Segmentation of sub-cortical Gray Matter (GM) structures in magnetic resonance brain images is crucial in clinic and research for many purposes such as early diagnosis of neurological diseases, guidance of surgical operations and longitudinal volumetric studies. Unfortunately, the algorithms that segment the brain into 3 tissues usually suffer from poor performance in the sub-cortical region. In order to increase the detection of sub-cortical GM structures, an evolutionary hierarchical region merging approach, abbreviated as EHRM, is proposed in this study. Through EHRM, an intensity based region merging is utilized while merging is allowed to proceed among disconnected regions. Texture information is also incorporated into the scheme to prevent the region merging between tissues with similar intensity but different texture properties. The proposed algorithm is tested on real and simulated datasets. The performance is compared with a popular segmentation algorithm, which is also intensity driven: the FAST algorithm [1] in the widely used FSL suite. EHRM is shown to make a significant improvement the detection of sub-cortical GM structures. Average improvements of 10%, 36% and 22% are achieved for caudate, putamen and thalamus respectively. The accuracy of volumetric estimations also increased for GM and WM. Performance of EHRM is robust in presence of bias field. In addition, EHRM operates in O(N) complexity. Furthermore, the algorithm proposed here is simple, because it does not incorporate spatial priors such as an atlas image or intensity priors. With these features, EHRM may become a favorable alternative to the existing brain segmentation tools.

A detecção de gestos é uma etapa importante em aplicações de interação humanocomputador. Se a mão do usuário é detectada com precisão, tanto a análise quanto o reconhecimento do gesto de mão se tornam mais simples e confiáveis. Neste trabalho, descrevemos um novo método para detecção de pele humana, destinada a ser empregada como uma etapa de pré-processamento para segmentação de gestos de mão em sistemas que visam o seu reconhecimento. Primeiramente, treinamos os modelos de cor e textura de pele (material a ser identificado) a partir de um conjunto de treinamento formado por imagens de pele. Nessa etapa, construímos um modelo de mistura de Gaussianas (GMM), para determinar os tons de cor da pele e um dicionário de textons, para textura de pele. Em seguida, introduzimos um estratégia de fusão estocástica de regiões de texturas, para determinar todos os segmentos de diferentes materiais presentes na imagem (cada um associado a uma textura). Tendo obtido todas as regiões, cada segmento encontrado é classificado com base nos modelos de cor de pele (GMM) e textura de pele (dicionário de textons). Para testar o desempenho do algoritmo desenvolvido realizamos experimentos com o conjunto de imagens SDC, projetado especialmente para esse tipo de avaliação (detecção de pele humana). Comparado com outras técnicas do estado-daarte em segmentação de pele humana disponíveis na literatura, os resultados obtidos em nossos experimentos mostram que a abordagem aqui proposta é resistente às variações de cor e iluminação decorrentes de diferentes tons de pele (etnia do usuário), assim como de mudanças de pose da mão, mantendo sua capacidade de discriminar pele humana de outros materiais altamente texturizados presentes na imagem. / Gesture detection is an important task in human-computer interaction applications. If the hand of the user is precisely detected, both analysis and recognition of hand gesture become more simple and reliable. This work describes a new method for human skin detection, used as a pre-processing stage for hand gesture segmentation in recognition systems. First, we obtain the models of color and texture of human skin (material to be identified) from a training set consisting of skin images. At this stage, we build a Gaussian mixture model (GMM) for identifying skin color tones and a dictionary of textons for skin texture. Then, we introduce a stochastic region merging strategy, to determine all segments of different materials present in the image (each associated with a texture). Once the texture regions are obtained, each segment is classified based on skin color (GMM) and skin texture (dictionary of textons) model. To verify the performance of the developed algorithm, we perform experiments on the SDC database, specially designed for this kind of evaluation (human skin detection). Also, compared with other state-ofthe- art skin segmentation techniques, the results obtained in our experiments show that the proposed approach is robust to color and illumination variations arising from different skin tones (ethnicity of the user) as well as changes of pose, while keeping its ability for discriminating human skin from other highly textured background materials.

