Instituto Nacional de Estadística y Censos” (INEC).