Visual Simultaneous Localisation and Mapping (SLAM) is concerned with simultaneously estimating the pose of a camera and a map of the environment from a sequence of images. Traditionally, sparse maps comprising isolated point features have been employed, which facilitate robust localisation but are not well suited to advanced applications. In this thesis, we present map representations that allow a more dense description of the environment. In one approach, planar features are used to represent textured planar surfaces in the scene. This model is applied within a visual SLAM framework based on the Extended Kalman Filter. We presents solutions to several challenges which arise from this approach.

A website with a good e-commerce design will positively influence a customer’s attitude, strengthen the trust of the customers towards the organisation, increase the satisfaction of the customer, draw consumers, and bring forth purchases/repeat purchases from them. The purpose of this study was to determine viable e-commerce opportunities for the Ficksburg Cherry Festival website. The study endeavoured to ascertain how the Ficksburg Cherry Festival could improve its website by identifying e-commerce opportunities for the Ficksburg Cherry Festival website. In this study a non-random self-administered survey approach was used where attendees were intercepted at the venue of the Ficksburg Cherry Festival and the exhibitors in their stalls. The existing Ficksburg Cherry Festival website was analysed for e-commerce features that are present and opportunities for further implementation of e-commerce features were identified with help from the attendees and exhibitors. The findings of this study presented a demographic profile and attendance characteristics of both the attendees and exhibitors, which the management team of the Ficksburg Cherry Festival can use to better market the festival using traditional and online communication for both attendees and exhibitors. The recommendations will assist the management of the festival to improve the website and to progress from a straightforward information-only website to a fully-developed e-commerce site, with positive effects for all the stakeholders. Similar South African festivals could find the research findings and recommendations of value for their own website development as well. / Business Management / M. Com. (Business Management)

The State as the representative of the victims of crime is expected to protect those vulnarable group of people with due regard to the rights of the perpetrators’s of crime. It is imperative that the law of general application which is aimed at protecting victims of crime, be sufficiently effective to protect the victims. The Criminal Procedure Act 51 of 1977 is aimed at assisting the police to conduct pre-trial criminal procedure in order to bring perpetrators of crime to book. Sections 36A, 36B, 36C and 37 (both previous and as amended) of the Criminal Procedure Act including chapter 5A of the South African Police Act, 1995 are explored in this dissertation. This dissertation examines the areas in the Criminal Procedure Act that make it problematic for the police to conduct efficient and effective crime detection through the ascertainment of bodily features of the suspected or accused person. The law in three foreign jurisdictions relating to this topic are investigated and compared in order to make recommendations and suggest possible solutions. / Criminal & Procedural Law / LL.M.

South African indigenous language teaching and learning materials do not provide sufficient information to help additional language learners learn the target languages effectively. While there are institutions that are tasked with developing and sharpening the skills of students in speaking South African indigenous languages, such students hardly, if at all master the art of speaking them eloquently. Students who study these languages in order to converse proficiently with their mother-tongue speakers experience insurmountable difficulties, in spite of various efforts made by the teachers who train them to read books on their own. Passing their examinations does not mean that the students’ ability to communicate with mother-tongue speakers will improve to the extent of eliminating the prevailing misunderstanding between the two groups. The persistence of this problem reveals a discrepancy between the studies of indigenous languages in South Africa and the way of speaking them, whereby important linguistic elements that make communication more authentic are excluded in language materials. This study analyses the use and significance of CIFWs in daily interactions by investigating the two Xhosa CIFWs words wethu and bethu. The overall aim of this study is to explore the use of a corpus in the examination of CIFWs in general, and wethu and bethu in particular. Both a quantitative approach based on the Gothenburg-Unisa spoken corpus and a qualitative approach based on Allwoods’ ACA theoretical framework were used in the analysis and description of the functions and significances of wethu and bethu as communicative and interactive function words. / Linguistics / MA ((Applied Linguistics)

