Mapas de visibilidade em grandes terrenos representados por grades regulares

• Autor(a) principal: Ferreira, Cháulio de Resende
• Data de Publicação: 2014
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/2665
• topic: Geoinformática; Visibilidade; Algoritmos para memória extera; Computação de alto desempenho; Geoinformatcs; Visibility; Algorithm for external memory; High performance computing; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

In geographical information science (GIS) it is usual to compute the viewshed of a given point on a terrain. This point is usually called observer, and its viewshed indicates which terrain regions are visible from it. In this work we present two novel algorithms for viewshed computation on large grid terrains, both more efficient than other approaches found in related work. The first algorithm is called TiledVS . It was specially designed for external memory processing, that is, it performs a smaller number of in/out (I/O) operations. To do that, it uses a special library called TiledMatrix, which subdivides the terrain into several rectangular blocks and efficiently manages the accesses to the terrain cells. According to our experimental results, TiledVS is more than 4 times faster than all other previous algorithms. The second one is a parallel algorithm that uses the shared memory model (OpenMP). It subdivides the terrain into several sectors around the observer, such that each sector may be processed independently. Our experimental results showed that, using a personal computer with 4 cores, it is possible to compute the viewshed 4 times faster than with the serial implementation of the same algorithm. Using a computer with 16 cores, we obtained up to 12 times speedups.