Generación de cartografía vectorial automática para cuerpos de agua aplicando análisis multiresolución en imágenes satelitales de alta resolución espacial en la Sabana de Bogotá

Abstract

A methodology is proposed for the generation of automatic vectorial cartography water surfaces taking advantage of the photointerpretation parameters for satellite images that combined with a multiresolution analysis allow to exalt, locate and automatically segment the bodies of water present in the images. Starting from the hypothesis that water surfaces, due to their spectral response in satellite images have significantly higher values in textural analysis of homogeneity and energy, these descriptors are calculated derived from the co-occurrence matrix on the horizontal and vertical details of second level of wavelet decomposition, once the water surface is located inside the image, the snake is implemented, an active contour technique to refine the selection and contour of the object to finally vectorize the selected water surface and generate graphic outputs. The methodology was implemented in 10 square scenes covering the zone of the sabana de Bogotá, 9 of the scenes were taken from PlanetScope images with a spatial resolution of 3 meters and a size of 2233 pixels side, the last scene was extracted from an IKONOS image with a spatial resolution of 1 meter and a size of 2248 pixels side. The evaluation of the methodology was made by calculating five metrics: Dice Score, Kappa, global accuracy, user accuracy and producer accuracy; For the calculation of these metrics, a manual digitalization of the water surfaces was carried out and compared with the automatic segmentation obtained with the proposed methodology, the average of all the evaluated metrics exceeds 86%, which indicates that the methodology is effective in terms of thematic accuracy for the generation of vector cartography of water surfaces.