Abstract

Paper is a material that suits many different purposes, and is therefore used extensively. To be able to optimize the manufacturing parameters during the production in a paper mill, it is of great importance to know as much as possible about the material. Paper exists in many different qualities, and different paper grades can have very different properties. There already exists many techniques to measure paper properties, but nearly all are based on computerized image analysis of two spatial dimensions, or not based on image analysis at all. Recently, the processing power of computers has reached a level which makes it possible to analyze huge data sets, by 3D computerized image analysis. Most grades of paper consists of fibres connected in a 3D network, together with fillers and surface coatings. We have focused on the 3D network of fibres, to provide techniques to look into, and perform calculations on this network. Advantages of using image analysis is that if we can transform samples of paper into an accurate digital format, a computer can then be programmed to analyze some of the properties not easily accessible by other means. Both mechanical strength and opacity depend on the geometry of the fibre network, so the digital model could be useful for further analysis of these properties. Disadvantages are that the conversion from physical to digital form will introduce artefacts and this conversion is often time-consuming. Our work has been to develop a method to construct 3D digital models of paper and to develop some useful measurements. The digital models consist of efficient representations that enable our measurements and also provide a good base for further development. The measurements consist of a detailed fibre analysis, a few fibre network measures, and experiments to measure the fibre pore network. Both 2D and 3D based fibre segmentation algorithms have been implemented. The image capturing process is crucial, but we still believe that the core problem is to improve the fibre segmentation. Especially, a good reconstruction of the network requires a high percentage of the fibres to be found, something that need further work to be accomplished. This thesis will emphasize the image analysis part of the problem, even if there are as much to be said about paper as a material.

Keywords

volume images; image registration; filter design; fibre network; individual fibre segmentation; fibre properties estimation; visualization

Published in

Acta Universitatis Agriculturae Sueciae. Silvestria
2002, number: 254
ISBN: 91-576-6338-6
Publisher: Swedish University of Agricultural Sciences

SLU Authors

• Aronsson, Mattias

  • Centre for Image Analysis, Swedish University of Agricultural Sciences
    

UKÄ Subject classification

Paper, Pulp and Fiber Technology


Permanent link to this page (URI)

https://res.slu.se/id/publ/107644