[pt] SLAM (Mapeamento e Localização Simultânea) é uma das áreas mais pesquisadas na Robótica móvel. Trata-se do problema, num robô móvel, de construir um mapa sem conhecimento prévio do ambiente e ao mesmo tempo manter a sua localização nele. Embora a tecnologia ofereça sensores cada vez mais precisos, pequenos erros na medição são acumulados comprometendo a precisão na localização, sendo estes evidentes quando o robô retorna a uma posição inicial depois de percorrer um longo caminho. Assim, para melhoria do desempenho do SLAM é necessário representar a sua formulação usando teoria das probabilidades. O SLAM com Filtro Extendido de Kalman (EKF-SLAM) é uma solução básica, e apesar de suas limitações é a técnica mais popular. O Fast SLAM, por outro lado, resolve algumas limitações do EKF-SLAM usando uma instância do filtro de partículas conhecida como Rao-Blackwellized. Outra solução bem sucedida é o DP-SLAM, o qual usa uma representação do mapa em forma de grade de ocupação, com um algoritmo hierárquico que constrói mapas 2D bastante precisos. Todos estes algoritmos usam informação de dois tipos de sensores: odômetros e sensores de distância. O Laser Range Finder (LRF) é um medidor laser de distância por varredura, e pela sua precisão é bastante usado na correção do erro em odômetros. Este trabalho apresenta uma detalhada implementação destas três soluções para o SLAM, focalizado em ambientes fechados e estruturados. Apresenta-se a construção de mapas 2D e 3D em terrenos planos tais como em aplicações típicas de ambientes fechados. A representação dos mapas 2D é feita na forma de grade de ocupação. Por outro lado, a representação dos mapas 3D é feita na forma de nuvem de pontos ao invés de grade, para reduzir o custo computacional. É considerado um robô móvel equipado com apenas um LRF, sem nenhuma informação de odometria. O alinhamento entre varreduras laser é otimizado fazendo o uso de Algoritmos Genéticos. Assim, podem-se construir mapas e ao mesmo tempo localizar o robô sem necessidade de odômetros ou outros sensores. Um simulador em Matlab é implementado para a geração de varreduras virtuais de um LRF em um ambiente 3D (virtual). A metodologia proposta é validada com os dados simulados, assim como com dados experimentais obtidos da literatura, demonstrando a possibilidade de construção de mapas 3D com apenas um sensor LRF. / [en] Simultaneous Localization and Mapping (SLAM) is one of the most widely researched areas of Robotics. It addresses the mobile robot problem of generating a map without prior knowledge of the environment, while keeping track of its position. Although technology offers increasingly accurate position sensors, even small measurement errors can accumulate and compromise the localization accuracy. This becomes evident when programming a robot to return to its original position after traveling a long distance, based only on its sensor readings. Thus, to improve SLAM s performance it is necessary to represent its formulation using probability theory. The Extended Kalman Filter SLAM (EKF-SLAM) is a basic solution and, despite its shortcomings, it is by far the most popular technique. Fast SLAM, on the other hand, solves some limitations of the EKFSLAM using an instance of the Rao-Blackwellized particle filter. Another successful solution is to use the DP-SLAM approach, which uses a grid representation and a hierarchical algorithm to build accurate 2D maps. All SLAM solutions require two types of sensor information: odometry and range measurement. Laser Range Finders (LRF) are popular range measurement sensors and, because of their accuracy, are well suited for odometry error correction. Furthermore, the odometer may even be eliminated from the system if multiple consecutive LRF scans are matched. This works presents a detailed implementation of these three SLAM solutions, focused on structured indoor environments. The implementation is able to map 2D environments, as well as 3D environments with planar terrain, such as in a typical indoor application. The 2D application is able to automatically generate a stochastic grid map. On the other hand, the 3D problem uses a point cloud representation of the map, instead of a 3D grid, to reduce the SLAM computational effort. The considered mobile robot only uses a single LRF, without any odometry information. A Genetic Algorithm is presented to optimize the matching of LRF scans taken at different instants. Such matching is able not only to map the environment but also localize the robot, without the need for odometers or other sensors. A simulation program is implemented in Matlab to generate virtual LRF readings of a mobile robot in a 3D environment. Both simulated readings and experimental data from the literature are independently used to validate the proposed methodology, automatically generating 3D maps using just a single LRF.

