A multi-scale terrain mapping method using hillshade abstraction

Terrain generalization aims to selectively simplify terrain details across scales while preserving the overall terrain form and key geomorphic features. An effective representation of terrain structure provides an important basis for preserving terrain structure during generalization. To support this goal, this study introduces hillshading to represent terrain form and structural information through variations in illumination. A multiscale terrain generalization method was developed based on terrain-structure representation, terrain-unit partitioning, and terrain-surface fitting. First, we construct a multi-azimuth color hillshade (MACH) by mapping multi-view hillshades to RGB channels, thereby enhancing the separability of terrain orientations, gradients, and structural boundaries. Terrain-aware superpixel segmentation was then used to derive terrain units aligned with major ridges and valleys, with the number of units controlling the generalization granularity. Finally, minimum bending energy surface fitting was applied within each unit to suppress secondary relief variations while preserving dominant geomorphic structures and surface continuity. Experiments across terrains with different relief characteristics show that the proposed method achieves effective and scale-controllable terrain generalization, progressively suppressing secondary relief variations as the generalization intensity increases while preserving major geomorphic structures and surface continuity. Overall, the proposed method provides a scale-controllable approach to structure-preserving terrain generalization.