Les chargements thermomécaniques cycliques conduisent à l'apparition de fissures courtes de fatigue dont la phase d'amorçage est prépondérante par rapport à la durée de vie totale du composant. Les mécanismes liés à l'évolution de ces fissures présentent une forte dépendance vis-à-vis de la microstructure du matériau, notamment à faible amplitude de chargement. Afin d'étudier cette dépendance, une étude statistique basée sur des résultats de calculs éléments finis d'agrégats polycristallins a été conduite. Le problème a d'abord été traité en deux dimensions, avec un modèle de plasticité cristalline simplifié. L'analyse des résultats a permis de mettre en évidence un effet de voisinage dans les polycristaux et de retranscrire la dispersion des résultats de fatigue observés en surface. En passant à un modèle en trois dimensions et étudiant différents chargements, il a été possible de corréler les structures de localisation de la déformation plastique dans le volume de l'agrégat avec les observations expérimentales. Également, les marches d'intrusion/extrusion induites en surface par les chargements de fatigue, sites préférentiels d'amorçage, ont été reproduites. Elles mettent en exergue la nocivité des chargements de type équibiaxial. Le problème des effets de bords a été soulevé, car il réduit l'échantillon des résultats exploitables par un traitement statistique. Ensuite, l'introduction de la rugosité de surface dans les agrégats simulés a mis en lumière la compétition entre les paramètres microstructuraux du matériau et les singularités géométriques de sa surface libre. Cette analyse a permis de montrer qu'il doit exister un état de surface limite, à partir duquel l'effet de la rugosité prend le dessus sur l'aspect cristallographique concernant la localisation de la déformation plastique en surface des polycristaux. Néanmoins, cet effet s'estompe très rapidement dans la profondeur et devient presque nul après la première rangée de grains surfaciques. Enfin, une modélisation à la fois périodique et rendant compte de l'effet de surface libre, appelée modélisation semi-périodique, a été mise en place pour éliminer les effets de bords et exploiter pleinement les résultats de calculs. Cette méthode apporte aussi de nouvelles perspectives pour l'élaboration d'un modèle à champs moyens pour les polycristaux en surface.

Cette thèse présente d'une part, la construction d'approximations parcimonieuses de champs acoustiques, d'autre part l'utilisation de ces approximations pour la résolution de problèmes inverses. L'approximation de solutions de l'équation d'Helmholtz est étendue à des modèles de vibration de plaques, ce qui, entre autres, permet également de concevoir une méthode de calcul de mode propres alternative aux méthodes de référence. Différents problèmes inverses sont ensuite étudiés, en se basant sur ces résultats d'approximation. Le premier est l'holographie acoustique en champ propre, pour laquelle nous développons une nouvelle méthode de régularisation, ainsi qu'une antenne aléatoire permettant de réduire le nombre de mesures nécessaires à la reconstruction de déformées opérationnelles de plaques. Le deuxième problème inverse étudié est l'interpolation spatiale de réponses impulsionnelles de plaques, où nous montrons qu'en mesurant le champ vibratoire sur un ensemble d'échantillons bien choisi (justifié par une analyse théorique), les réponses impulsionnelles d'une plaque peuvent être obtenues avec moins de mesures que demandées par le théorème d'échantillonnage de Shannon. Enfin, le problème de la localisation de sources dans un espace clos réverbérant est étudié. Nous montrons qu'en utilisant des modèles parcimonieux d'ondes, la localisation est possible sans connaissances a priori sur les conditions aux limites du domaine de propagation des ondes.

Les crossing-overs (CO) sont indispensables dans la création variétale pour permettre l'introgression de régions d'intérêt dans les variétés agronomiques d'espèces cultivées telles que le blé tendre (Triticum aesstivum L.). Afin d'évaluer l'impact des facteurs qui influencent leur formation, nous avons entrepris une caractérisation fine de leur distribution sur le plus grand chromosome ( chromosome 3B, 995Mb) du blé tendre en s'appuyant sur la carte physique récemment développée et le séquençage de quelques régions de plusieurs mégabases. La comparaison entre une carte génétique dense (102 marqueurs) et une carte physique de délétion montre que 77% des CO sont présents dans les régions distales couvrant seuleument 25% du chromosome. La comparaison de différentes cartes génétiques montre de plus que cette distribution est conservée entre populations avec cependant des différences de taux de CO locaux entre populations mais également entre méiose mâle et femelle. Cette distribution est influencée par une interférence positive forte à des distances inférieures à 10 cM. Cependant, les faibles fréquences de CO observées au sein des régions proximales restent inexpliquées. En effet, la faible augmentation du taux de CO observée au sein des régions proximales placées en position distale suggère que la position proximale de ces régions sur le chromosome ne semble pas responsable de leur faible fréquence de CO. De plus, nous avons montré que ces faibles fréquences ne seraient pas non plus dues à une divergence de séquence entre chromosomes homologues au sein des régions proximales, la fréquence de CO étant toujours aussi faible au sein de celles-ci entre deux chromosomes homozygotes. En revanche, l'analyse à l'échelle d'une région séquencée de 3.1 Mb indique que les fréquences de CO importantes sont fortement corrélées avec la présence de gènes. L'inhibition de la formation des CO au sein des régions proximales pourrait ainsi s'expliquer par la présence de gènes en quantité moins importante dans ces régions comparées aux régions distales.

