By Kofi Nyarko, June 21, 2009

     NASA’s goal of improving aviation safety through the integration of advanced technologies within the cockpit is slowly being realized.  One such technology, termed Synthetic Vision System (SVS), using computer hardware and sensors onboard the aircraft to create a model of the terrain below that is viewable from the cockpit.  This model would enable pilots to always have a clear view of their surroundings.  One of the challenges associated with this technology is being able to dynamically create this terrain model in real-time from the database that holds the raw information about the terrain’s elevation.

     A graduate student at Morgan State University, Amaro Thiam, has been working on this very problem. His research focuses on real-time rendering of large scale, detail rich terrains. Although terrain rendering algorithms have been studied for a long time, it is still a very active field in computer graphics. High performance terrain rendering techniques all depend on special pre-computed data for a fixed size, static terrain, which makes them unable to deal with the detailed surfaces generated at run-time. His approach involves the implementation of open source solutions, efficient large-scale terrain rendering techniques using out-of-core visualization and view dependent refinement algorithms. 

     GIS (geographical information systems) software packages exist for complex terrain and visualization analysis. However these packages are not easily customizable and rarely work in real-time. Mr. Thiam began his research by developing a custom platform for terrain data visualization through OpenGL, an open source graphics library that includes features of interests specified by NASA. He later incorporated a terrain rendering algorithm called Stateless One-pass Adaptive Refinement (SOAR), and later SOARX, which was better suited for displaying large scale, detail rich terrains in real-time.

     The SOAR algorithm was developed by Peter Lindstrom and Valerio Pascucci at the Lawrence Livermore National Laboratory.  It contains a relatively simple, yet very powerful terrain rendering framework. It has many independent components: an adaptive refinement algorithm with optional frustum culling and on-the-fly triangle stripping, together with smooth geomorphing based on projected error, and a specialized indexing scheme for efficient paging of data required for out-of-core rendering. The refinement algorithm generates a new mesh (the structural underpinning for the terrain) from scratch with every frame. Mr. Thiam experimented with various methods of displaying huge datasets by employing continuous level of detail (LOD) algorithms and RGB/HSL color space conversions.  It is expected that his research will aid NASA’s effort to incorporate terrain rendering technologies within the cockpits of general aviation aircrafts.