Automatic mesh transformation method for musculoskeletal model

The roadmap [25] states importance of registration of data sets for creation of the Virtual Physiological Human, a model of a human body. It also mentions usage of morphing technique for interpolation of new data. This thesis focuses on the transformations tied with these operations and tries to find an automatic solution which does not need user set up parameters. The deformation filter for surface models of muscles in musculoskeletal model of human body developed in the previous work was chosen as testing application. It has difficulties with damaged input meshes, especially those containing non-manifold edges and vertices. Therefore, the goal is an automatic detection and removal of such artifacts, and the combination of several such inputs into one finer mesh surface gained using a multi-morphing method. To make this possible, approaches for mutual registration of input meshes are analysed, a suitable parametric domain is searched for and appropriate way of final interpolation is chosen. A solution for making such actions in fully automatic manner for general damaged input meshes with similar shape specified by underlying real-world object, but with various initial position and unknown number of topological artifacts consisting of holes and isolated components on top of previously mentioned non-manifold edges and vertices, is then suggested. The resulting method is then implemented and both partial steps and complete design is tested in various experiments. The results are discussed and conclusion is stated at the end of this thesis.