Pour être autonome, un robot mobile doit être capable de décrire l'environnement dans lequel il évolue. Pour ce faire, le robot doit pouvoir se localiser et construire une carte de l'environnement si celle-ci n'existe pas. Grâce à ces informations, il pourra par exemple éviter les obstacles et donc naviguer en toute sécurité. Cette capacité correspond à une étape obligatoire sur la route de son autonomie totale. Le sujet traité dans ce manuscrit répond à cette problématique. Il concerne la localisation d'un véhicule et la cartographie d'un environnement en simultané, SLAM pour simultaneous localization and mapping. Après un etat de l'art des approches proposées durant la dernière décennie, deux points importants ont été constatés lors de l'emploi d'un filtre EKF-SLAM : les estimations obtenues avec ce filtre ne sont pas consistantes et cette approche, telle qu'elle a été introduite, ne répond pas aux critères d'observabilité d'un système. Pour chacun de ces problèmes, une solution est proposée et critiquée à l'aide de résultats de simulation et d'expérimentation

.

Robust long-term positioning for autonomous mobile robots is essential for many applications. In many environments this task is challenging, as errors accumulate in the robot’s position estimate over time. The robot must also build a map so that these errors can be corrected when mapped regions are re-visited; this is known as Simultaneous Localisation and Mapping, or SLAM. Successful SLAM schemes have been demonstrated which accurately map tracks of tens of kilometres, however these schemes rely on expensive sensors such as laser scanners and inertial measurement units. A more attractive, low-cost sensor is a digital camera, which captures images that can be used to recognise where the robot is, and to incrementally position the robot as it moves. SLAM using a single camera is challenging however, and many contemporary schemes suffer complete failure in dynamic or featureless environments, or during erratic camera motion. An additional problem, known as scale drift, is that cameras do not directly measure the scale of the environment, and errors in relative scale accumulate over time, introducing errors into the robot’s speed and position estimates. Key to a successful visual SLAM system is the ability to continue operation despite these difficulties, and to recover from positioning failure when it occurs. This thesis describes the development of such a scheme, which is known as BoWSLAM. BoWSLAM enables a robot to reliably navigate and map previously unknown environments, in real-time, using only a single camera. In order to position a camera in visually challenging environments, BoWSLAM combines contemporary visual SLAM techniques with four new components. Firstly, a new Bag-of-Words (BoW) scheme is developed, which allows a robot to recognise places it has visited previously, without any prior knowledge of its environment. This BoW scheme is also used to select the best set of frames to reconstruct positions from, and to find efficient wide-baseline correspondences between many pairs of frames. Secondly, BaySAC, a new outlier- robust relative pose estimation scheme based on the popular RANSAC framework, is developed. BaySAC allows the efficient computation of multiple position hypotheses for each frame. Thirdly, a graph-based representation of these position hypotheses is proposed, which enables the selection of only reliable position estimates in the presence of gross outliers. Fourthly, as the robot explores, objects in the world are recognised and measured. These measurements enable scale drift to be corrected. BoWSLAM is demonstrated mapping a 25 minute 2.5km trajectory through a challenging and dynamic outdoor environment in real-time, and without any other sensor input; considerably further than previous single camera SLAM schemes.

