Made available in DSpace on 2014-10-09T12:36:19Z (GMT). No. of bitstreams: 0 / Made available in DSpace on 2014-10-09T13:59:16Z (GMT). No. of bitstreams: 1 12908.pdf: 2121499 bytes, checksum: c2d2abf4d18418855c9e93a43337b8ad (MD5) / Dissertacao (Mestrado) / IEA/D / Escola Politecnica, Universidade de Sao Paulo - POLI/USP

Statistical learning offers a trove of opportunities for problems where a large amount of data is available but falls short when data are limited. For example, in medicine, statistical learning has been used to outperform dermatologists in diagnosing melanoma visually from millions of photos of skin lesions. However, many other medical applications of this kind of learning are made impossible due to the lack of sufficient learning data, for example, performing similar diagnosis of soft tissue tumors within the body based on radiological imagery of blood vessel development. A key challenge underlying this situation is that many statistical learning approaches utilize unstructured data representations such as strings of text or raw images, that don’t intrinsically incorporate structural information. Shape grammar is a way of using visual rules to define the underlying structure of geometric data, pioneered by the design community. Shape grammar rules are replacement rules in which the left side of the rule is a search pattern and the right side is a replacement pattern which can replace the left side where it is found. Traditionally shape grammars have been assembled by hand through observation, making it slow to use them and limiting their use with complex data. This work introduces a way to automate the generation of shape grammars and a technique to use grammars for classification in situations with limited data. A method for automatically inducing grammars from graph based data using a simple recursive algorithm, providing non-probabilistic rulesets, is introduced. The algorithm uses iterative data segmentation to establish multi scale shape rules, and can do so with a single dataset. Additionally, this automatic grammar induction algorithm has been extended to apply to high dimensional data in a nonvisual domain, for example, graphs like social networks. We validated our method by comparing our results to grammars made of historic buildings and products and found it performed comparably grammars made by humans. The induction method was extended by introducing a classification approach based on mapping grammar rule occurrences to dimensions in a high dimensional vector space. With this representation data samples can be analyzed and quickly classified, without the need for data intensive statistical learning. We validated this method by performing sensitivity tests on key graph augmentations and found that our method was comparably sensitive and significantly faster at learning than related existing methods at detecting graph differences across cases. The automated grammar technique and the grammar based classification technique were used together to classify magnetic resonance imaging (MRI) of the brain of 17 individuals and showed that our methods could detect a variety of vasculature borne condition indicators with short and long-term health implications. Through this study we demonstrate that automated grammar based representations can be used for efficient classification of anomalies in abstract domains such as design and biological tissue analysis.

