Cette thèse porte sur la segmentation supervisée de flux vidéo dans un contexte applicatif lié à la reconnaissance d'actions de la vie courante.La méthode de segmentation proposée est dérivée la méthode des modèles de formes implicites (Implicit Shape Model) et s'obtient en optimisant les votes présents dans cette méthode d'élection.Nous démontrons que cette optimisation (dans un contexte de fenêtre temporelle glissante) peut être exprimée de manière équivalente dans le formalisme des SVM en imposant une contrainte de cohérence temporelle à l'apprentissage, ou, en représentant la fenêtre glissante selon une décomposition pyramidale dense.Tout ce processus est validé expérimentalement sur un jeu de données de la littérature de segmentation supervisée.Il y surpasse les autres méthodes de type modèles de formes implicites et le SVM linéaire standard.La méthode proposée est ensuite mise en œuvre dans le cadre de la segmentation supervisée d'actions.Pour cela, des primitives dédiées sont extraites des données squelette de la personne d'intérêt obtenues grâce à des logiciels standards.Ces primitives sont ensuite quantifiées puis utilisées par la méthode d'élection.Ce système de segmentation d'actions obtient les meilleurs scores de l'état de l'art sur un jeu de données de la littérature de reconnaissance d'actions, ce qui valide cette combinaison des primitives et de la méthode d'élection. / This thesis focuses on the supervised segmentation of video streams within the application context of daily action recognition.A segmentation algorithm is obtained from Implicit Shape Model by optimising the votes existing in this polling method.We prove that this optimisation can be linked to the sliding windows plus SVM framework and more precisely is equivalent with a standard training by adding temporal constraint, or, by encoding the data through a dense pyramidal decomposition. This algorithm is evaluated on a public database of segmentation where it outperforms other Implicit Shape Model like methods and the standard linear SVM.This algorithm is then integrated into a action segmentation system.Specific features are extracted from skeleton obtained from the video by standard software.These features are then clustered and given to the polling method.This system, combining our feature and our algorithm, obtains the best published performance on a human daily action segmentation dataset.

Submitted by Fabio Sobreira Campos da Costa (fabio.sobreira@ufpe.br) on 2016-07-01T12:22:19Z No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Tese_Filipe_Cordeiro.pdf: 19608976 bytes, checksum: a0ff2fa1256af4323f10bfcbb3df974d (MD5) / Made available in DSpace on 2016-07-01T12:22:19Z (GMT). No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Tese_Filipe_Cordeiro.pdf: 19608976 bytes, checksum: a0ff2fa1256af4323f10bfcbb3df974d (MD5) Previous issue date: 2015-12-16 / De acordo com a Organização Mundial de Saúde, o câncer de mama é a forma mais comum de câncer entre as mulheres no mundo todo, sendo um dos tipos de câncer mais fatal. Estudos mostram que o diagnóstico precoce pode contribuir para a redução da taxa de mortalidade e aumentar as opções de tratamento. Apesar da existência de várias técnicas de obtenção de imagens no auxílio ao diagnóstico de câncer de mama, a mamografia digital é ainda a tecnologia mais eficaz e utilizada para esse fim. Consequentemente, a segmentação de imagens de mamografia é uma tarefa fundamental para auxiliar o diagnóstico, levando em consideração a forma da lesão mamária e suas bordas. No entanto, a segmentação de imagens de mamografia é uma tarefa complexa, uma vez que ela é muito dependente dos tipos de tecido mamário e da lesão. O algoritmo GrowCut é um método de segmentação de propósito geral baseado em autômatos celulares, capaz de realizar uma segmentação precisa através da seleção adequada de pontos internos e externos à região de interesse. Neste trabalho é apresentado um novo algoritmo semi-supervisionado baseado na modificação do algoritmo GrowCut para realizar segmentação de imagens de mamografia de forma semi-automática. No método proposto é utilizada uma função de pertinência fuzzy Gaussiana para modificar a regra de evolução do