The increased temperatures in the Arctic accelerate the loss of land based ice stored in glaciers. The Greenland Ice Sheet is the largest ice mass in the Northern Hemisphere and holds ~10% of all the freshwater on Earth, equivalent to ~7 metres of global sea level rise. A few decades ago, the mass balance of the Greenland Ice Sheet was poorly known and assumed to have little impact on global sea level rise. The development of regional climate models and remote sensing of the ice sheet during the past decade have revealed a significant mass loss. To monitor how the Greenland Ice Sheet will affect sea levels in the future requires understanding the physical processes that govern its mass balance and movement. In the southeastern and central western regions, mass loss is dominated by the dynamic behaviour of ice streams calving into the ocean. Changes in surface mass balance dominate mass loss from the Greenland Ice Sheet in the central northern, southwestern and northeastern regions. Little is known about what the hydrological system looks like beneath the ice sheet; how well the hydrological system is developed decides the water’s impact on ice movement. In this thesis, I have focused on radar sounding measurements to map the subglacial topography in detail for a land-terminating section of the western Greenland Ice Sheet. This knowledge is a critical prerequisite for any subglacial hydrological modelling. Using the high-resolution ice thickness and bed topography data, I have made the following specific studies: First, I have analysed the geological setting and glaciological history of the region by comparing proglacial and subglacial spectral roughness. Second, I have analysed the subglacial water drainage routing and revealed a potential for subglacial water piracy between adjacent subglacial water catchments with changes in the subglacial water pressure regime. Finally, I have looked in more detail into englacial features that are commonly observed in radar sounding data from western Greenland. In all, the thesis highlights the need not only for accurate high-resolution subglacial digital elevation models, but also for regionally optimised interpolation when conducting detailed hydrological studies of the Greenland Ice Sheet. / De ökade temperaturerna i Arktis påskyndar förlusten av landbaserad is lagrad i glaciärer och permafrost. Grönlands inlandsis är den största ismassan på norra halvklotet och lagrar ca 10% av allt sötvatten på jorden, vilket motsvarar ca 7 meter global havsnivåhöjning. För ett par decennier sedan var inlandsisens massbalans dåligt känd och antogs ha liten inverkan på dagens havsnivåhöjning. Utvecklingen av regionala klimatmodeller och satellitbaserad fjärranalys av inlandsisen har under de senaste decenniet påvisat en betydande massförlust. För att förutse vilken inverkan inlandsisen har på framtida havsnivåhöjningar krävs en förståelse för de fysikaliska processerna som styr dess massbalans och isrörelse. I de sydöstra och centrala västra delarna av inlandsisen domineras massförlusten av dynamiska processer i isströmmar som kalvar ut i havet. Massförlusten i de centrala norra, sydvästra och nordöstra delarna domineras av isytans massbalans. Ytterst lite är känt om hur det hydrologiska systemet ser ut under inlandsisen; hur väl det hydrologiska systemet är utvecklat avgör vattnets påverkan på isrörelsen. I denna doktorsavhandling har jag använt markbaserade radarmätningar för att kartlägga den subglaciala topografin för en del av den västra landbaserade inlandsisen. Denna kunskap är en viktig förutsättning för att kunna modellera den subglaciala hydrologin. Med hjälp av rumsligt högupplöst data över istjockleken och bottentopografin har jag gjort följande specifika studier: Först har jag analyserat de geologiska och glaciologiska förhållandena i regionen genom att jämföra proglacial och subglacial spektralanalys av terrängens ytojämnheter. Sedan har jag analyserat den subglaciala vattenavrinningen och påvisat en potential för att avrinningsområdena kan ändras beroende på vattentryckförhållandena på botten. Slutligen har jag tittat mer i detalj på englaciala radarstrukturer som ofta observerats i radardata från västra Grönland. Sammanfattningsvis belyser avhandlingen behovet av inte bara noggranna rumsligt högupplösta subglaciala digitala höjdmodeller, utan även regionalt optimerad interpolering när detaljerade hydrologiska studier ska utföras på Grönlands inlandsis.

