Ringo-arhitektura za trodimenzionalnu vizuelizaciju terena velikih razmera
ДокторандDimitrijević, Aleksandar M.
МенторRančić, Dejan D.
Чланови комисијеMilovanović, Emina
МетаподациПриказ свих података о дисертацији
This doctoral dissertation presents an original approach to the planet-sized terrain visualization. The developed algorithm and associated software architecture allow geodetically correct and precise representation of the entire planet in real time, regardless of the viewer’s distance from its surface. Verification and validation of the results are achieved by the use of formal and precise method, through the practical implementation based on the real data for the planet Earth, defining metrics, obtaining numerical values and finally publishing the results in a specialized scientific journal. The visualization is based on the modified geometry clipmap algorithm. The proposed modification enables the generation of sub-pixel precise WGS84 reference ellipsoid on the graphics processor, using single precision arithmetic. Regardless of the limited accuracy of this arithmetic, the proposed method allows very fast, accurate and consistent visualization of the WGS84 reference ellips...oid, from cosmic to extremely small distances to its surface. Dividing the ellipsoid into three partitions, with the use of equidistant cylindrical projection for each of the partitions, minimizes the distortion of the basic grid upon which the terrain is elevated, and thus the distortion of the applied textures. The described algorithm allows high and stable frame-rate, even when modest hardware is used, with a very low utilization of the CPU and an effective usage of the GPU. An efficient method for texture streaming and accompanying software pipeline are proposed, making the visualization independent of the applied texture-coverage resolution. A long-term research of terrain visualization methods has been followed by the constant development and improvement of appropriate software, which allowed the validation of all the ideas generated during the research. The software is developed in C++ programming language, using MS Visual Studio environment. As an interface to the graphics subsystem, the OpenGL API is used. Due to the specific implementation, OpenGL version 3.3 or later is required and graphics hardware that supports it. The visualization uses real data gathered by satellite and aerial remote sensing, as well as currently the most accurate models and data related to the evaluation of the size and shape of the planet Earth. Verification of the implementation accuracy was based on established metrics and values taken after the transformation on the graphics processor. The efficiency of implementation is estimated using a special library, developed for this purpose, that is able to accurately measure the time elapsed between two function calls, both at the central and the graphics processor. Since these two processors are completely asynchronous, it is necessary to monitor their execution time in parallel. The performance significantly depends on the current state of the system, hence, the analysis includes other parameters and graphics subsystem states, obtained by specialized library provided by hardware manufacturers. In addition, the efficiency of the proposed algorithm and its implementation are compared with the current state-of-the-art algorithms in this field. The final verification of the proposed ideas and implementation represents the acceptance and publication of research results in the eminent journal for computer graphics and visualization - Computers & Graphics Journal (UK)