Problems related to apparel fit stem from a variety of factors, such as an outdated anthropometric database from which sizing systems could be developed, lack of and/or inadequate classified body forms, non-standardised communication of sizing and fit and non-standardised fit quality management, amongst the clothing industries (Chun-Yoon&Jasper, 1996; Holzman, 1996; Winks, 1997; Desmarteau, 2000; Anderson, Brannon, Ulrich, Presley; Woronka; Grasso&Stevenson, 2001; Ashdown, 2003; Simmons&Istook, 2004). Anthropometric data in Kenya was taken in 1975 and the measurements were obtained from girls and women of Kenya’s learning institutions and organisations (KEBS, 2001). The source of the original data from which the size tables were derived is obscure, to authenticate the quality of the techniques and instruments used for the data collection. Apparently, there is no known research that has been carried out on clothing anthropometry, sizing (body measurements) and fit (body shape) for women. In the absence of representative sizing systems, wrong styles and sizes based on estimates and not on the actual sizes and body shapes of women consumers in Kenya, contribute to fit problems. Consumers’ lack of knowledge about size (body measurements) and fit (body shape) issues also contributes to the disillusionment, confusion and inappropriate apparel selection. Consumers’ fit preferences contribute to fit problems, if the available styles do not consider consumers’ body shapes, and even further, if the consumers are uninformed about their shapes and how to dress accordingly. The aim of this research was therefore to identify and describe distinctive female body shapes of career women in Kenya using body dimensions and photographs, to describe the differences between the emerging distinctive body shapes (measurements and photographs) and the Western distinctive shapes, and to finally describe and analyse implications for the fit of apparel associated with the emerging distinctive body shapes of Kenya’s career women. It also intended to assess and describe career women’s self-perceived fit issues with the ready-made apparel in Kenya, to determine and describe Kenyan career women’s knowledge about the communication of size (key body dimensions) and fit (body shapes), and also to determine and describe career women’s fit preferences for differently fitted apparel items in Kenya. This research is descriptive as an attempt is made to describe and understand body shape(s) and tendencies in consumers’ behaviour regarding fit issues. It is exploratory as it aims to obtain insight into a relatively new area of study, namely identification of the most prevalent (distinctive) body shape of Kenya’s career women, consumers’ perceived size and fit issues, their knowledge about size and fit, and their fit preferences. Various theories were consulted and adapted in this study, while practical training in anthropometry and photography was undertaken to ensure that measurements and photographs were taken accurately and reliably. Traditional anthropometric-related theories and standards of obtaining body measurements were consulted and applied. Photography rules were set and observed while photographing the women. Phase one of the study focused on the variables in the body characteristics thought to be appropriate for identifying and describing distinctive female body shapes. Phase two applied the quantitative research that focused on the variables obtained from fit problems with apparel, the communication of size and fit, and fit preferences. A structured questionnaire was used to get the broader picture of the respondents’ perceived fit problems, their knowledge about the communication of size and fit, as well as their fit preferences for differently fitted apparel items. The questionnaire measured specific dimensions of fit problems with apparel, the communication of size and fit, as well as fit preferences. The body dimensions that were recorded, body evaluations, and the responses to the questionnaire were coded, captured and analysed. It is apparent from this study that the most dominant body shape is a curvy rectangular shape that differs not only from the ideal (hourglass) body shape, but also from the Western (USA) prevalent straight rectangular shape. The fit problems such as tight hips, crotch, bust and stomach experienced by Kenya’s career women are therefore inevitable, as confirmed by the dissatisfaction with the unavailability of appropriate styles for their sizes and shapes. It is clear that most Kenyan female consumers are familiar with the non-informative lettered and numbered size labels, but unfamiliar with size label terms that represent established body types. They understand neither the meanings of various size and fit descriptions, nor their own key body dimensions; this leads to confusion as to where the cause of their problems lies. Apparently most Kenyan career women consumers prefer fitted and semi-fitted skirts and jackets. In the absence of a distinctive body shape in Kenya, it is possible that the available styles do not cater for their curvy rectangular body shape; hence, they experience fit problems. Consumers’ lack of knowledge about body shape may also lead to inappropriate fit preferences that do not take into account their distinctive body shape and its critical fit points. This study makes certain recommendations to the ready-made apparel industry in Kenya and foreign companies that export their apparel items to Kenya, government agencies such as the Kenya Bureau of Standards, and to consumer-oriented organisations. The results contribute to the body of knowledge regarding the theory of apparel size and fit, Ashdown’s sizing systems theory, research methodology theory, and consumer education theory. / Thesis (PhD)--University of Pretoria, 2009. / Consumer Science / unrestricted

This thesis is concerned with the physical principles, methodology and applications of neutron diffraction in the measurement of residual stress. Work on three main areas is presented. 1) Carbon steels 2) Data and Peak Broadening analysis and 3) Single lap glue shear joints. The Carbon steels section shows the drastic effect of the content of carbon on the measured stress. This is an aspect which has been somewhat neglected in the past. The carbon is in the form of cementite, which is a hard compound and causes the carbon steel to act like a composite material, the ferrite acting as a soft matrix and the cementite as a reinforcement. The consequence of this is that the two components develop high microstresses with plastic deformation. This is clearly illustrated in the work of [Bon 97] where values of approx. 460 MPa in the residual stress in the ferrite are balanced by negative residual stresses of 2300 MPa in cementite yielding an overall macro residual stress of zero. In this work it has been shown that even knowledge of the cementite and ferrite residual stresses and fractions may not be sufficient to accurately calculate the macro stress since the ferrite unloading curve is non linear. The use of a single valued constant modulus to convert from strain to stress is hence not valid. Peak shape analysis enables dislocation density and cell size estimates to be made. The thesis examines several methods of data weighting and deconvolution in order to asses the best means of extracting this information from standard residual stress data. Care should be taken for the peaks with very low backgrounds when finding the Gaussian and Lorentzian components. A weighting that avoids the strong bias of zero and I counts in the detector channels should be used e.g. W = I / ( 10 + Y). Lorentzian and Gaussian components can be successfully extracted from asymmetrical peaks (of peaks that broaden symmetrically), using deconvolution method 1, although the data should be of good quality. Reproducibility has been shown in the Gaussian, Lorentzian and FWHM for different instruments at different institutes. This is extremely important for the use of these values for peak broadening analysis and for estimation of the plastic deformation within a sample. The neutron diffraction technique has been used to investigate the longitudinal stresses in the adherend produced as a result of cure and due to the application of a tensile load in a single lap shear joint. The results throw doubt on widely used finite element predictions.

