Creation of a base mesh from captured data

This project will investigate ways in which we can produce a good control mesh from captured data that can be used for further manipulation of the model. The goal is to extract good coarse subdivision meshes from densely- scanned, real world models. During a design process computerised models are produced in two different ways. The designer either creates the model, ab initio, using professional modelling software or makes the model in clay and the model is then scanned in, typically with a laser scanner. In the ab initio case, we need powerful general tools that give designers sufficient freedom and flexibility to get the ideas from inside their heads into the computer. In the clay modelling case, we need tools that can create a coarse, editable control mesh, with fine detail provided in for example a multi- resolution manner. We hypothesise it is possible to create a good coarse subdivision mesh from densely captured data, with fine detail represented using a multi-resolution approach. Enough research has been done on 3D capture and we are able to produce, in a post-process, clean 3D meshes from scanned data with high accuracy. We will therefore start by looking at ways to produce a good coarse representation from post-processed scan data. For the designer to be able to manipulate the data sensibly and in a controlled manner the number of data points have to be reduced. A coarse mesh allows the efficient later manipulation of the model online. This will allow 3D designers to continue to use their traditional clay modelling methods, while dramatically reducing the time to convert these models to an editable digital form. The fine detail will be added in a multi-resolution fashion, for which subdivision is an obvious candidate. The key point is that we should aim to produce a very coarse mesh compared with previous model-capture methods.

Often, when designing products, a designer starts by making a real world model of their idea, e.g. a clay model. These models are then digitized, for example by using a laser scanner. The point clouds created in the scanning process are typically very dense, they can be visualized on a computer screen but they are too dense for further modification on the computer. For the designer to be able to manipulate the data sensibly and in a controlled manner the number of data points has to be reduced. Also, the model needs to enter into a larger product design pipeline, which requires the model to be converted to a mathematical representation of the shape, i.e. to a standard CAD representation like NURBS or subdivision surfaces. Both surface representations offer the designer a coarse control mesh as a manipulation interface, which is mathematically linked to a smooth limit surface describing the shape. Ideally, the control mesh is very sparse, so that the shape of the smooth limit surface can be controlled by moving just a few points of the control mesh. However, the control mesh also needs to be dense enough to be able to represent intricate features in the surface. In this project we explored ways to ease the conversion from a dense point cloud to a coarse mathematical representation of the shape, so a designer can transition easily from designing in the real world to manipulating his/her design on the computer. We solved this problem in two ways. First, we made it easy to manually create control meshes from scan data by providing a guidance system which enables the designer to directly draw a control mesh onto the scan data. The thus created polyhedron is then automatically converted to a subdivision control mesh. Detailed features can be incorporated by drawing onto the control mesh. Secondly, we developed an automatic conversion algorithm which lets the designer choose how closely details in the surface are represented in the control mesh. The designer can generate very coarse meshes which ignore intricate features in the surface and can then use our software tools to add features to the surface. Or the designer can generate a control mesh at a density which reproduces all the features in the surface. To avoid creating control meshes which have high density over the whole surface, the algorithm increases density only locally, so that rather flat regions are represented with a coarse mesh, while in regions with small detail the control mesh is much denser in order to represent detail information. We also experimented with bringing the design paradigm of the real world and the CAD world closer together by bringing the shape into a virtual world. There it can either treated as a CAD model with a control mesh interface, or as a shape with clay like material properties which respond to interaction like our real world clay model.