Demonstration of structure-from-motion (SfM) and multi-view stereo (MVS) close range photogrammetry technique for scour hole analysis

摘要

Comprehensive data collection remains a challenge in the field of sediment research. The manual acquisition of fine-gridded data is almost infeasible even for a laboratory setup. Therefore, this paper demonstrates a simple and cost-effective SfM-MVS technique to acquire accurate morphological data. This data further can be used for assessing the scour development around a bridge pier. For this purpose, the experiments are conducted for clear-water scour around circular and hexagonal piers for three different discharges. Before the start of the experimental run, a set of overlapped images is taken using the digital camera. Once the experiment run is completed, the water in the flume is completely drained off and then again another set of overlapped photos are taken. A total of eight ground control points (GCPs) is used to transform the generated relative three-dimensional cloud points to the absolute local coordinate system. Eventually, the high-spatial resolution digital elevation models (DEMs) are generated using the SfM-MVS photogrammetry technique. A statistical analysis is performed between the checkpoints (observed data) and DEM predicted points, which revealed that the generated DEMs show high accuracy in all the cases. It is therefore concluded that the SfM-MVS technique can be applied to understand the morghological changes around any shape of the piers. Thus, the proposed image analysis method can be adopted for obtaining the high spatial resolution data for sediment transport research.