在由立体像对生成数字高程模型(DEM)的过程中,由于匹配算法的限制和地物之间的相互遮挡,经常不能直接恢复出所有点的高程数据,因此采用某种插值算法由已有的数据点得到更为稠密的地表三维数据,便成为建模过程中一个不可缺少的重要环节.该文在分形几何学的基础之上,提出了一个基于地表分形特性的空间多面体变形插值算法.该算法的独特之处在于,它可以在原始数据点杂乱和不规则分布的情况下得到接近于真实地貌的插值数据.最终的实验结果令人满意.

In constructing Digital Elevation Model (DEM) from image pairs, often the elevation data of point cannot be directly recoverd, due to the limitation of the mapping algorithm adopted and the mutual hiding of the terrain itself. Therefore it is very important to construct a dense terrain model from these sparse data by means of interpolation. In this paper, based on the fractal features of natural terrain, a Spacial Polyhedron Model Deformation algorithm is presented, which can generate realistic dense terrain model from sparse points that are irregularly distributed. The experimental results are satisfactory.