As áreas de visualização e modelagem baseados em pontos têm sido pesquisadas ativamente na computação gráfica. Pontos com atributos (por exemplo, normais) são geralmente chamados de surfels e existem vários algoritmos para a manipulação e visualização eficiente deles. Um ponto chave para a eficiência de muitos métodos é o uso de estruturas de particionamento do espaço. Geralmente octrees e KD-trees, por utilizarem cortes alinhados com os eixos são preferidas em vez das BSP-trees, mais genéricas. Neste trabalho, apresenta-se uma estrutura chamada Constrained BSP-tree (CBSP-tree), que pode ser vista como uma estrutura intermediárias entre KD-trees e BSP-trees. A CBSP-tree se caracteriza por permitir cortes arbitrários desde que seja satisfeito um critério de validade dos cortes. Esse critério pode ser redefinido de acordo com a aplicação. Isso permite uma aproximação melhor de regões curvas. Apresentam-se algoritmos para construir CBSP-trees, valendo-se da flexibilidade que a estrutura oferece, e para realizar operações booleanas usando uma nova classificação de interior/exterior.

A partir das dimensões dos indivíduos pode-se definir dimensionamentos adequados para os produtos e postos de trabalho, proporcionando segurança e conforto aos usuários. Com o avanço da tecnologia de digitalização de imagens (escaneamento) 3D, é possível tirar algumas medidas de maneira mais rápida e com a redução da presença do entrevistado durante o processo. No entanto, faltam estudos que avaliem estas tecnologias no Brasil, sendo necessária a realização de uma comparação das tecnologias e das respectivas precisões para que seu uso em pesquisas. Com o objetivo de oferecer métodos comparativos para escolha dos marcadores e equipamentos a serem utilizados em uma pesquisa antropométrica tridimensional da população brasileira, no presente estudo estão comparadas duas tecnologias de escaneamento: o sistema a laser WBX da empresa norte americana Cyberware e o sistema MHT da empresa russa Artec Group. O método para avaliação da precisão dimensional dos dados advindos desses equipamentos de digitalização de imagens 3D teve cinco etapas: Estudo dos processos de escaneamento; Escaneamento dos marcadores de pontos anatômicos; Escaneamento utilizando um corpo de prova cilíndrico; Escaneamento de um manequim; Escaneamento de um voluntário que teve seus pontos anatômicos marcados para a retirada de medidas. Foi feita uma comparação entre as medidas retiradas manualmente, por meio de antropômetro e virtualmente, com o auxílio do software de modelagem tridimensional Rhinoceros. Em relação aos resultados obtidos na avaliação do manequim e do voluntário, concluiu-se que a magnitude do erro absoluto é semelhante para ambos os scanners, e permanece constante independentemente das dimensões sob análise. As principais diferenças são em relação às funcionalidades dos equipamentos. / Only from the dimensions of individuals it is possible to define appropriate sizing for products and workplaces, providing security and comfort to users. With the evolution of 3D digital imaging technology (3D scanning), it is possible to take some measurements faster and reduce the need of the interviewee during the process. However, there are few studies that evaluate these technologies in Brazil. It is necessary to compare these equipments in order to know their precision so they can be used in researches. In order to choose anatomical markers and equipments, this study compares two different equipments: Cyberware WBX laser scanner and Artec Group MHT white light scanner. The method for assessing the dimensional accuracy of the data obtained from those scanning 3D imaging equipment had five steps: Study of the scanning processes; Scanning using a cylindrical object; Scanning a mannequin; Scanning a volunteer who had his anatomical points marked for taking measurements. The comparison was made between the measurements taken manually with an anthropometer and virtually using the 3D modeling software Rhinoceros. Based on results obtained in the evaluation of the mannequin and volunteer, it was concluded that the absolute error is similar for both scanners and remains constant regardless of the size under consideration. The main differences are the features of each equipment.

