Structure from Motion in alpine environment - evaluation of the method and assessment of channel topography change on the example of the mountain stream Ove dal Fuorn
Project Number: CH-6623
||01/01/2019 - 05/31/2019 project completed
||Herr Samuel Wiesmann
Phone: +41 (0) 81 851 41 11
In the last years Digital Terrain Models (DTMs) created with Structure from Motion- Multi-View Stereo (SfM-MVS) have been widely used in the field of geomorphology. So far, little attention has been paid to the distribution and number of Ground Control Points (GCPs) for georeferencing the model in a longish research area. Therefore, the aim of this study is to test three different distributions and three different numbers of GCPs using two commercial software programmes (Agisoft PhotoScan and Pix4Dmapper) in a mountain stream environment (Ova dal Fuorn) in Switzerland. For the evaluation of the GCP settings, DTMs of Difference (DoDs) and two error metrics (Root Mean Square Error (RMSE) and Mean Absolute Error (MAE)) were deployed. For the software comparison a number of aspects have been taken into account, namely he user friend- liness, the processing time, the model quality and the software pricing. It was found that the best model quality is achieved when using the highest number of GCPs – in this case 42–spread equally over the study area. Here for both programmes a low RMSE of 0.05m and a low MAE of 0.03m for Pix4Dmapper and 0.04m for PhotoScan could be reached. The software comparison showed that neither software is clearly superior over the other. Pix4Dmapper is slightly better in terms of accuracy and user-friendliness, whereas PhotoScan runs more stable while processing and has a more attractive pricing system, especially for research.
One field where SfM-MVS DTMs are used is for geomorphic change detection. Due to a heavy thunderstorm in the summer of 2017 the stream bed of the Ova dal Fuorn experi- enced severe changes. Therefore, in a second step the best created DTM was compared via a DoD with a Light Detection and Ranging (LiDAR) DTM from the year 2009 to detect and quantify the changes. To be able to classify the changes and to be able to take a look at the temporal development, the 2009 DTM was compared with a LiDAR DTMs from 2003. Via the DoDs, the areas where changes occurred were located and their volume in cubic metre was calculated. The results showed that the changes between 2009 and 2018 were more severe than between 2003 and 2009 and can most likely be linked to the 2017 event.
Thiex, Daniel. 2019. Structure from Motion in an alpine environment – evaluation of the method and assessment of channel topography change on the example of the mountain stream Ova dal Fuorn (Swiss National Park, SNP). Master thesis, Eberhard Karls University, Tübingen.pdf Master thesis
Last update: 3/23/20
Source of data: ProClim- Research InfoSystem (1993-2020)
Update the data of project: CH-6623