Freight transportation is an integral element of various supply chains and has a complex and dynamical interrelationship with human economic activities. Modern logistical strategies paved way to the current supply chain organisation and logistics network design resulting in a more global economy and huge economies of scale. Recent trends of volatility of oil price have major implications in the movement of commodities across the supply chains. Likewise, climate change issues have presented urgent challenges in reducing carbon emissions for the transport and logistics sector. Pressure on the sector comes from both governments and consumers alike, demanding future sustainability as well as corporate environmental and social responsibility. The original contribution of this research is to investigate the system-wide dynamics of freight transportation and production in the context of supply chains. A theoretical framework called the ‘Geographic Adaptive Potential’ or GAP is built to understand how constraints in energy and emissions affect the production and distribution of commodities. The changes in the supply chain were investigated in four different components, namely a) the potential to shift to less energy and emissions intensive modes for long-haul freight, b) logistical strategies in the last leg of the chain or urban freight and c) local production and distribution, and d) the accessibility of potential customers to the markets. The design of the GAP components is in correspondence with the links of the supply chain. The analyses yielded an evaluation of the adaptive capacity of the freight transport and production system. For long-haul freight, a GIS-based model was created called the ‘New Zealand Intermodal Freight Network’ or NZIFN. It is an optimisation tool integrating the road, rail and shipping network of New Zealand and calculates that minimum time, operating costs, energy and emissions routes between 2 given locations. The case studies of Auckland to Wellington and Auckland to Christchurch distributions of non-perishable products established that even a marginal increase of rail and coastal shipping share produced around 10% reduction in both freight energy and greenhouse gas emissions. In the study of the last leg of the supply chain, the truck trip generation rates of different food stores were investigated. The strongest factors influencing the trip rates to a store are its size and product variation, the latter being a new parameter introduced in the dissertation. It is defined as the total number of brands for 6 chosen commodities commonly found in the stores. The trip rates together with the truck type and distance travelled were used to compute the freight energy usage of the stores. Results revealed that supermarkets consume the most energy for their delivery operations but relative to its physical size, they are more energy efficient than smaller stores. This is due to the utilisation of advanced logistical strategies such as freight consolidation and the effective use of distribution centres. The localised production chapter was explored in the context of Farmers’ markets and their difference with the conventional supermarket distribution system. Using a freight transport energy audit, the energy intensities of both systems were compared. The findings showed that Farmers’ markets were more energy-intensive than supermarkets owing to the low volumes of goods delivered to the market and the lack of freight consolidation effort in the system. The study on the active mode access of potential customers to both Farmers’ markets and supermarkets captured the interplay between freight and personal transport and is the final component of GAP. The results of the ArcGIS based model called ‘Active Mode Access’ or AMA demonstrated that both Farmers’ markets and supermarkets have the same level of accessibility for walking or biking customers. However, the calculations also showed that almost 87% of New Zealanders have no AMA to stores and are at risk for fuel price increase. Finally, the key result of this dissertation is the assessment that there is actually limited adaptive capacity of the freight transport and production system. This is due to network infrastructure and geographical constraints as well as commodity type and mode compatibility and other operational concerns. Due to these limitations, the GAP model assessed that reduction in energy and allowable emissions will ultimately reduce the amount of commodities moved in the system.

This thesis explores the use of invariant features for learning sprites from image sequences, and for recognising object categories in images. A popular framework for the interpretation of image sequences is the layers or sprite model of e.g. Wang and Adelson (1994), Irani et al. (1994). Jojic and Frey (2001) provide a generative probabilistic model framework for this task, but their algorithm is slow as it needs to search over discretised transformations (e.g. translations, or affines) for each layer. We show that by using invariant features (e.g. Lowe’s SIFT features) and clustering their motions we can reduce or eliminate the search and thus learn the sprites much faster. The algorithm is demonstrated on example image sequences. We introduce the Generative Template of Features (GTF), a parts-based model for visual object category detection. The GTF consists of a number of parts, and for each part there is a corresponding spatial location distribution and a distribution over ‘visual words’ (clusters of invariant features). We evaluate the performance of the GTF model for object localisation as compared to other techniques, and show that such a relatively simple model can give state-of- the-art performance. We also discuss the connection of the GTF to Hough-transform-like methods for object localisation.

