Uso de subárvores especializadas para obtenção de desfazer e tutorial localizados em sistemas de controle de versão de modelos 3D

• Ano de defesa: 2016
• Autor(a) principal: Vieira, Rafael Siqueira Telles
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Controle de versão 3D; Subárvores especializadas; Desfazeres localizados
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/21524

A system that records changes made to a file is called a Version Control System (VCS). Even though VCSs may store all kinds of files, our thesis focus on changes made to polygonal mesh files. The data structure used by those systems are trees or directed acyclic graphs, and our work focuses on version control trees (VCTs). Our method allows the user to track the history of topological and geometrical changes made to a mesh region to generate a tutorial or to undo some operations made to that region. The region to be tracked is selected through a bounding box selection mechanism, which is propagated to all VCT nodes so we can find the intersection sets between the bounding-box selection and Changesets, i.e., the difference between two consecutive versions. To identify the Changesets, the algorithm maps elements from one version to another using topological propagation and hash tables, which are responsible for propagating the border between mapped and unmapped elements, and to create a unique key for elements that do not change between consecutive versions. With these data, it is possible to construct a subtree associated with the selected region to serve as a tutorial on how that region was modeled. That subtree also allows the user to undo local changes, i.e., operations that do not change all elements of a mesh version, to save modeling time, which is normally high even for small meshes. This undo operation acts just over a specific region and it is a specialized undo, as such, we call it bounded undo, which is an important contribution in this Thesis. Despite the important contributions, we also point out some current limitations to our method and discuss ways to overcome them. Our technique is also compared with the current state of the art version control systems for meshes, and is analyzed with respect to efficiency of unique key generation and speed, which are two important points of the algorithm, since it deals with large datasets.