Evaluación y propuesta de metodologías de clasificación a partir del procesado combinado de datos LiDAR e imágenes aéreas georreferenciadas

Abstract

In the past few decades, LIDAR technology, Light Distance and Ranging, has contributed to represent the land surface in a massive way. It constitutes a revolution in the field of Earth sciences. The volume of the captured data justifies the need to develop methodologies that allow for the classification of the point clouds, with the object of facilitating its treatment. For this purpose, a methodology based on data mining is proposed, specifically focused on the Random Forest algorithm. It sets out to classify these points according to their mapping use at three categories: buildings, transport networks and vegetation. The results show that this is an appropriate methodology for the classification of the points in buildings and areas with vegetation. It should experiment with new procedures for the prediction of the transport networks.