Controls on denudation in the eastern Bolivian Andes are evaluated by synthesis of new and existing denudation estimates from basin-morphometry, stream - powered fluvial incision, landslide mapping, sediment flux, erosion surfaces, thermochronology, foreland basin sediment volumes, and structural restorations. Centered at 17.5 °S, the northeastern Bolivian Andes exhibit high relief, a wet climate, and a narrow fold- thrust belt. In contrast, the southeastern Bolivian Andes have low relief, a semi-arid climate, and a wide fold-thrust belt. Basin -morphometry indicates a northward increase in relief and relative denudation. Stream-power along river profiles shows greater average incision rates in the north by a factor of 2 to 4. In the south, profile knickpoints with high incision rates are controlled by fold-thrust belt structures such as the surface expressions of basement megathrusts, faults, folds, and lithologic boundaries. Landslide and sediment-flux data are controlled by climate, elevation, basin morphology, and size and show a similar trend; short -term denudation-rate averages are greater in the north (1- 9 mm/yr) than the south (0.3-0.4 mm/yr). Long-term denudation-rate estimates including fission track, basin fill, erosion surfaces, and structural restorations also exhibit greater values in the north (0.2-0.8 mm/yr) compared to the south (0.04-0.3 mm/yr). Controls on long-term denudation rates include relief, orographic and global atmospheric circulation patterns of precipitation, climate change, glaciation, and fold-thrust belt geometry and kinematics. The denudation synthesis supports two conclusions: 1) denudation rates have increased towards the present 2) an along-strike disparity in denudation (greater in the north) has existed since at least the Miocene and has increased towards the present. Denudation rates and controls suggest that Bolivian mountain morphology is controlled by both its orientation at mid-latitude, and the feedbacks between uplift, kinematics, orographic effects on precipitation, glaciation, and the increased erosion that accompanies orogenesis.

Le cancer épithélial de l’ovaire (CÉO) est le cancer gynécologique le plus létal. Le CÉO de type séreux, la forme la plus commune avec plus de 50% des cas, est souvent diagnostiqué tardivement et associé à un mauvais pronostic. Le CÉO avancé, surtout traité par chimiothérapie, va devenir chimiorésistant chez la majorité des patientes traitées. Bien que des lignées cellulaires du CÉO aient été dérivées à partir de tumeurs solides et d’ascites de patientes ayant ou non subi une chimiothérapie, aucune des lignées cellulaires du CÉO provenant d’une même patiente avant et après ses traitements de chimiothérapie n’ont été établies précédemment. Notre laboratoire est le premier à développer de telles lignées cellulaires. Nos nouvelles lignées cellulaires sont dérivées de trois patientes différentes (1369, 2295 et 3133) et classées selon leur provenance, soit la tumeur solide (TOV) ou l’ascite (OV). Nous avons donc caractérisé ces nouvelles lignées de cellules pré-chimiothérapie (TOV1369TR, OV2295, TOV3133D et TOV3133G) et post-chimiothérapie (OV1369(2), OV2295(2), TOV2295, OV3133 et OV3133(2)) par diverses approches. Par immunohistochimie et immunobuvardage de type Western, nous avons caractérisé les niveaux d’expression de marqueurs épithéliaux typiques de kératines (KRT7, KRT8, KRT18, KRT19, KRT20) pour confirmer l’origine épithéliale et ovarienne des cellules. Nous avons également analysé le niveau d’expression de HER2 et p53, deux marqueurs importants dans le CÉO. Cependant, il ne semble pas y avoir d’expression différentielle évidente de ces marqueurs entre les lignées pré-chimiothérapie et post-chimiothérapie. Plus encore, nous avons étudié plusieurs caractéristiques tumorigéniques des lignées cellulaires, dont la prolifération cellulaire (par compte cellulaire), la migration cellulaire (par recouvrement de plaie), la capacité à former des sphéroïdes en 3D (par la méthode des gouttelettes inversées), et la formation de tumeurs in vivo dans des souris SCID (xénogreffes sous-cutanées). En général, il ne semble pas y avoir de différences claires entre les cellules pré-chimiothérapie et post-chimiothérapie au niveau du comportement cellulaire, à l’exception du fait qu’aucune des lignées post-chimiothérapie semblent être en mesure de former des structures tridimensionnelles compactes, contrairement à certaines lignées post-chimiothérapie. Nos résultats pourront servir à mieux comprendre les différents mécanismes régissant les tumeurs malignes du CÉO de type séreux et à mieux comprendre la progression de la maladie à travers les différents traitements, ce qui nous permettra d’acquérir des informations essentielles pour mieux évaluer et traiter différentes patientes. / Epithelial ovarian cancer (EOC) is the deadliest of all gynecologic cancers. The serous type of EOC is the most common form of the disease, and it accounts for more than 50% of the cases. It is often diagnosed at advanced stages where its prognosis is poor. Advanced EOC is treated mainly with chemotherapy. However, chemoresistance development eventually impedes the success of the treatments for most patients. Researchers have derived cell lines from EOC from solid tumors or from ascites. So far, there has not been EOC cell lines established from samples taken before and after chemotherapy treatments within the same patient. Our laboratory is thus the first to develop a new and powerful model of pre-chemotherapy and post-chemotherapy cell lines. All