关键词: 实时优化适应性网格（ROAM）, 数字高程模型（DEM）, 帧一致性, 二叉树, 多分辨率, 误差树

Abstract: A real-time multiresolution triangulation algorithm of large DEM is established based on ROAM（Real-time Optimally Adapting Meshes）.With frame-coherence and double-tree-referencial triangulation scheme the height field of vertex and deviation of triangles in memory are reused efficiently.As a result，this algorithm can run on out of core mode with data necessary resided in memory.The triangulation deviation is calculated more accurately and processed offline beforehand.A hierarchical deviation tree is built correspondingly.The deviation is fetched from disk immediately during runtime instead of calculated online，from which the algorithm benefits much and is accelerated significantly.And the DEM is approximated more accurately also.A test demo is conducted.The result shows that this algorithm is applicable to multiresolution visualization of large DEM with relative high speed.

Key words: Real-time Optimally Adapting Meshes（ROAM）, Digital Elevation Model（DEM）, frame-coherence, binary tree, multiresolution, deviation tree