Abstract
A digital elevation model (DEM) is a quantitative representation of terrain and an important tool for Earth science and hydrological applications. A high-resolution DEM provides accurate basic Geodata and plays a crucial role in related scientific research and practical applications. However, in reality, high-resolution DEMs are often difficult to obtain. Due to the self-similarity present within terrains, we proposed a method using the original DEM itself as a sample to expand the DEM using sliding windows method (SWM) and generate a higher resolution DEM. The main processes of SWM include downsampling the original DEM and constructing mapping sets, searching for the optimal matching, window replacement. Then, we repeat these processes with the small-scale expansion factor. In this paper, the grid resolution of the Taitou Basin was expanded from 30 to 10 m. Overall, the superresolution reconstruction results showed that the method could achieve better outcomes than other commonly used techniques and exhibited a slight deviation (root mean square error (RMSE) = 3.38) from the realistic DEM. The generated high-resolution DEM prove to be significant in the application of flood simulation modeling.
Highlights
Each sliding window corresponded to a 3 × 3 sample window of the original digital elevation model (DEM)
As we can see that sliding windows method (SWM) works well in the Taitou Basin, we apply SWM on another terrain case to test whether this model is still useful
It is concluded that SWM has an advantage in quantitative accuracy and generalization capability when reconstructing a high-resolution DEM
Summary
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. Xu et al [7] proposed a CNN-based model that is broadly derived from the enhanced deep superresolution network (EDSR) [33]. This network is pretrained with natural images, it still requires a number of DEM samples. We propose considering both elevation and slope information when sliding windows on the input DEM to obtain the terrain self-similarity rules. By using the SWM to obtain the terrain self-similarity rule to perform superresolution reconstruction of a low-resolution DEM, we can obtain higher resolution elevation information.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have