La capture de mouvement ambulatoire est un sujet en plein essor pour des applications aussi diverses que le suivi des personnes âgées, l'assistance des sportifs de haut niveau, la réhabilitation fonctionnelle, etc. Ces applications exigent que le mouvement ne soit pas contraint par un système externe, qu’il puisse être réalisé dans différentes situations, y compris en extérieur, que l’équipement soit léger et à un faible coût, qu’il soit réellement ambulatoire et sans procédure complexe de calibration.Actuellement, seuls les systèmes utilisant un exosquelette ou bien des modules inertiels (souvent combinés avec des modules magnétiques) permettent d'effectuer de la capture de mouvement de façon ambulatoire. Le poids de l’exosquelette est très important et il impose des contraintes sur les mouvements de la personne, ce qui le rend inutilisable pour certaines applications telles que le suivi de personnes âgées. La technologie inertielle est plus légère. Elle permet d'effectuer la capture du mouvement sans contrainte sur l’espace de mesure ou sur les mouvements réalisés. Par contre, elle souffre de dérives des gyromètres, et le système doit être recalibré.L'objectif de cette thèse est de développer un système de capture de mouvement de chaînes articulées, bas-coût et temps réel, réellement ambulatoire, ne nécessitant pas d'infrastructure de capture spécifique, permettant une utilisation dans de nombreux domaines applicatifs (rééducation, sport, loisirs, etc.).On s'intéresse plus particulièrement à des mesures intra-corporelles. Ainsi, tous les capteurs sont placés sur le corps et aucun dispositif externe n'est utilisé. Outre un démonstrateur final permettant de valider l'approche proposée, on s'astreint à développer également des outils qui permettent de dimensionner le système en termes de technologie, nombre et position des capteurs, mais également à évaluer différents algorithmes de fusion des données. Pour ce faire, on utilise la borne de Cramer-Rao.Le sujet est donc pluridisciplinaire. Il traite des aspects de modélisation et de dimensionnement de systèmes hybrides entièrement ambulatoires. Il étudie des algorithmes d'estimation adaptés au domaine de la capture de mouvement corps entier en traitant les problématiques d'observabilité de l'état et en tenant compte des contraintes biomécaniques pouvant être appliquées. Ainsi, un traitement adapté permet de reconstruire en temps réel la posture du sujet à partir de mesures intra-corporelles, la source étant également placée sur le corps. / Ambulatory motion capture is of great interest for applications ranging for the monitoring of elderly people, sporty performances monitoring, functional rehabilitation, etc. These applications require that the movement is not constrained by an external system, that it can be performed in different situations, including outdoor environment. It requires lightweight and low cost equipment; it must be truly ambulatory without complex process of calibration.Currently, only systems using an exoskeleton or inertial modules (often combined with magnetic modules) can be used in such situations. Unfortunately, the exoskeleton weight is not affordable and it imposes constraints on the movements of the person, which makes it unusable for certain applications such as monitoring of the elderly.Inertial technology is lighter. Itcan be used for the capture of movement without constraints on the capture space or on the movements. However, it suffers from gyros drift, and the system must be recalibrated.The objective of this thesis is to develop a system of motion capture for an articulated chain, low-cost, real-time truly ambulatory that does not require specific capture infrastructure, that can be used in many application fields (rehabilitation, sport, leisure, etc.).We focus on intra-corporal measurements. Thus, all sensors are placed on the body and no external device is used. In addition to a final demonstrator to validate the proposed approach, we also develop tools to evaluate the system in terms of technology, number and position of sensors, but also to evaluate different algorithms for data fusion. To do this, we use the Cramer-Rao lower bound. \\The subject is multidisciplinary. It addresses aspects of modelling and design of fully ambulatory hybrid systems. It studies estimation algorithms adapted to the field of motion capture of a whole body by considering the problem of observability of the state and taking into account the biomechanical constraints that can be taken into account. Thus, with an appropriate treatment, the pose of a subject can be reconstructed in real time from intra-body measurements.

Thesis deals with the detection and tracking of small moving objects from static images. This work shows a general overview of methods and approaches to detection and tracking of objects. There are also described some other approaches to the whole solution. It also included basic definitions, such a noise, convolution and mathematical morphology. The work described Bayesian filtering and Kalman filter. It described the theory of Wavelets, wavelets filters and transformations. The work deals with different ways of the blob`s detection. It is here the design and implementation of applications, which is based on the wavelets filters and Kalman filter. It`s implemented several methods of background subtraction, which are compared by testing. Testing and application are designed to detect vehicles, which are moving faraway (at least 200 m away).

No description available.