Application for Geospatial Data Processing Based on CUDA

Abstract - In this day and age, and especially over the past decade, satellite imagery (geodata) has become increasingly important in exploring our planet, creating weather forecasts, predicting natural disasters, as well as in utilizing online maps or apps showcasing traffic jams. However, due to the fact that such images have a very high resolution and therefore consume a lot of memory, they cannot be processed quickly. Researchers indicate that the solution for the acceleration of geodata processing lies within the use of the graphic processing unit (GPU) instead of the processor. Nevertheless, there are virtually no programs on the market that use a GPU to address this issue, and those that do are very expensive. The object of development is a parallel CUDA application for processing satellite imagery. The main goals are for this software to have cross-platform GUI for Windows and Linux, as well as having the ability to run as terminal application on servers. C++ is used for both CPU and CUDA parts of the software. The paper consists of 43 pages, 4 chapters, 4 tables, 25 images, 13 references, 4 appendices. Keywords: Geodata, GPU, Image Processing