An indirect interpolation model and its application for digital elevation model generation

Spatial interpolation is an important facet of generating digital elevation models (DEMs). When faced with geographical big data, a DEM with high accuracy is required in many types of investigations and applications. This requirement has encouraged researchers to enhance current interpolation methods and improve DEM construction as much as possible. Currently, almost all spatial interpolation methods use given sampling points as inputs and then calculate the missing sampling points individually. All of them consider terrain similarities in local areas but fail to consider the influence of the varying tendency in the overall terrain. In this paper, we propose a new concept for spatial interpolation that also considers local terrain similarity. However, in our approach, the unknown elevation points are expressed using both sampling points and unknown neighboring elevation points. In this way, the unknown elevation points of the entire computed region are connected formulaically, which introduces the trend features of terrain changes into the DEM construction process. We select a typical experimental area and conduct a DEM experiment based on elevation data sources. The experiment reveals that compared with traditional methods, the new method constructs a more accurate DEM, with morphological characteristics that are more consistent with the real terrain surface.