Modelo de sensibilidad de los parámetros de frecuencia de la Transformada de Trazo para la clasificación de imágenes aéreas.

Abstract

Aerial or satellital images can be used to produce terrain coverage maps, which in time are a very useful and important tool for decision-making people in several fields including biodiversity, telecommunications and natural disaster management. The Trace Transform method can be used to process these images. This method extracts features from the images by applying a series of functionals to produce a numeric representation that will be used for classification later on. This model depends on several factors in order to have an efficient operation, among them, the frequency parameters of the traces, classifier type and land coverage. The purpose of this research is to create a model of sensibility to the frequency parameters of the Trace Transform for the classification of textures in aerial images. Experimentation on feature extraction time and precision rate revealed that the frequency parameters, specially the pixel selection paramenter, Dt, and the classifier type have a big effect both of them.