Shape memory alloys (SMAs) exhibit unique combination of structural and functional properties and hence have a variety of current and potential applications. The mechanical behaviour of SMAs, in particular the influence of processing on the microstructure, which in turn influences the performance of the alloy, mechanical properties at the nano-scale, and under cyclic loading conditions, are of great current interest. In this thesis, specific issues within each of these broad areas are examined with a view to suggest further optimize/characterize SMAs. They are the following: (a) For thermo-mechanical secondary processing of SMAs, can we identify the optimum combination of temperature- strain rate window that yields a desirable microstructure? (b) How can indentation be used to obtain information about functional properties of shape memory alloys so as to complement traditional methods? (c) How can the information obtained from indentation be utilized for the identification of the alloy composition that yields a high temperature SMA through the combinatorial diffusion couple approach? Towards achieving the first objective, we study the hot deformation behavior of a cast NiTi alloy with a view of controlling the final microstructure. The “processing maps” approach is used to identify the optimum combination of temperature and strain rate for the thermomechanical processing of a SMA system commonly used in actuators applications (NiTiCu). Uniaxial compressions experiments are conducted in the temperature range of 800- 1050 °C and at strain rate range of 10-3 and 102 s-1. 2-D power dissipation efficiency and instability maps are generated and various deformation mechanisms, which operate in different temperature–strain rate regimes, are identified with the aid of these maps. Complementary microstructural analysis of specimens (post deformation) is performed with the help of electron backscattered diffraction (EBSD) analysis to arrive at a processing route which produces stress free grains. A safe window suitable for industrial processing of this alloy which leads to grain refinement and strain-free grains (as calculated by various methods of misorientation analysis representation) is suggested. Regions of the instability (characterized by the same analysis) result in strained microstructure, which in turn can affect the performance of the SMA in a detrimental manner. Next, to extract useful information from indentation responses, microindentation experiments at a range of temperatures (as the shape memory transformation is in progress) are conducted underneath the Vickers indenter. SME was observed to cause a change in the calculated recovery ratios at temperatures above As. Spherical indentation of austenite and martensite show different characteristics in elastic and elasto- plastic regimes but are similar in the plastic regime. NanoECR experiments are also conducted under a spheroconical indenter at room temperature, where the resistance measured is observed to increase during the unloading of room temperature austenite SMA. This is a signature of the reverse transformation back to austenite during the withdrawal of the indenter. Lastly, recovery ratios are monitored in the case of a NiTiPd diffusion couple before and after heat treatment at different temperature intervals using non- contact optical profilometry. The recovery ratio approach is successfully used to determine the useful temperature and %Pd range for a potential NiTiPd high temperature SMA. The method makes high throughput identification of high temperature shape memory alloys possible due to promising alloy compositions being identified at an early stage.

The segmentation of medical images is a difficult task due to the inhomogeneous intensity variations that occurs during digital image acquisition, the complicated shape of the object, and the medical expert’s lack of semantic knowledge. Automated segmentation algorithms work well for some medical images, but no algorithm has been general enough to work for all medical images. In practice, most of the time the segmentation results are corrected by the experts before the actual use. In this work, we are motivated to determine how to make use of manually segmented data in automatic segmentation. The key idea is to transfer the ground truth segmentation from the database of train images to a given test image. The ground truth segmentation of MR images is done by experts. The process includes a hierarchical image decomposition approach that performs the shape matching of test images at several levels, starting with the image as a whole (i.e. level 0) and then going through a pyramid decomposition (i.e. level 1, level 2, etc.) with the database of the train images and the given test image. The goal of pyramid decomposition is to find the section of the training image that best matches a section of the test image of a different level. After that, a re-composition approach is taken to place the best matched sections of the training image to the original test image space. Finally, the ground truth segmentation is transferred from the best training images to their corresponding location in the test image. We have tested our method on a hip joint MR image database and the experiment shows successful results on level 0, level 1 and level 2 re-compositions. Results improve with deeper level decompositions, which supports our hypotheses.

