A Global-Information-Constrained Deep Learning Network for Digital Elevation Model Super-Resolution

High-resolution DEMs can provide accurate geographic information and be widely used in hydrological analysis, path planning, urban design. As the main complementary means of producing high-resolution DEMs, DEM super-resolution (SR) method based on deep learning has reached a bottleneck. The reason for this phenomenon is that lacks part global it requires. Specifically, multilevel aggregation process difficulty sufficiently capturing low-level features with dependencies, which leads to lack relationships high-level information. To address problem, we propose global-information-constrained network SR (GISR). our proposed GISR consists supplement module local feature generation module. former uses Kriging information, considering spatial autocorrelation rule. latter includes residual PixelShuffle module, restore detailed terrain. Compared bicubic, Kriging, SRCNN, SRResNet, TfaSR methods, experimental results show better ability retain terrain features, effect more consistent ground truth DEM. Meanwhile, compared method, RMSE improved by 20.5% 68.8%.