Using multidimensional views of photographs for flood modelling

Abstract

Using physically based computational models coupled with remote sensing technologies, photogrammetry techniques, and GIS applications are important tools for flood hazard mapping and flood disaster prevention. Also, information processing of massive input data with refined accuracy allows us to develop and to improve urban-flood-modeling at a detailed level. The topographical information from digital surface model (DSM) or digital terrain model (DTM) is essential for flood managers who actually require this high accuracy and resolution of input data to set up their practical applications. Light detecting and ranging (LiDAR) techniques are mainly used, but these costly techniques can be appraised by equipments, maintenance, and operations which include aircraft. Recent advances in photogrammetry and computer vision technologies like structure form motion (SfM) technique are widely used and offer cost-effective approaches to reconstruct 3D-topographical information from simple 2D photos, so-called 3D reconstruction. In terms of input data for flood modeling, the SfM technique can be comparable to other acquisition-techniques. In this paper, there are one experimental and two case studies. Firstly, a result of the experiment showed a similarity between flood maps by applying the SfM process form the 3D-reconstruction and using benchmark information. These 3D-reconstruction processes started from 2D photos, which were taken from virtual scenes by using multidimensional-view approach. These photos can be used to generate 3D information which is later used to create the DSM from multidimensional fusion of views (MFV-DSM). Then, the DSM was used as input data to set up 2D flood modeling. Thereafter, when using the DSMs as topographical input data, comparison between a benchmark DSM and MFV-DSM shows similarity flood-map results in both flood depths and flood extends. Secondary, the two cases from real world scenes also showed possibilities of using the SfM technique as an alternative acquisition tool, providing 3D information. This information can be used as input data for setting up modeling and can possibly be comparable or even outcompete with other acquisition techniques, such as LiDAR. As a result, using the SfM technique can be extended to become promising methods in practicable applications for modeling real flood events in real world scenes.