Intelligent Transportation Systems (ITSs) play an important role in modern traffic management, which can be divided into intelligent infrastructure systems and intelligent vehicle systems. Automatic Number Plate Recognition systems (ANPRs) are one of infrastructure systems that allow users to track, identify and monitor moving vehicles by automatically extracting their number plates. ANPR is a well proven technology that is widely used throughout the world by both public and commercial organisations. There are a wide variety of commercial uses for the technology that include automatic congestion charge systems, access control and tracing of stolen cars. The fundamental requirements of an ANPR system are image capture using an ANPR camera and processing of the captured image. The image processing part, which is a computationally intensive task, includes three stages: Number Plate Localisation (NPL), Character Segmentation (CS) and Optical Character Recognition (OCR). The common hardware choice for its implementation is often high performance workstations. However, the cost, compactness and power issues that come with these solutions motivate the search for other platforms. Recent improvements in low-power high-performance Field Programmable Gate Arrays (FPGAs) and Digital Signal Processors (DSPs) for image processing have motivated researchers to consider them as a low cost solution for accelerating such computationally intensive tasks. Current ANPR systems generally use a separate camera and a stand-alone computer for processing. By optimising the ANPR algorithms to take specific advantages of technical features and innovations available within new FPGAs, such as low power consumption, development time, and vast on-chip resources, it will be possible to replace the high performance roadside computers with small in-camera dedicated platforms. In spite of this, costs associated with the computational resources required for complex algorithms together with limited memory have hindered the development of embedded vision platforms. The work described in this thesis is concerned with the development of a range of image processing algorithms for NPL, CS and OCR and corresponding FPGA architectures. MATLAB implementations have been used as a proof of concept for the proposed algorithms prior to the hardware implementation. The proposed architectures are speed/area efficient architectures, which have been implemented and verified using the Mentor Graphics RC240 FPGA development board equipped with a 4M Gates Xilinx Virtex-4 LX40. The proposed NPL architecture can localise a number plate in 4.7 ms whilst achieving a 97.8% localisation rate and consuming only 33% of the available area of the Virtex-4 FPGA. The proposed CS architecture can segment the characters within a NP image in 0.2-1.4 ms with 97.7% successful segmentation rate and consumes only 11% of the Virtex-4 FPGA on-chip resources. The proposed OCR architecture can recognise a character in 0.7 ms with 97.3% successful recognition rate and consumes only 23% of the Virtex-4 FPGA available area. In addition to the three main stages, two pre-processing stages which consist of image binarisation, rotation and resizing are also proposed to link these stages together. These stages consume 9% of the available FPGA on-chip resources. The overall results achieved show that the entire ANPR system can be implemented on a single FPGA that can be placed within an ANPR camera housing to create a stand-alone unit. As the benefits of this are drastically improve energy efficiency and removing the need for the installation and cabling costs associated with bulky PCs situated in expensive, cooled, waterproof roadside cabinets.

In recent years, advances in signal processing have led to small, low power, inexpensive Wireless Sensor Network (WSN). The signal processing in WSN is different from the traditional wireless networks in two critical aspects: firstly, the signal processing in WSN is performed in a fully distributed manner, unlike in traditional wireless networks; secondly, due to the limited computation capabilities of sensor networks, it is essential to develop an energy and bandwidth efficient signal processing algorithms. Target localisation and tracking problems in WSNs have received considerable attention recently, driven by the necessity to achieve higher localisation accuracy, lower cost, and the smallest form factor. Received Signal Strength (RSS) based localisation techniques are at the forefront of tracking research applications. Since tracking algorithms have been attracting research and development attention recently, prolific literature and a wide range of proposed approaches regarding the topic have emerged. This thesis is devoted to discussing the existing WSN-based localisation and tracking approaches. This thesis includes five studies. The first study leads to the design and implementation of a triangulation-based localisation approach using RSS technique for indoor tracking applications. The presented work achieves low localisation error in complex environments by predicting the environmental characteristics among beacon nodes. The second study concentrates on investigating a fingerprinting localisation method for indoor tracking applications. The proposed approach offers reasonable localisation accuracy while requiring a short period of offline computation time. The third study focuses on designing and implementing a decentralised tracking approach for tracking multiple mobile targets with low resource requirements. Despite the interest in target tracking and localisation issues, there are few systems deployed using ZigBee network standard, and no tracking system has used the full features of the ZigBee network standard. Tracking through the ZigBee is a challenging task when the density of router and end-device nodes is low, due to the limited communication capabilities of end-device nodes. The fourth study focuses on developing and designing a practical ZigBee-based tracking approach. To save energy, different strategies were adopted. The fifth study outlines designing and implementing an energy-efficient approach for tracking applications. This study consists of two main approaches: a data aggregation approach, proposed and implemented in order to reduce the total number of messages transmitted over the network; and a prediction approach, deployed to increase the lifetime of the WSN. For evaluation purposes, two environmental models were used in this thesis: firstly, real experiments, in which the proposed approaches were implemented on real sensor nodes, to test the validity for the proposed approaches; secondly, simulation experiments, in which NS-2 was used to evaluate the power-consumption issues of the two approaches proposed in this thesis.