Dense DSM Generation Using the GPU

Generating dense digital surface models (DSM) is a vital step in all photogrammetric work. Its precise production influences the quality of subsequent products including the digital terrain model (DTM) and orthomosaic. SimActive’s proven DSM capability is employed throughout the industry for its precision and high processing speed. With today’s demands for larger and denser DSMs, a considerable strain is placed on the performance. This is driving the need for improved processing speeds for dense DSM generation. The problem has been tackled with new algorithms and the use of multicore CPUs, but have had little success on the large commercial production scale. This paper discusses the suitability of the GPU to resolve the performance dilemma with dense DSM generation. GPU computing will be discussed, followed by a careful look at the challenges involved and the benefits provided. Finally, sample results will be presented and analyzed.