A detecção de gestos é uma etapa importante em aplicações de interação humanocomputador. Se a mão do usuário é detectada com precisão, tanto a análise quanto o reconhecimento do gesto de mão se tornam mais simples e confiáveis. Neste trabalho, descrevemos um novo método para detecção de pele humana, destinada a ser empregada como uma etapa de pré-processamento para segmentação de gestos de mão em sistemas que visam o seu reconhecimento. Primeiramente, treinamos os modelos de cor e textura de pele (material a ser identificado) a partir de um conjunto de treinamento formado por imagens de pele. Nessa etapa, construímos um modelo de mistura de Gaussianas (GMM), para determinar os tons de cor da pele e um dicionário de textons, para textura de pele. Em seguida, introduzimos um estratégia de fusão estocástica de regiões de texturas, para determinar todos os segmentos de diferentes materiais presentes na imagem (cada um associado a uma textura). Tendo obtido todas as regiões, cada segmento encontrado é classificado com base nos modelos de cor de pele (GMM) e textura de pele (dicionário de textons). Para testar o desempenho do algoritmo desenvolvido realizamos experimentos com o conjunto de imagens SDC, projetado especialmente para esse tipo de avaliação (detecção de pele humana). Comparado com outras técnicas do estado-daarte em segmentação de pele humana disponíveis na literatura, os resultados obtidos em nossos experimentos mostram que a abordagem aqui proposta é resistente às variações de cor e iluminação decorrentes de diferentes tons de pele (etnia do usuário), assim como de mudanças de pose da mão, mantendo sua capacidade de discriminar pele humana de outros materiais altamente texturizados presentes na imagem. / Gesture detection is an important task in human-computer interaction applications. If the hand of the user is precisely detected, both analysis and recognition of hand gesture become more simple and reliable. This work describes a new method for human skin detection, used as a pre-processing stage for hand gesture segmentation in recognition systems. First, we obtain the models of color and texture of human skin (material to be identified) from a training set consisting of skin images. At this stage, we build a Gaussian mixture model (GMM) for identifying skin color tones and a dictionary of textons for skin texture. Then, we introduce a stochastic region merging strategy, to determine all segments of different materials present in the image (each associated with a texture). Once the texture regions are obtained, each segment is classified based on skin color (GMM) and skin texture (dictionary of textons) model. To verify the performance of the developed algorithm, we perform experiments on the SDC database, specially designed for this kind of evaluation (human skin detection). Also, compared with other state-ofthe- art skin segmentation techniques, the results obtained in our experiments show that the proposed approach is robust to color and illumination variations arising from different skin tones (ethnicity of the user) as well as changes of pose, while keeping its ability for discriminating human skin from other highly textured background materials.

A detecção de gestos é uma etapa importante em aplicações de interação humanocomputador. Se a mão do usuário é detectada com precisão, tanto a análise quanto o reconhecimento do gesto de mão se tornam mais simples e confiáveis. Neste trabalho, descrevemos um novo método para detecção de pele humana, destinada a ser empregada como uma etapa de pré-processamento para segmentação de gestos de mão em sistemas que visam o seu reconhecimento. Primeiramente, treinamos os modelos de cor e textura de pele (material a ser identificado) a partir de um conjunto de treinamento formado por imagens de pele. Nessa etapa, construímos um modelo de mistura de Gaussianas (GMM), para determinar os tons de cor da pele e um dicionário de textons, para textura de pele. Em seguida, introduzimos um estratégia de fusão estocástica de regiões de texturas, para determinar todos os segmentos de diferentes materiais presentes na imagem (cada um associado a uma textura). Tendo obtido todas as regiões, cada segmento encontrado é classificado com base nos modelos de cor de pele (GMM) e textura de pele (dicionário de textons). Para testar o desempenho do algoritmo desenvolvido realizamos experimentos com o conjunto de imagens SDC, projetado especialmente para esse tipo de avaliação (detecção de pele humana). Comparado com outras técnicas do estado-daarte em segmentação de pele humana disponíveis na literatura, os resultados obtidos em nossos experimentos mostram que a abordagem aqui proposta é resistente às variações de cor e iluminação decorrentes de diferentes tons de pele (etnia do usuário), assim como de mudanças de pose da mão, mantendo