The topographic skeleton is the primary expression and intuitive understanding of topographic relief. This study integrated a topographic skeleton into deep learning for terrain reconstruction. Firstly, a topographic skeleton, such as valley, ridge, and gully lines, was extracted from a global digital elevation model (GDEM) and Google Earth Image (GEI). Then, the Conditional Generative Adversarial Network (CGAN) was used to learn the elevation sequence information between the topographic skeleton and high-precision 5 m DEMs. Thirdly, different combinations of topographic skeletons extracted from 5 m, 12.5 m, and 30 m DEMs and a 1 m GEI were compared for reconstructing 5 m DEMs. The results show the following: (1) from the perspective of the visual effect, the 5 m DEMs generated with the three combinations (5 m DEM + 1 m GEI, 12.5 m DEM + 1 m GEI, and 30 m DEM + 1 m GEI) were all similar to the original 5 m DEM (reference data), which provides a markedly increased level of terrain detail information when compared to the traditional interpolation methods; (2) from the perspective of elevation accuracy, the 5 m DEMs reconstructed by the three combinations have a high correlation (>0.9) with the reference data, while the vertical accuracy of the 12.5 m DEM + 1 m GEI combination is obviously higher than that of the 30 m DEM + 1 m GEI combination; and (3) from the perspective of topographic factors, the distribution trends of the reconstructed 5 m DEMs are all close to the reference data in terms of the extracted slope and aspect. This study enhances the quality of open-source DEMs and introduces innovative ideas for producing high-precision DEMs. Among the three combinations, we recommend the 12.5 m DEM + 1 m GEI combination for DEM reconstruction due to its relative high accuracy and open access. In regions where a field survey of high-precision DEMs is difficult, open-source DEMs combined with GEI can be used in high-precision DEM reconstruction.

Previous studies have shown that the use of appropriate ground control points (GCPs) and camera calibration models can optimize photogrammetry. However, the synergistic effects of GCPs and camera calibration models on UAV-SfM photogrammetry are still unknown. This study used camera models with varying complexities under different GCP conditions (in terms of number and quality) for UAV-SfM photogrammetry. The correlation matrix and root mean squared error (RMSE) were used to analyze the synergistic effects of GCPs and camera models. The results show that (1) without GCPs, complex camera models reduce distortion parameter correlation and improve terrain modeling accuracy by about 70%, with Model C (with F, Cx, Cy, K1–K4, and P1–P4) being the most widely applicable. (2) Increasing the number of GCPs enhances the terrain modeling accuracy more effectively than increasing the camera model complexity, reducing the RMSE by 45–70%, while the model complexity does not affect the required GCP number. (3) A strong interaction exists between the GCP quality and camera models: High-quality GCPs enhance camera model performance, while complex camera models reduce the requirement of GCP quality. This study provides both theoretical insights and practical guidance for efficient and low-cost UAV-SfM photogrammetry in different scenarios.

Identifying and extracting check dams is of great significance for soil and water conservation, agricultural management, and ecological assessment. In the Yellow River Basin, the check dam, as a system, generally comprises dam locations and dam-controlled areas. Previous research, however, has focused on dam-controlled areas and has not yet identified all elements of check dam systems. This paper presents a method for automatically identifying check dam systems from digital elevation model (DEM) and remote sensing images. We integrated deep learning and object-based image analysis (OBIA) methods to extract the dam-controlled area’s boundaries, and then extracted the location of the check dam using the hydrological analysis method. A case study in the Jiuyuangou watershed shows that the precision and recall of the proposed dam-controlled area extraction approach are 98.56% and 82.40%, respectively, and the F1 score value is 89.76%. The completeness of the extracted dam locations is 94.51%, and the correctness is 80.77%. The results show that the proposed method performs well in identifying check dam systems and can provide important basic data for the analysis of spatial layout optimization and soil and water loss assessment.