cell lines were derived sequentially from 3 different patients (1369, 2295 and 3133), from either solid tumors (TOV) or ascites (OV). We therefore characterized these new pre-chemotherapy cell lines (TOV1369TR, OV2295, TOV3133D and TOV3133G) and post-chemotherapy cell lines (OV1369(2), OV2295(2), TOV2295, OV3133 and OV3133(2)) through several approaches. Using immunohistochemistry and Western blot, we have characterized the level of expression of typical epithelial keratin markers (KRT7, KRT8, KRT18, KRT19, KRT20) to confirm the epithelial and ovarian nature of the cells. We have also analysed the expression level of important EOC markers, such as that of HER2 and p53, and found no clear difference between the pre-chemotherapy and post-chemotherapy EOC cells. Moreover, we have studied various tumorigenic features of the cell lines, such as cell proliferation (by cell count), cell migration (by the wound healing assay), 3D spheroid formation (by the hanging drop method), in vivo tumor formation in SCID mice (subcutaneous xenografts). In general, there were no notable differences between the two categories of cell lines at the cellular level, except that post-chemotherapy cell lines seemed to be unable to form compact 3D structures, contrary to some pre-chemotherapy cell lines. The obtained results would aid in better understanding the different mechanisms that malignant serous EOC tumors undergo and the progression of the disease with respect to the different treatments. Such study would allow us to gain valuable insight into the optimal treatment decisions to take for different EOC patients.

A measure of the quality of Voronoi tessellations resulting from various mesh generators founded on feature-driven models is introduced in this work. A planar tessellation covers an image with polygons of various shapes and sizes. Tessellations have potential utility due to their geometry and the opportunity to derive useful information from them for object recognition, image processing and classification. Problem domains including images are generally feature-endowed, non-random domains. Generators modeled otherwise may easily guarantee quality of meshes but certainly bear no reference to features of the meshed problem domain. They are therefore unsuitable in point pattern identification, characterization and subsequently the study of meshed regions. We therefore found generators on features of the problem domain. This provides a basis for element quality studies and improvement based on quality criteria. The resulting polygonal meshes tessellating an n-dimensional digital image into convex regions are of varying element qualities. Given several types of mesh generating sets, a measure of overall solution quality is introduced to determine their effectiveness. Given a tessellation of general and mixed shapes, this presents a challenge in quality improvement. The Centroidal Voronoi Tessellation (CVT) technique is developed for quality improvement and guarantees of mixed, general-shaped elements and to preserve the validity of the tessellations. Mesh quality indicators and entropies introduced are useful for pattern studies, analysis, recognition and assessing information. Computed features of tessellated spaces are explored for image information content assessment and cell processing to expose detail using information theoretic methods. Tessellated spaces also furnish information on pattern structure and organization through their quality distributions. Mathematical and theoretical results obtained from these spaces help in understanding Voronoi diagrams as well as for their successful applications. Voronoi diagrams expose neighbourhood relations between pattern units. Given this realization, the foundation of near sets is developed for further applications. / February 2017

L’annotation en rôles sémantiques est une tâche qui permet d’attribuer des étiquettes de rôles telles que Agent, Patient, Instrument, Lieu, Destination etc. aux différents participants actants ou circonstants (arguments ou adjoints) d’une lexie prédicative. Cette tâche nécessite des ressources lexicales riches ou des corpus importants contenant des phrases annotées manuellement par des linguistes sur lesquels peuvent s’appuyer certaines approches d’automatisation (statistiques ou apprentissage machine). Les travaux antérieurs dans ce domaine ont porté essentiellement sur la langue anglaise qui dispose de ressources riches, telles que PropBank, VerbNet et FrameNet, qui ont servi à alimenter les systèmes d’annotation automatisés. L’annotation dans d’autres langues, pour lesquelles on ne dispose pas d’un corpus annoté manuellement, repose souvent sur le FrameNet anglais. Une ressource telle que FrameNet de l’anglais est plus que nécessaire pour les systèmes d’annotation automatisé et l’annotation manuelle de milliers de phrases par des linguistes est une tâche fastidieuse et exigeante en temps. Nous avons proposé dans cette thèse un système automatique pour aider les linguistes dans cette tâche qui pourraient alors se limiter à la validation des annotations proposées par le système. Dans notre travail, nous ne considérons que les verbes qui sont plus susceptibles que les noms d’être accompagnés par des actants réalisés dans les phrases. Ces verbes concernent les termes de spécialité d’informatique et d’Internet (ex. accéder, configurer, naviguer, télécharger) dont la structure actancielle est enrichie manuellement par des rôles sémantiques. La structure actancielle des lexies verbales est décrite selon les principes de la