sua capacidade de discriminar pele humana de outros materiais altamente texturizados presentes na imagem. / Gesture detection is an important task in human-computer interaction applications. If the hand of the user is precisely detected, both analysis and recognition of hand gesture become more simple and reliable. This work describes a new method for human skin detection, used as a pre-processing stage for hand gesture segmentation in recognition systems. First, we obtain the models of color and texture of human skin (material to be identified) from a training set consisting of skin images. At this stage, we build a Gaussian mixture model (GMM) for identifying skin color tones and a dictionary of textons for skin texture. Then, we introduce a stochastic region merging strategy, to determine all segments of different materials present in the image (each associated with a texture). Once the texture regions are obtained, each segment is classified based on skin color (GMM) and skin texture (dictionary of textons) model. To verify the performance of the developed algorithm, we perform experiments on the SDC database, specially designed for this kind of evaluation (human skin detection). Also, compared with other state-ofthe- art skin segmentation techniques, the results obtained in our experiments show that the proposed approach is robust to color and illumination variations arising from different skin tones (ethnicity of the user) as well as changes of pose, while keeping its ability for discriminating human skin from other highly textured background materials.

The emergence of cloud computing has transformed the potential of robotics by enabling multi-robotic teams to fulfil complex tasks in the cloud. This paradigm is known as “cloud robotics” and relieves robots from hardware and software limitations, as large amounts of available resources and parallel computing capabilities are available in the cloud. The introduction of cloud-enabled robots alleviates the need for computationally intensive robots to be built, as many, if not all, of the CPU-intensive tasks can be offloaded into the cloud, resulting in multi-robots that require much less power, energy consumption and on-board processing units. While the benefits of cloud robotics are clearly evident and have resulted in an increase in interest among the scientific community, one of the biggest challenges of cloud robotics is the inherent communication challenges brought about by disconnections between the multi-robotic system and the cloud. The communication delays brought about by the cloud disconnection results in robots not being able to receive and transmit data to the physical cloud. The unavailability of these robotic services in certain instances could prove fatal in a heterogeneous environment that requires multi-robotic teams to assist with the saving of human lives. This niche area is relatively unexplored in the literature. This work serves to assist with the challenge of disconnection in cloud robotics by proposing a survivable cloud multi-robotics (SCMR) framework for heterogeneous environments. The SCMR framework leverages the combination of a virtual ad hoc network formed by the robot-to-robot communication and a physical cloud infrastructure formed by the robot-to-cloud communications. The Quality of Service (QoS) on the SCMR framework is tested and validated by determining the optimal energy utilization and Time of Response (ToR) on drivability analysis with and without cloud connection. The experimental results demonstrate that the proposed framework is feasible for current multi-robotic applications and shows the survivability aspect of the framework in instances of cloud disconnection. / School of Computing / M.Sc. (Computer Science)

The emergence of cloud computing has transformed the potential of robotics by enabling multi-robotic teams to fulfil complex tasks in the cloud. This paradigm is known as “cloud robotics” and relieves robots from hardware and software limitations, as large amounts of available resources and parallel computing capabilities are available in the cloud. The introduction of cloud-enabled robots alleviates the need for computationally intensive robots to be built, as many, if not all, of the CPU-intensive tasks can be offloaded into the cloud, resulting in multi-robots that require much less power, energy consumption and on-board processing units. While the benefits of cloud robotics are clearly evident and have resulted in an increase in interest among the scientific community, one of the biggest challenges of cloud robotics is the inherent communication challenges brought about by disconnections between the multi-robotic system and the cloud. The communication delays brought about by the cloud disconnection results in robots not being able to receive and transmit data