Shape parameterisation for identification and characterisation of polymer surface features

点击翻译按钮获取中文摘要

While we may perceive the surfaces of commonplace objects as smooth or flat, when investigated at a microscopic level there is always quite a complicated landscape. There will be hills ('asperities') and valleys, and the sizes and shapes of these geometrical objects may be important for a variety of reasons. In micromoulding of polymers, the surface characteristics may be the crucial feature that controls the function of a product, as is the case (for instance) with a microlens consisting of an array of very small lenses. There is therefore a need to control the surface of the product, and to do this automatically we need to characterise the surface mathematically. Within the polymer IRC laboratories at Bradford, we have instruments that can measure surfaces and produce digital images of them. Each image consists of many thousands of points in three-dimensional space, and is cumbersome to manipulate. In the case of a long sequence of images, such as is proposed in the monitoring of a mass production process, there will be many of these image files that will occupy a prohibitively large amount of computer storage. To handle these files routinely as part of a control and optimisation scheme for a process, a means of characterising the surfaces using a small number of parameters is required.A method of characterisation of three-dimensional surfaces, the PDE method, is under development in the School of Informatics. Solutions of partial differential equations are used to form surfaces in three-dimensional space that approximate. A highly diverse set of shapes can be generated that are controlled by the small number of parameters that are associated with the PDE solutions, and these can be fitted to real three-dimensional objects. The most recent application has been in fitting to scanned human faces for recognition purposes. We propose to use this method to characterise polymer surfaces. This will be a highly significance advance over the conventional means of characterising surfaces via the surface roughness, which uses a single number, the mean asperity radius. Rich and realistic approximations of surfaces will be made possible, and ideal surfaces created. Then, robotically controlled measurement and real-time computing will be used to characterise the produced surfaces, compare them with the ideal and thus optimise the micromoulding process.