Elaboración de modelos 3D de diferentes morfologías y escalas utilizando técnicas Structure-from-Motion y fotografías terrestres

Abstract

Three-dimensional photo-reconstruction methods based on the use of Structure-from-Motion (SfM) and MultiView-Stereo techniques (MVS) are tested together for measuring, monitoring and quantifying three different geomorphological features: i) the Corral del Veleta rock glacier (Granada, Spain), ii) calanchi type badlands (Sicily, Italy) and iii) five small gully headcuts (Cáceres, Spain). Results about the accuracy, usefulness and applicability of these techniques are presented here. In order to quantify the accuracy, the root mean square error of the control points used during the georeferencing procedure and the average absolute distance between each point cloud and a benchmark cloud, calculated with the cloud-to-cloud and the multiscale model-to-model comparison techniques, were used. For the cases i) and iii) the benchmark point clouds were acquired by means of a terrestrial laser scanner, while for the case ii) a digital elevation model obtained by means of classical photogrammetry applied to aerial pictures taken by an unmanned aerial vehicle was used as benchmark. The results of these analyses showed centimetre-level accuracies with average distances to the benchmark models ranging from 0.01 m to 0.42 m. The estimated accuracies show that models are suitable for estimating geomorphic changes only in the case of gully headcuts. In the case of the glacier and the badlands, the obtained models are not suitable to estimate changes at an annual scale, but could be useful to produce high-quality maps or estimating morphometric parameters.