algoritmo GrowCut original, visando estimar as probabilidades de um pixel pertencer ao objeto ou fundo da imagem. Esse modelo permite uma maior flexibilidade na inicialização das sementes quando comparado à trabalhos no estado da arte, pois a marcação realizada pelo especialista é utilizada extraindo-se informação do conjunto de sementes, e não informações do posicionamento individual, como o presente no GrowCut clássico. Foi também desenvolvido uma etapa de geração automática de sementes, onde apenas pontos internos da região de interesse são selecionados, através do uso do método de otimização Evolução Diferencial. Além disso, foi desenvolvido um método de ajuste de parâmetros adaptativo, que a partir da extração de características da imagem ajusta os melhores parâmetros para o algoritmo. A abordagem desenvolvida é comparada qualitativamente e quantitativamente com técnicas de segmentação do estado da arte BEMD, BMCS, WAGA, Abordagem Topográfica e MCW, usando métricas relacionadas à forma das regiões segmentadas. As análises são avaliadas utilizando regiões de interesse da base IRMA, totalizando 1.165 mamogramas. Resultados mostram que o algoritmo proposto obteve melhores resultados, considerando similaridade com imagem ouro, para as métricas utilizadas. Para validar a proposta , foi construído um classificador de imagem usando o Perceptron Multicamadas clássico. Resultados mostraram que a técnica proposta obteve taxa de classificação de 94,77%, evidenciado a viabilidade do método proposto. / According to the World Health Organization, breast cancer is the most common cancer in women worldwide, becoming one of the most fatal types of cancer. Several studies show that the early diagnosis technologies can contribute to reduce the mortality rates and improve treatment options. Despite the existence of several imaging techniques to aid at the diagnosis of breast cancer, digital mammography is still the most used and effective imaging technology. Consequently, mammographic image segmentation is a fundamental task to support image diagnosis, considering shape analysis of mammary lesions and their borders. However, mammogram image segmentation is a hard task, once it is highly dependent on the types of mammary tissues. The GrowCut algorithm is a general-purpose segmentation method based on cellular automata, able to perform accurate segmentation through the adequate selection of internal and external points of the region of interest. Herein this work we present a new semi-supervised segmentation algorithm based on the modification of the GrowCut algorithm to perform semi-automatic mammographic image segmentation. In our proposal, we used a fuzzy Gaussian membership function to modify the evolution rule of the original GrowCut algorithm, in order to estimate the probabilities of a pixel being object or background. This model allows flexibility in the seeds initialization when compared to state of the art techniques, because the annotation executed by the specialist is used through the extraction of information of set os seeds, in opposite to the individual seeds information present in classical GrowCut .An automatic seed generation step was developed, where only the seeds internal to the region of interest are selected, using the Differential Evolution algorithm. Furthermore, we developed an adaptive parameter tuning method, which from the image characteristics it find the best parameters to the algorithm. The proposed approach was qualitatively and quantitatively compared with other state-of-the-art segmentation techniques BEMD, BMCS, WAGA, Topographic Approach and MCW, using metrics related to the shape of segmented regions. The analysis are evaluated using regions of interest from IRMA database, totaling 1.165 mammograms. Results show that the proposed algorithm achieved better results, considering similarity with ground truth, for the used metrics. In order to validate our proposal we built an image classifier using a classical Multilayer Perceptron. This analysis employed 1.165 mammograms from IRMA breast cancer database Results show that the proposed technique could achieve a classification rate of 94.77%, evidencing the feasibility of our approach.