Proteins and the formation of large protein complexes are essential parts of living organisms. Proteins are present in all aspects of life processes, performing a multitude of various functions ranging from being structural components of cells, to facilitating the passage of certain molecules between various regions of cells. The 'protein docking problem' refers to the computational method of predicting the appropriate matching pair of a protein (receptor) with respect to another protein (ligand), when attempting to bind to one another to form a stable complex. Research shows that matching the three-dimensional (3D) geometric structures of candidate proteins plays a key role in determining a so-called docking pair, which is one of the key aspects of the Computer Aided Drug Design process. However, the active sites which are responsible for binding do not always present a rigid-body shape matching problem. Rather, they may undergo sufficient deformation when docking occurs, which complicates the problem of finding a match. To address this issue, we present an isometry-invariant and topologically robust partial shape matching method for finding complementary protein binding sites, which we call the ProtoDock algorithm. The ProtoDock algorithm comes in two variations. The first version performs a partial shape complementarity matching by initially segmenting the underlying protein object mesh into smaller portions using a spectral mesh segmentation approach. The Heat Kernel Signature (HKS), the underlying basis of our shape descriptor, is subsequently computed for the obtained segments. A final descriptor vector is constructed from the Heat Kernel Signatures and used as the basis for the segment matching. The three different descriptor methods employed are, the accepted Bag of Features (BoF) technique, and our two novel approaches, Closest Medoid Set (CMS) and Medoid Set Average (MSA). The second variation of our ProtoDock algorithm aims to perform the partial matching by utilizing the pointwise HKS descriptors. The use of the pointwise HKS is mainly motivated by the suggestion that, at adequate times, the Heat Kernel Signature of a point on a surface sufficiently describes its neighbourhood. Hence, the HKS of a point may serve as the representative descriptor of its given region of which it forms a part. We propose three (3) sampling methods---Uniform, Random, and Segment-based Random sampling---for selecting these points for the partial matching. Random and Segment-based Random sampling both prove superior to the Uniform sampling method. Our experimental results, run against the Protein-Protein Benchmark 4.0, demonstrate the viability of our approach, in that, it successfully returns known binding segments for known pairing proteins. Furthermore, our ProtoDock-1 algorithm still still yields good results for low resolution protein meshes. This results in even faster processing and matching times with sufficiently reduced computational requirements when obtaining the HKS.

Le titane, comme ses alliages, présente des caractéristiques remarquables qui peuvent être modulées du fait des nombreuses microstructures qu’il est possible d’obtenir. Grâce à cette grande variété, le titane et ses alliages possèdent un grand nombre de propriétés. Parmi les plus intéressantes, on peut citer leur résistance à la corrosion, leur biocompatibilité, mais aussi leurs excellentes propriétés mécaniques (résistance, ductilité, ténacité, fluage…). Pour toutes ces raisons, l’attrait pour les alliages de titane n’a cessé de croître dans de nombreux secteurs. En effet ils sont maintenant largement utilisés dans les industries aéronautique et chimique, mais aussi l’architecture, le naval, l’industrie automobile, le sport ou encore la médecine. La nanoindentation est utilisée couramment de nos jours pour déterminer les propriétés mécaniques locales des matériaux. Elle permet notamment de caractériser des alliages métalliques possédant une microstructure polycrystalline. La taille de l’indenteur en nanoindentation étant faible (de quelques micromètres à quelques dizaines de micromètres), cette technique est idéale pour caractériser les propriétés mécaniques de surface des différents grains d’un matériau. Elle permet notamment de mesurer simultanément la dureté et le module d’élasticité. Si les essais de nanoindentation sont associés à un banc motorisé X-Y, une matrice étendue d’indents peut être réalisée avec un pas de quelques micromètres. Grâce à cette technique et dans le cadre de ce travail de thèse, nous avons réalisé dans un premier temps des cartographies de dureté et de module d’élasticité (HIPF et EIPF). Dans un second temps, nous avons évalué des propriétés non-conventionnelles d’alliages de titane, telles que l’effet mémoire de forme et la superélasticité. Dans la première partie de l’étude, la nanoindentation a été corrélée à l’EBSD (diffraction des électrons rétro-diffusés) afin d’identifier la relation entre l’orientation cristallographique d’un grain et ses propriétés mécaniques. L’étude a été menée sur les alliages de composition Ti-30Nb et Ti-27Nb (%at) de structure cubique centrée (phase ), et sur le titane de pureté commerciale T40, de structure hexagonale compacte (phase ). Dans la seconde partie de l’étude, la nanoindentation a été utilisée pour mesurer l’effet mémoire de forme (SM) et la superélasticité (SE) de différents alliages de titane à travers une large gamme de profondeur d’indentation. La mesure de ces propriétés non-conventionnelles a été réalisée à partir de l’étude des courbes charge-déplacement obtenues pour chaque essai d’indentation. L’amplitude de l’effet SE et SM a été caractérisée par des ratios de hauteur et de travail déterminés par l’étude des courbes de nanoindentation ainsi que des profils AFM réalisés au microscope à force atomique. / Titanium and titanium alloys presents remarkable characteristics which can be modulated due to the many different microstructures that is possible to obtain. Thanks to this huge variety, titanium and its alloys can exhibit many properties. Among the most interesting, there may be mentioned their corrosion resistance, biocompatibility, but also their excellent mechanical properties (strength, ductility, toughness, creep…). For all these reasons, interest for of titanium alloys has been growing in many areas. Indeed they are now widely used in the aerospace and chemical industries, but also in architecture, naval, automotive, sports or medicine. Nanoindentation is commonly used nowadays to determine local mechanical properties of materials. For example, this technique allows the characterization of metallic alloys having a polycrystalline microstructure. The size of the indenter in nanoindentation being small (few microns to few tens microns), and consequently this technique is ideal for characterizing the surface mechanical properties of different grains of a material. It allows simultaneous measurement of the hardness and the elastic modulus. If nanoindentation tests are associated with a XY motorized test bed, a wide array of indents can be achieved with a step of few micrometers. Thanks to this technique and as part of this thesis, we have realized at first hardness and elastic modulus mapping (HIPF and EIPF). In a second time, we have evaluated unconventional properties of titanium alloys, such as shape memory effect and superelasticity. In the first part of the study, nanoindentation was correlated with EBSD (Electron backscattered diffraction) to identify the relationship between the crystallographic orientation of a grain and its mechanical properties. The study was conducted on the Ti-30Nb and Ti-27Nb (at.%) alloy compositions having a bodycentered cubic structure ( phase), and the commercially pure titanium (CP-Ti) having a hexagonal close packed structure ( phase). In the second part of the study, nanoindentation was used to measure the shape memory effect (SM) and the superelasticity (SE) of various titanium alloys through a range of indentation depth. The measurement of these unconventional properties was performed from the study of load-displacement curves for each indentation test. The magnitude of the SE and SM effect was characterized by depth and work ratios determined from the study of nanoindentation curves and AFM profiles.