Tesis para optar al título de Realizador en Cine y Televisión / En un futuro distópico, una chica llamada Xanat encuentra un brazalete que contiene una inteligencia artificial llamada B­02. Ésta tecnología le permite explorar la mente de otras personas mientras duermen o se encuentran en estado inconsciente. Es el año 2256 y doscientos años antes el mundo era muy distinto al que es ahora, Xanat espera que con la habilidad de B­02 pueda encontrar alguna forma de conocer ese mundo y tener información sobre el porqué del presente. Sin embargo, no existen personas que recuerden ese mundo o que se dediquen a la indagación del pasado pues éste se ha prohibido de manera de no volver a cometer los mismos errores. No obstante, B­02 sabe que arriba de una torre hay un lugar en donde aún queda una esperanza para los planes de Xanat. En la torre más alta de la ciudad se encuentra una sala de criogenia en donde descansan personas que vivieron hace mucho tiempo. Un científico llamado Yael que custodia la sala, le ofrece a Xanat un trato: ella deberá entrar a las mentes de los congelados y extraerles sus recuerdos, de esta forma ambos conocerán sobre el mundo antiguo. Xanat acepta y entra a la mente de uno de los congelados descubriendo secretos del mundo antiguo. Esto provoca en ella un momento de reflexión donde se da cuenta de lo mal que está su mundo y comienza a ver una forma de repararlo. Para efectos de la obra de la obra de título, esta se centrará en el inicio de la historia y la primera visita a la mente de un congelado, Howard, por parte de Xanat con la ayuda de B­02.

Computer vision is a research field in which images are the main object of study. One of its category of problems is shape description. Object classification is one important example of applications using shape descriptors. Usually, these processes were performed on 2D images. With the large-scale development of new technologies and the affordable price of equipment that generates 3D images, computer vision has adapted to this new scenario, expanding the classic 2D methods to 3D. However, it is important to highlight that 2D methods are mostly dependent on the variation of illumination and color, while 3D sensors provide depth, structure/3D shape and topological information beyond color. Thus, different methods of shape descriptors and robust attributes extraction were studied, from which new attribute extraction methods have been proposed and described based on 3D data. The results obtained from well known public datasets have demonstrated their efficiency and that they compete with other state-of-the-art methods in this area: the RPHSD (a method proposed in this dissertation), achieved 85:4% of accuracy on the University of Washington RGB-D dataset, being the second best accuracy on this dataset; the COMSD (another proposed method) has achieved 82:3% of accuracy, standing at the seventh position in the rank; and the CNSD (another proposed method) at the ninth position. Also, the RPHSD and COMSD methods have relatively small processing complexity, so they achieve high accuracy with low computing time. / Visão computacional é uma área de pesquisa em que as imagens são o principal objeto de estudo. Um dos problemas abordados é o da descrição de formatos (em inglês, shapes). Classificação de objetos é um importante exemplo de aplicação que usa descritores de shapes. Classicamente, esses processos eram realizados em imagens 2D. Com o desenvolvimento em larga escala de novas tecnologias e o barateamento dos equipamentos que geram imagens 3D, a visão computacional se adaptou para este novo cenário, expandindo os métodos 2D clássicos para 3D. Entretanto, estes métodos são, majoritariamente, dependentes da variação de iluminação e de cor, enquanto os sensores 3D fornecem informações de profundidade, shape 3D e topologia, além da cor. Assim, foram estudados diferentes métodos de classificação de objetos e extração de atributos robustos, onde a partir destes são propostos e descritos novos métodos de extração de atributos a partir de dados 3D. Os resultados obtidos utilizando bases de dados 3D públicas conhecidas demonstraram a eficiência dos métodos propóstos e que os mesmos competem com outros métodos no estado-da-arte: o RPHSD (um dos métodos propostos) atingiu 85:4% de acurácia, sendo a segunda maior acurácia neste banco de dados; o COMSD (outro método proposto) atingiu 82:3% de acurácia, se posicionando na sétima posição do ranking; e o CNSD (outro método proposto) em nono lugar. Além disso, os métodos RPHSD têm uma complexidade de processamento relativamente baixa. Assim, eles atingem uma alta acurácia com um pequeno tempo de processamento.

As áreas de visualização e modelagem baseados em pontos têm sido pesquisadas ativamente na computação gráfica. Pontos com atributos (por exemplo, normais) são geralmente chamados de surfels e existem vários algoritmos para a manipulação e visualização eficiente deles. Um ponto chave para a eficiência de muitos métodos é o uso de estruturas de particionamento do espaço. Geralmente octrees e KD-trees, por utilizarem cortes alinhados com os eixos são preferidas em vez das BSP-trees, mais genéricas. Neste trabalho, apresenta-se uma estrutura chamada Constrained BSP-tree (CBSP-tree), que pode ser vista como uma estrutura intermediárias entre KD-trees e BSP-trees. A CBSP-tree se caracteriza por permitir cortes arbitrários desde que seja satisfeito um critério de validade dos cortes. Esse critério pode ser redefinido de acordo com a aplicação. Isso permite uma aproximação melhor de regões curvas. Apresentam-se algoritmos para construir CBSP-trees, valendo-se da flexibilidade que a estrutura oferece, e para realizar operações booleanas usando uma nova classificação de interior/exterior.