to the physical cloud. The unavailability of these robotic services in certain instances could prove fatal in a heterogeneous environment that requires multi-robotic teams to assist with the saving of human lives. This niche area is relatively unexplored in the literature. This work serves to assist with the challenge of disconnection in cloud robotics by proposing a survivable cloud multi-robotics (SCMR) framework for heterogeneous environments. The SCMR framework leverages the combination of a virtual ad hoc network formed by the robot-to-robot communication and a physical cloud infrastructure formed by the robot-to-cloud communications. The Quality of Service (QoS) on the SCMR framework is tested and validated by determining the optimal energy utilization and Time of Response (ToR) on drivability analysis with and without cloud connection. The experimental results demonstrate that the proposed framework is feasible for current multi-robotic applications and shows the survivability aspect of the framework in instances of cloud disconnection. / School of Computing / M.Sc. (Computer Science)

Conselho Nacional de Desenvolvimento CientÃfico e TecnolÃgico / A cobertura regular de quase todo o planeta por sistemas de radar de abertura sintÃtica (synthetic aperture radar - SAR) orbitais e o uso de sistemas aerotransportados tÃm propiciado novos meios para obter informaÃÃes atravÃs do sensoriamento remoto de vÃrias regiÃes de nosso planeta, muitas delas inacessÃveis. Este trabalho trata do processamento de imagens digitais geradas por radar de abertura sintÃtica, especificamente da segmentaÃÃo, que consiste do isolamento ou particionamento dos objetos relevantes presentes em uma cena. A segmentaÃÃo de imagens digitais visa melhorar a interpretaÃÃo das mesmas em procedimentos subseqÃentes. As imagens SAR sÃo corrompidas por ruÃdo coerente, conhecido por speckle, que mascara pequenos detalhes e zonas de transiÃÃo entre os objetos. Tal ruÃdo à inerente ao processo de formaÃÃo dessas imagens e dificulta tarefas como a segmentaÃÃo automÃtica dos objetos existentes e a identificaÃÃo de seus contornos. Uma possibilidade para efetivar a segmentaÃÃo de imagens SAR consiste na filtragem preliminar do ruÃdo speckle, como etapa de tratamento dos dados. A outra possibilidade, aplicada neste trabalho, consiste em segmentar diretamente a imagem ruidosa, usando seus pixels originais como fonte de informaÃÃo. Para isso, à desenvolvida uma metodologia de segmentaÃÃo baseada em crescimento e fusÃo estatÃstica de regiÃes, que requer alguns parÃmetros para controlar o processo. As vantagens da utilizaÃÃo dos dados originais para realizar a segmentaÃÃo de imagens de radar sÃo a eliminaÃÃo de etapas de prÃ-processamento e o favorecimento da detecÃÃo das estruturas presentes nas mesmas. à realizada uma avaliaÃÃo qualitativa e quantitativa das imagens segmentadas, sob diferentes situaÃÃes, aplicando a tÃcnica proposta em imagens de teste contaminadas artificialmente com ruÃdo multiplicativo. Este segmentador à aplicado tambÃm no processamento de imagens SAR reais e os resultados sÃo promissores. / The regular coverage of the planet surface by spaceborne synthetic aperture radar (SAR)and also airborne systems have provided alternative means to gather remote sensing information of various regions of the planet, even of inaccessible areas. This work deals with the digital processing of synthetic aperture radar imagery, where segmentation is the main subject. It consists of isolating or partitioning relevant objects in a scene, aiming at improving image interpretation and understanding in subsequent tasks. SAR images are contaminated by coherent noise, known as speckle, which masks small details and transition zones among the objects. Such a noise is inherent in radar image generation process, making difficult tasks like automatic segmentation of the objects, as well as their contour identification. To segment radar images, one possible way is to apply speckle filtering before segmentation. Another one, applied in this work, is to perform noisy image segmentation using the original SAR pixels as input data, without any preprocessing,such as filtering. To provide segmentation, an algorithm based on region growing and statistical region merging has been developed, which requires some parameters to control the process. This task presents some advantages, as long as it eliminates preprocessing steps and favors the detection of the image structures, since original pixel information is exploited. A qualitative and quantitative performance evaluation of the segmented images is also executed, under different situations, by applying the proposed technique to simulated images corrupted with multiplicative noise. This segmentation method is also applied to real SAR images and the produced results are promising.

