Clasificación de cobertura del suelo sobre área urbana a partir de imágenes satelitales de mediana resolución empleando regresión-kriging: una comparación con otros métodos no convencionales
Fecha
Autores
Autor corporativo
Título de la revista
ISSN de la revista
Título del volumen
Editor
Compartir
Director
Altmetric
Resumen
In this paper the classification of land coverage on urban area from medium resolution satellite images using regression-kriging was proposed. In order to evaluate the performance of the regression-kriging method was made a classification of a satellite image, it was compared with other classifications obtained by the Support Vector Machines, Distance Mahalanobis, Decision Trees, Artificial Neural Networks methods and Random Forests. The classification of land coverage was made on a SPOT 5 satellite image. The study area corresponds to a central area of the city of Bogota, Colombia; the program selected to process the image was the software "R". Within the results of work is implementing a methodology for classifying satellite images using regression-kriging, the code developed in "R" to classify images by the six with the above methods, six classifications maps, the confusion matrices, the confidence intervals and other indexes to evaluate the accuracy of the classification. All methods showed a good performance in the classification task and thus it was found that in the case of the classification of land cover in urban areas is more effective the regresión-kriging method that considers not only the spectral characteristics of the image also the structure of spatial correlation existing between the data.