Abstract

Improving terrain tile data selection efficiency, real-time loading of visible tile data and building GPU-based continuous Level of Details (LOD) are the key technologies for large scale terrain rendering on GPU. In this article, in or- der to reduce terrain tile data selection time, we build double layers tile quad tree for massive terrain data and organize tile data by designing Z-order space filling curve. According to the visible region coordinates obtained by GPU offscreen ren- der to texture, we realize real-time loading of visible tile data from CPU to GPU. Map visible tile quad tree into two- dimensional texture on GPU making full use of the characteristics of GPU multi-channel parallel processing. In order to execute tile error metric computation and LOD selection on GPU, we design GPU-based cuboid bounding unsaturated er- ror metric, reduce CPU computational burden and enhance the terrain rendering performance. Experiments show that our algorithm can improve the utilization rate of GPU in the terrain rendering and achieve a good visual effect and a high frame rate.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call