Bag-of-Features Approach to Unsupervised Visual Object Categorisation

The large and growing number of digital images is making manual image search laborious. Only a fraction of the images contain metadata that can be used to search for a particular type of image. Thus, the main research question of this thesis is whether it is possible to learn visual object categories directly from images. Computers process images as long lists of pixels that do not have a clear connection to high-level semantics which could be used in the image search. There are various methods introduced in the literature to extract low-level image features and also approaches to connect these low-level features with high-level semantics. One of these approaches is called Bag-of-Features which is studied in the thesis. In the Bag-of-Features approach, the images are described using a visual codebook. The codebook is built from the descriptions of the image patches using clustering. The images are described by matching descriptions of image patches with the visual codebook and computing the number of matches for each code. In this thesis, unsupervised visual object categorisation using the Bag-of-Features approach is studied. The goal is to find groups of similar images, e.g., images that contain an object from the same category. The standard Bag-of-Features approach is improved by using spatial information and visual saliency. It was found that the performance of the visual object categorisation can be improved by using spatial information of local features to verify the matches. However, this process is computationally heavy, and thus, the number of images must be limited in the spatial matching, for example, by using the Bag-of-Features method as in this study. Different approaches for saliency detection are studied and a new method based on the Hessian-Affine local feature detector is proposed. The new method achieves comparable results with current state-of-the-art. The visual object categorisation performance was improved by using foreground segmentation based on saliency information, especially when the background could be considered as clutter.

Dynamic Analysis Model of Spherical Roller Bearings with Defects

Rolling element bearings are essential components of rotating machinery. The spherical roller bearing (SRB) is one variant seeing increasing use, because it is self-aligning and can support high loads. It is becoming increasingly important to understand how the SRB responds dynamically under a variety of conditions. This doctoral dissertation introduces a computationally efficient, three-degree-of-freedom, SRB model that was developed to predict the transient dynamic behaviors of a rotor-SRB system. In the model, bearing forces and deflections were calculated as a function of contact deformation and bearing geometry parameters according to nonlinear Hertzian contact theory. The results reveal how some of the more important parameters; such as diametral clearance, the number of rollers, and osculation number; influence ultimate bearing performance. Distributed defects, such as the waviness of the inner and outer ring, and localized defects, such as inner and outer ring defects, are taken into consideration in the proposed model. Simulation results were verified with results obtained by applying the formula for the spherical roller bearing radial deflection and the commercial bearing analysis software. Following model verification, a numerical simulation was carried out successfully for a full rotor-bearing system to demonstrate the application of this newly developed SRB model in a typical real world analysis. Accuracy of the model was verified by comparing measured to predicted behaviors for equivalent systems.

Development of a Filter Aid Proportioning, Mixing, and Feeding System for a Polishing Filter

The purpose of this work was to study the characteristics of the most commonly used filter aid materials and their influences on the design of proportioning, mixing, and feeding system for polishing filter family. Based on the literature survey and hands-on experience a system was designed with defined equipment and capital and operating costs. The system was designed to serve precoating and bodyfeeding applications and is easily extended to be used in multiple filter processes. Also a test procedure was carried out where influences of flux and filter cloths to accumulated cake were studied. Filter aid is needed in challenging conditions to improve filtration efficiency and cleaning, and thus extend the operating life of the filter media. Filter aid preparation and feeding system was designed for the use of two different filter aids; precoat and bodyfeed. Precoating is used before the filtration step initiates. If the solids in the filterable solution have a tendency to clog the filter bag easily, precoat is used on the filter bag to obtain better filtration efficiency and quality. Diatomite or perlite is usually used as a precoating substance. The intention is to create a uniform cake to the overall surface of the filter cloth, with predetermined thickness, 2 – 5 mm. This ensures that the clogging of the filter cloth is reduced and the filtration efficiency is increased. Bodyfeed is used if the solids in the filterable solution have a tendency to form a sticky impermeable filter cake. The cake properties are enhanced by maintaining the permeability of the accumulating cake by using the filter aid substance as bodyfeed during the filtration process.

Conformal invariants

This master thesis investigates the moduli of families of curves and the capacities of the Gr¨otzsch and Teichm¨uller rings, which are applied in the main parts of this master thesis. The extremal properties of these rings are discussed in connection with the spherical symmetrization. Applications are given to the study of distortion of quasiconformal maps in the euclidean n-dimensional space.