Lexicologie Explicative et Combinatoire, LEC de Mel’čuk et fait appel partiellement (en ce qui concerne les rôles sémantiques) à la notion de Frame Element tel que décrit dans la théorie Frame Semantics (FS) de Fillmore. Ces deux théories ont ceci de commun qu’elles mènent toutes les deux à la construction de dictionnaires différents de ceux issus des approches traditionnelles. Les lexies verbales d’informatique et d’Internet qui ont été annotées manuellement dans plusieurs contextes constituent notre corpus spécialisé. Notre système qui attribue automatiquement des rôles sémantiques aux actants est basé sur des règles ou classificateurs entraînés sur plus de 2300 contextes. Nous sommes limités à une liste de rôles restreinte car certains rôles dans notre corpus n’ont pas assez d’exemples annotés manuellement. Dans notre système, nous n’avons traité que les rôles Patient, Agent et Destination dont le nombre d’exemple est supérieur à 300. Nous avons crée une classe que nous avons nommé Autre où nous avons rassemblé les autres rôles dont le nombre d’exemples annotés est inférieur à 100. Nous avons subdivisé la tâche d’annotation en sous-tâches : identifier les participants actants et circonstants et attribuer des rôles sémantiques uniquement aux actants qui contribuent au sens de la lexie verbale. Nous avons soumis les phrases de notre corpus à l’analyseur syntaxique Syntex afin d’extraire les informations syntaxiques qui décrivent les différents participants d’une lexie verbale dans une phrase. Ces informations ont servi de traits (features) dans notre modèle d’apprentissage. Nous avons proposé deux techniques pour l’identification des participants : une technique à base de règles où nous avons extrait une trentaine de règles et une autre technique basée sur l’apprentissage machine. Ces mêmes techniques ont été utilisées pour la tâche de distinguer les actants des circonstants. Nous avons proposé pour la tâche d’attribuer des rôles sémantiques aux actants, une méthode de partitionnement (clustering) semi supervisé des instances que nous avons comparée à la méthode de classification de rôles sémantiques. Nous avons utilisé CHAMÉLÉON, un algorithme hiérarchique ascendant. / Semantic role annotation is a process that aims to assign labels such as Agent, Patient, Instrument, Location, etc. to actants or circumstants (also called arguments or adjuncts) of predicative lexical units. This process often requires the use of rich lexical resources or corpora in which sentences are annotated manually by linguists. The automatic approaches (statistical or machine learning) are based on corpora. Previous work was performed for the most part in English which has rich resources, such as PropBank, VerbNet and FrameNet. These resources were used to serve the automated annotation systems. This type of annotation in other languages for which no corpora of annotated sentences are available often use FrameNet by projection. Although a resource such as FrameNet is necessary for the automated annotation systems and the manual annotation by linguists of a large number of sentences is a tedious and time consuming work. We have proposed an automated system to help linguists in this task so that they have only to validate annotations proposed. Our work focuses on verbs that are more likely than other predicative units (adjectives and nouns) to be accompanied by actants realized in sentences. These verbs are specialized terms of the computer science and Internet domains (ie. access, configure, browse, download) whose actantial structures have been annotated manually with semantic roles. The actantial structure is based on principles of Explanatory and Combinatory Lexicology, LEC of Mel’čuk and appeal in part (with regard to semantic roles) to the notion of Frame Element as described in the theory of frame semantics (FS) of Fillmore. What these two theories have in common is that they lead to the construction of dictionaries different from those resulting from the traditional theories. These manually annotated verbal units in several contexts constitute the specialized corpus that our work will use. Our system designed to assign automatically semantic roles to actants is based on rules and classifiers trained on more than 2300 contexts. We are limited to a restricted list of roles for certain roles in our corpus have not enough examples manually annotated. In our system, we addressed the roles Patient, Agent and destination that the number of examples is greater than 300. We have created a class that we called Autre which we bring to gether the other roles that the number of annotated examples is less than 100. We subdivided the annotation task in the identification of participant actants and circumstants and the assignment of semantic roles to actants that contribute to the sense of the verbal lexical unit. We parsed, with Syntex, the sentences of the corpus to extract syntactic informations that describe the participants of the verbal lexical unit in the sentence. These informations are used as features in our learning model. We have proposed two techniques for the task of participant detection: the technique based in rules and machine learning. These same techniques are used for the task of classification of these participants into actants and circumstants. We proposed to the task of assigning semantic roles to the actants, a partitioning method (clustering) semi supervised of instances that we have compared to the method of semantic role classification. We used CHAMELEON, an ascending hierarchical algorithm.