D’un point de vue général, les travaux de recherche de cette thèse s’inscrivent dans le cadre d’une approche globale quiconsiste à extraire des informations relatives aux toitures de bâtiments à partir de photos aériennes (orthophotoplans). L’objectifétant de pouvoir reconnaître des toitures extraites d’images aériennes en utilisant une base de connaissances, puisaffiner/déformer des modèles 3D générés automatiquement à partir de données géographiques. Pour cela, une premièreétape consiste tout d’abord à partitionner l’image aérienne en différentes régions d’intérêt (pans de toiture, cheminées,chiens assis, fenêtres, etc.), c’est la contribution de cette thèse.La méthodologie permettant d’atteindre cet objectif est composée de trois étapes : (i) Une étape de simplification qui consisteà simplifier l’image initiale avec un couple invariant/gradient approprié et optimisé pour l’application. Pour cela, unesérie de tests permettant de choisir, d’une part, l’invariant colorimétrique le plus approprié parmi 24 invariants et, d’autrepart, le meilleur gradient parmi 14 gradients issus de la littérature est réalisée. (ii) La deuxième étape comporte deux stratégiesde segmentation par LPE. L’image simplifiée est segmentée d’une part par une LPE-régions couplée à une stratégiede fusion de régions, et d’autre part, par une LPE-contours. Le processus de fusion de régions intègre des critères defusion fondés sur des grandeurs radiométriques et géométriques adaptés aux particularités des orthophotoplans traités.Une technique de caractérisation 2D des arêtes de toitures par une analyse des segments est proposée afin de calculerl’un des critères de fusion. (iii) La troisième étape consiste à combiner les avantages de chaque méthode dans un mêmeschéma de segmentation coopératif afin d’aboutir à un résultat de segmentation fiable. Les tests ont été effectués sur unorthophotoplan contenant 100 toitures de complexité variée et évaluées avec le critère de VINET utilisant une segmentationde référence afin de prouver la robustesse et la fiabilité de l’approche proposée. Une étape de comparaison permettantde situer les résultats obtenus via notre approche proposée par rapport à ceux obtenus pas les principales méthodes desegmentation de la littérature est finalement effectuée. / The work presented in this thesis is developed in a global approach that consists in recognizing roofs extracted from aerialimages using a knowledge database, and bending out 3D models automatically generated from geographical data. Themain step presented in this thesis consists in segmenting roof images in different regions of interest in order to provideseveral measures of roofs (section of roofs, chimneys, roof light, etc).The method aimed at achieving this goal is composed of three principal steps: (i) A simplification step that consists insimplifying the image with an appropriate (optimized for the application) couple of invariant/gradient. For that, several testshave been performed to choose a suitable colorimetric invariant among a set of 24 invariants and define the best gradientamong 14 gradients (eight gray level gradients and six color gradients) of the literature. (ii) The second step is composedof two main treatments: On the one hand, the preliminary orthophotoplan segmentation is produced using the watershedregions applied on the simplified image. An efficient region merging strategy is then applied in order to deal with theover-segmentation problem. The regions merging procedure includes a merging criteria adapted to the orthophotoplanparticularities. In order to calculate one of the merging criteria, a 2D modeling of roof ridges strategy is proposed. Onthe other hand, the simplified image is segmented by the watershed lines. (iii) The third step consists in integrating bothsegmentation strategies by watershed algorithm into a single cooperative segmentation scheme to achieve satisfactorysegmentation results. Tests have been performed on an orthophotoplan containing 100 roofs with varying complexity andevaluated with VINET criteria using a ground truth image segmentation. Comparison results with five popular segmentationtechniques of the literature demonstrates the effectiveness and the reliability of the proposed approach.

This master´s thesis deals with depiction aplication of camera like sensitive element for assisting system of car. It was proposed, how find a road, a lane and a obstacle on roadways. Only one camera was aplication for it. Solution is realized by methods, that are based on color and gradient of image. It applies simple methods and methods with mathematical model. Result is sum of method and its test and comparing. Realization of my program is present at resume of this thesis.