Real-time visualization for large scale terrain based on linear quadtree

Abstract

To solve the problem of low frame rate and high memory space cost in large scale terrain visualization, a real-time rendering algorithm based on linear quadtree is proposed in this paper. It consists of three parts. Firstly, digital elevation models (DEMs) are de-sampled to get the same size blocks, terrain pyramid model is built with different scales and biases factors offline. Then linear quadtree index of terrain blocks is created for fast query, a more reasonable scheduling scheme is defined, and every frame only allowed loading one terrain block in order to avoid unstable rendering quality. Finally, GPU-based morphing method is applied to slow down vertex popping; vertical skirt is used to eliminate crack. Compared with Geomipmapping algorithm, the experimental results show the proposed algorithm could drop down memory space, GPU processing time and GPU rendering batches significantly, and the rendering rate is high and smooth. © 2012 Springer-Verlag.