Ce mémoire de maîtrise présente une nouvelle approche non supervisée pour détecter et segmenter les régions urbaines dans les images hyperspectrales. La méthode proposée n ́ecessite trois étapes. Tout d’abord, afin de réduire le coût calculatoire de notre algorithme, une image couleur du contenu spectral est estimée. A cette fin, une étape de réduction de dimensionalité non-linéaire, basée sur deux critères complémentaires mais contradictoires de bonne visualisation; à savoir la précision et le contraste, est réalisée pour l’affichage couleur de chaque image hyperspectrale. Ensuite, pour discriminer les régions urbaines des régions non urbaines, la seconde étape consiste à extraire quelques caractéristiques discriminantes (et complémentaires) sur cette image hyperspectrale couleur. A cette fin, nous avons extrait une série de paramètres discriminants pour décrire les caractéristiques d’une zone urbaine, principalement composée d’objets manufacturés de formes simples g ́eométriques et régulières. Nous avons utilisé des caractéristiques texturales basées sur les niveaux de gris, la magnitude du gradient ou des paramètres issus de la matrice de co-occurrence combinés avec des caractéristiques structurelles basées sur l’orientation locale du gradient de l’image et la détection locale de segments de droites. Afin de réduire encore la complexité de calcul de notre approche et éviter le problème de la ”malédiction de la dimensionnalité” quand on décide de regrouper des données de dimensions élevées, nous avons décidé de classifier individuellement, dans la dernière étape, chaque caractéristique texturale ou structurelle avec une simple procédure de K-moyennes et ensuite de combiner ces segmentations grossières, obtenues à faible coût, avec un modèle efficace de fusion de cartes de segmentations. Les expérimentations données dans ce rapport montrent que cette stratégie est efficace visuellement et se compare favorablement aux autres méthodes de détection et segmentation de zones urbaines à partir d’images hyperspectrales. / This master’s thesis presents a new approach to urban area detection and segmentation in hyperspectral images. The proposed method relies on a three-step procedure. First, in order to decrease the computational complexity, an informative three-colour composite image, minimizing as much as possible the loss of information of the spectral content, is computed. To this end, a non-linear dimensionality reduction step, based on two complementary but contradictory criteria of good visualization, namely accuracy and contrast, is achieved for the colour display of each hyperspectral image. In order to discriminate between urban and non-urban areas, the second step consists of extracting some complementary and discriminant features on the resulting (three-band) colour hyperspectral image. To attain this goal, we have extracted a set of features relevant to the description of different aspects of urban areas, which are mainly composed of man-made objects with regular or simple geometrical shapes. We have used simple textural features based on grey-levels, gradient magnitude or grey-level co-occurence matrix statistical parameters combined with structural features based on gradient orientation, and straight segment detection. In order to also reduce the computational complexity and to avoid the so-called “curse of dimensionality” when clustering high-dimensional data, we decided, in the final third step, to classify each individual feature (by a simple K-means clustering procedure) and to combine these multiple low-cost and rough image segmentation results with an efficient fusion model of segmentation maps. The experiments reported in this report demonstrate that the proposed segmentation method is efficient in terms of visual evaluation and performs well compared to existing and automatic detection and segmentation methods of urban areas from hyperspectral images.