The tools used by wood scientists today to understand structural differences in the wood cell are manual and therefore uncertain and time consuming. This implicates that they are not well suited for statistical analysis. One aim of this master thesis project has been to explore the possibility of letting image analyze methods speed up and increase reliability of wood cell characteristics evaluation.
One usable application of a faster wood cell structure evaluating tool is in the production of paper pulp. Wood cells consist of a cell wall, which is divided into three layers, the S1- S2- and S3 layer. Tree scientists believe that the first layer plays a determining role when wood fibers (cells) are made ready for paper pulp production. The thicker it is, the more energy is needed to be consumed in order to make the fibers ready for paper pulp production. Is it possible with image analyze methods to distinguish a width difference between different species, thus the best suited raw material is used in paper pulp production?
One problem with regular low-pass noise filtering is that also the sharpness of the image details, e.g. the fibre edges, are reduced or blurred. This effect of the fibre edges might be negative for the following measurements. In this Master's Thesis report so called anisotropic filtering methods were tested to suppress the noise and conserve the fibre edges. Anisotropic filtering methods locally adapts the smoothing of the image. Hence, well tuned they can conserve edges while they suppress the noise. Four different anisotropic filtering methods were tested and compared, Bilateral filtering, SUSAN smoothing, anisotropic diffusion and Adaptive filtering with structure tensors.
It is shown that all four methods, applied on the paper images, can be tuned to give a result with suppressed noise and conserved edges. Also the pros and cons of the different methods are discussed.
The aim of this thesis is first to try to evaluate and second to further develop the usability of, a new method for measuring the distribution of filler content in paper. Digital image analysis, based on images from induced fluorescence photography on paper sheets, split by a lamination technique developed at STFI-Packforsk, is to be used to accomplish this.
A review of the literature on the topic is done, focus is put on the spatial distribution of filler. Later a method for splitting of paper samples is introduced, followed by a description of the appropriate ways of acquiring images. The proper set-up for the image acquisition process is determined, and a suggestion of suitable digital image analysis tools are designed and evaluated.
After attempting to acquire suitable photographs, image segmentation is tried in order to extract images of filler material clusters. A combination of image filters and morphology transforms is evaluated as a method to extract this information. A 3D visualisation of the filler distribution follows. Finally a check for validity of the photographies is done using a optical sample mass measurement and a statistical approach. During this project, MATLAB-code is developed to perform these image analysis operations.
Accurate photographies describing the filler distribution could not be obtained. Therefore this method could not be tested thoroughly. The recommendation is nevertheless not to develop this method further due to the unreliability discovered. Interesting - and possibly usable - discoveries might have been made in the process of coming to this conclusion. If so, the effects of these remain to be seen.
In the northern parts of Sweden the water in dams freeze during winter. In the spring, when the ice breaks up, the water level gauge gets damaged due to ice floes. One consequence of this phenomenon is that the engine operators of the power plant repeatedly have to replace the broken gauge with new ones. Another consequence is that the quality of the measurements of the water level is low while the gauge is in bad condition.
To explore the possibilities of replacing the gauge with a camera surveillance system both the technincal and sociotechnical conditions were investigated. From a technical point of view there was no problem with the project and a program that calculates the water level in the dam was created in MATLAB. From a sociotechnical point of view on the other hand, the project could not be carried out. Due to the water decree, the water level gauges could not be replaced. Therefore, no camera surveillance system was implemented.
The goal of this master thesis project is to produce a low budget 3D pointer with two standard webcams connected to a normal PC.
By filtering out visible light and using an IR diode as pointer the position in the images are determined with thresholding and connected component labeling followed by a camera calibration. A 3D position is calculated from the two 2D images with a home made geometric solution.
The result is a VC++ program that delivers new 3D coordinates in a user defined coordinate system every 25 ms. Despite the low resolution of the webcams a relatively high precision is achieved.