Geometric morphometrics, the science about the study of shape, has developed much in the last twenty years. In this thesis I first study the reliability of the phylogenies built using geometric morphometrics. The effect of different evolutionary models, branch-length combinations, dimensionality and degrees of integration is explored using computer simulations. Unfortunately in the most common situations (presence of stabilizing selection, short distance between internal nodes and presence of integration) the reliability of the phylogenies is very low. Different empirical studies are analysed to estimate the degree of evolutionary integration usually found in nature. This gives an idea about how powerful the effect of integration is over the reliability of the phylogenies in empirical studies. Evolutionary integration is studied looking at the decrease of variance in the principal components of the tangent shape space using the independent contrasts of shape. The results suggest that empirical data usually show strong degrees of integration in most of the organisms and structures analysed. These are bad news, since strong degree of integration has devastating effects over the phylogenetic reliability, as suggested by our simulations. However, we also propose the existence of other theoretical situations in which strong integration may not translate into convergence between species, like perpendicular orientation of the integration patterns or big total variance relative to the distance between species in the shape space. Finally, geometric morphometrics is applied to the study of the evolution of shape in proteins. There are reasons to think that, because of their modular nature and huge dimensionality, proteins may show different patterns of evolutionary integration. Unfortunately, proteins also show strong functional demands, which influence their evolution and that cause strong integration patterns. Integration is then confirmed as a widespread property in the evolution of shape, which causes poor phylogenetic estimates.

We present a method to determine the precise shape of a dynamic object from video. This problem is fundamental to computer vision, and has a number of applications, for example, 3D video/cinema post-production, activity recognition and augmented reality. Current tracking algorithms that determine precise shape can be roughly divided into two categories: 1) Global statistics partitioning methods, where the shape of the object is determined by discriminating global image statistics, and 2) Joint shape and appearance matching methods, where a template of the object from the previous frame is matched to the next image. The former is limited in cases of complex object appearance and cluttered background, where global statistics cannot distinguish between the object and background. The latter is able to cope with complex appearance and a cluttered background, but is limited in cases of camera viewpoint change and object articulation, which induce self-occlusions and self-disocclusions of the object of interest. The purpose of this thesis is to model self-occlusion/disocclusion phenomena in a joint shape and appearance tracking framework. We derive a non-linear dynamic model of the object shape and appearance taking into account occlusion phenomena, which is then used to infer self-occlusions/disocclusions, shape and appearance of the object in a variational optimization framework. To ensure robustness to other unmodeled phenomena that are present in real-video sequences, the Kalman filter is used for appearance updating. Experiments show that our method, which incorporates the modeling of self-occlusion/disocclusion, increases the accuracy of shape estimation in situations of viewpoint change and articulation, and out-performs current state-of-the-art methods for shape tracking.