<p>  </p> <p>This research is motivated by the potential of autonomous mobile robots (AMRs) in enhancing safety, productivity, and efficiency in the construction industry. The dynamic and complex nature of construction sites presents significant challenges to AMRs, particularly in localization and mapping – a process where AMRs determine their own position in the environment while creating a map of the surrounding area. These capabilities are crucial for autonomous navigation and task execution but are inadequately addressed by existing solutions, which primarily rely on visual Simultaneous Localization and Mapping (SLAM) methods. These methods are often ineffective in construction sites due to their underlying assumption of a static environment, leading to unreliable outcomes. Therefore, there is a pressing need to enhance the applicability of AMRs in construction by addressing the limitations of current localization and mapping methods in addressing the dynamic nature of construction sites, thereby empowering AMRs to function more effectively and fully realize their potential in the construction industry.</p> <p>The overarching goal of this research is to fulfill this critical need by developing a novel visual SLAM framework that is capable of not only detecting and segmenting diverse dynamic objects in construction environments but also effectively interpreting the semantic structure of the environment. Furthermore, it can efficiently integrate these functionalities into a unified system to provide an improved SLAM solution for dynamic, complex, and unstructured environments. The rationale is that such a SLAM system could effectively address the dynamic nature of construction sites, thereby significantly improving the efficiency and accuracy of robot localization and mapping in the construction working environment. </p> <p>Towards this goal, three specific objectives have been formulated. The first objective is to develop a novel methodology for comprehensive dynamic object segmentation that can support visual SLAM within highly variable construction environments. This novel method integrates class-agnostic objectness masks and motion cues into video object segmentation, thereby significantly improving the identification and segmentation of dynamic objects within construction sites. These dynamic objects present a significant challenge to the reliable operation of AMRs and, by accurately identifying and segmenting them, the accuracy and reliability of SLAM-based localization is expected to greatly improve. The key to this innovative approach involves a four-stage method for dynamic object segmentation, including objectness mask generation, motion saliency estimation, fusion of objectness masks and motion saliency, and bi-directional propagation of the fused mask. Experimental results show that the proposed method achieves a highest of 6.4% improvement for dynamic object segmentation than state-of-the-art methods, as well as lowest localization errors when integrated into visual SLAM system over public dataset. </p> <p>The second objective focuses on developing a flexible, cost-effective method for semantic segmentation of construction images of structural elements. This method harnesses the power of image-level labels and Building Information Modeling (BIM) object data to replace the traditional and often labor-intensive pixel-level annotations. The hypothesis for this objective is that by fusing image-level labels with BIM-derived object information, a segmentation that is competitive with pixel-level annotations while drastically reducing the associated cost and labor intensity can be achieved. The research method involves initializing object location, extracting object information, and incorporating location priors. Extensive experiments indicate the proposed method with simple image-level labels achieves competitive results with the full pixel-level supervisions, but completely remove the need for laborious and expensive pixel-level annotations when adapting networks to unseen environments. </p> <p>The third objective aims to create an efficient integration of dynamic object segmentation and semantic interpretation within a unified visual SLAM framework. It is proposed that a more efficient dynamic object segmentation with adaptively selected frames combined with the leveraging of a semantic floorplan from an as-built BIM would speed up the removal of dynamic objects and enhance localization while reducing the frequency of scene segmentation. The technical approach to achieving this objective is through two major modifications to the classic visual SLAM system: adaptive dynamic object segmentation, and semantic-based feature reliability update. Upon the accomplishment of this objective, an efficient framework is developed that seamlessly integrates dynamic object segmentation and semantic interpretation into a visual SLAM framework. Experiments demonstrate the proposed framework achieves competitive performance over the testing scenarios, with processing time almost halved than the counterpart dynamic SLAM algorithms.</p> <p>In conclusion, this research contributes significantly to the adoption of AMRs in construction by tailoring a visual SLAM framework specifically for dynamic construction sites. Through the integration of dynamic object segmentation and semantic interpretation, it enhances localization accuracy, mapping efficiency, and overall SLAM performance. With broader implications of visual SLAM algorithms such as site inspection in dangerous zones, progress monitoring, and material transportation, the study promises to advance AMR capabilities, marking a significant step towards a new era in construction automation.</p>