Interactive visual analysis of hermetic and dense virtual models: the case of grain structures.

• Ano de defesa: 2021
• Autor(a) principal: Rodrigues, André Montes
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Grãos (Estrutura),; Microstructure visualization; Modelos herméticos; Multiscale visualization; Space-filling models; Visualização de microestruturas; Visualização em múltiplas escalas
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/3/3142/tde-20122021-103140/

Novel interactive techniques were designed pursuing improvements on the comprehension of grain microstructures and similar hermetic (space-filling) honeycomb structures, based on cognitive principles and in the state of the art of interactive systems. A systematic review was carried out on multiscale visualization, which was considered an ideal paradigm in this work. In a second stage, development shifted to scientific experiments to investigate perceptual performance on fundamental aspects for grain structure comprehension - grain size extremes, regions with localized size heterogeneities, and shape of individual grains. Techniques were consequently adapted for experiments, centered on these three visual search tasks. Human-computer interaction knowledge and preliminary experiments informed the definition of experimental logistics and four dependent variables to attest performance, as well as appropriate statistical methods to analyze mixed factorial experiments results. The main experiment was carried out with two groups of 30 participants, comprising members of the intended target audience. Spatial ability pre-tests indicated homogeneity of spatial skills between the groups. Techniques are Discrete Sections (S), Dynamic Sectioning (DS), and GrainCrawler (GC). S simulates a visualization based on sequences of parallel sections performed over short distances, technically feasible to execute on real materials. DS improves upon S on spatial resolution, and the user can cut through the model interactively in any direction. GC was developed after the models peculiarities, multiscale visualization potentials, and educational objectives. On GC, grain representation changes according to distance and perceptual objectives, a distinctive feature of the multiscale paradigm. Visualization modes were developed pursuing performance improvements, based on known depth perception principles, leveraging the models three-dimensional nature. The main goal and hypothesis was that GC would surpass DS and particularly S, and 3D modes would overcome 2D. However, all techniques performed well overall and are deemed useful in improving spatial understanding of this type of structure. They allowed detecting size extremes and heterogeneous regions in a space-filling model containing more than a thousand objects. However, each technique had specific strengths. S excels in the agile detection of large grains and regions, as long as the sections are not widely spaced, and can be applied to real materials. DS allows rotation of the section plane and is ideal for digital models, being the most balanced and flexible technique. Finally, GC excelled in small grain detection and shape analysis, tasks characterized by high complexity and heavy cognitive load. Still, the multiscale paradigm allows combining each techniques strengths according to the intended applications analytical goals.