DONATE

IAHR Document Library


« Back to Library Homepage « Proceedings of the 38th IAHR World Congress (Panama, 2019)

Bed Deformation Measurement Based on Digital Image Processing in a Physical Model

Author(s): Chunjing Liu

Linked Author(s): Chunjing Liu

Keywords: River model test; Physical model with a mobile bed; Structure from Motion; Accuracy;

Abstract: Structure from Motion (SfM) technique, which can get three-dimension information from a series two-dimension photos taken for a specific area with different view angles, have been proven to be very powerful and useful in fields of photogrammetry for field measurement. However, conventional methods such as point gauge is still as the dominating tools for river deformation measurement in physical model experiments, exception few 3D laser scanner applications, which is very expensive and cannot afford for common researcher. In this presentation, SfM technique were introduced for measuring of the topography in a physical model with mobile bed. The model sediment was model sand with different type, size and color. About 60 images were taken around a segment from the physical model with about 1.5m length and 4m width. Then the images were utilized to estimate the three-dimension structure of the physical river bed surface by SfM algorithm. Elevation of a reference cross section in the same model segment was also measured by a point gauge for comparation. The results show that the elevation get from SfM method is very close to the point gauge methods, the maximum deviation was less than 2mm to 3mm. Appropriate control points selection in the images would significantly improve the accuracy of Digital Elevation Model (DEM)generation process during SfM calculation. Compare to the point gauge, the SfM methods can get DEM of bed deformation very quickly and effectively, it would be a very promising tool in physical model experiments.

DOI: https://doi.org/10.3850/38WC092019-0334

Year: 2019

Copyright © 2021 International Association for Hydro-Environment Engineering and Research. All rights reserved. | Terms and Conditions