Binocular eye-gaze tracking can be used to estimate the point-of-gaze (POG) of a subject in real-world three-dimensional (3D) space using the vergence of the eyes. In this thesis, a novel non-contact, model-based technique for 3D POG estimation is presented. The non-contact system allows people to select real-world objects in 3D physical space using their eyes, without the need for head-mounted equipment. Using a model-based POG estimation algorithm allows for free head motion and a single stage of calibration. The users were free to naturally move and reorient their heads while operating the system, within an allowable headspace of 3.2 x 9.2 x 14 cm. A rela tively high precision, as measured by the standard deviation of the 3D POG estimates, was measured to be 0.26 cm and was achieved with the use of high speed sampling and digital filtering techniques. When observing points in a 3D volume, large head and eye rotations are far more common than when observing a 2D screen. A novel corneal reflection pattern matching algorithm is presented for increasing image feature tracking reliability in the presence of large eye rotations. It is shown that an average accuracy of 3.93 cm was achieved over seven different subjects and a workspace volume of 30 x 23 x 25 cm (width x height x depth). An example application is presented illustrating the use of the 3D POG as a human computer interface in a 3D game of Tic-Tac-Toe on a 3 x 3 x 3 volumetric display. / Applied Science, Faculty of / Electrical and Computer Engineering, Department of / Graduate

Recent advancement in cameras and image processing technology has generated a paradigm shift from traditional 2D and 3D video to Multi-view Video (MVV) technology, while at the same time improving video quality and compression through standards such as High Efficiency video Coding (HEVC). In multi-view, cameras are placed in predetermined positions to capture the video from various views. Delivering such views with high quality over the Internet is a challenging prospect, as MVV traffic is several times larger than traditional video since it consists of multiple video sequences each captured from a different angle, requiring more bandwidth than single view video to transmit MVV. Also, the Internet is known to be prone to packet loss, delay, and bandwidth variation, which adversely affects MVV transmission. Another challenge is that end users’ devices have different capabilities in terms of computing power, display, and access link capacity, requiring MVV to be adapted to each user’s context. In this paper, we propose an HEVC Multi-View system using Dynamic Adaptive Streaming over HTTP (DASH) to overcome the above mentioned challenges. Our system uses an adaptive mechanism to adjust the video bitrate to the variations of bandwidth in best effort networks. We also propose a novel scalable way for the Multi-view video and Depth (MVD) content for 3D video in terms of the number of transmitted views. Our objective measurements show that our method of transmitting MVV content can maximize the perceptual quality of virtual views after the rendering and hence increase the user’s quality of experience.

In recent years, deep convolutional neural networks (CNNs) have shown remarkable results in the image domain. However, most of the neural networks in action recognition do not have very deep layer compared with the CNN in the image domain. This thesis presents a 3D Densely Connected Convolutional Network (3D-DenseNet) for action recognition that can have more than 100 layers without exhibiting performance degradation or overfitting. Our network expands Densely Connected Convolutional Networks (DenseNet) [32] to 3D-DenseNet by adding the temporal dimension to all internal convolution and pooling layers. The internal layers of our model are connected with each other in a feed-forward fashion. In each layer, the feature-maps of all preceding layers are concatenated along the last dimension and are used as inputs to all subsequent layers. We propose two different versions of 3D-DenseNets: general 3D-DenseNet and lite 3D-DenseNet. While general 3D-DenseNet has the same architecture as DenseNet, lite 3D-DenseNet adds a 3D pooling layer right after the first 3D convolution layer of general 3D-DenseNet to reduce the number of training parameters at the beginning so that we can reach a deeper network. We test on two action datasets: the MERL shopping dataset [69] and the KTH dataset [63]. Our experiment results demonstrate that our method performs better than the state-of-the-art action recognition method on the MERL shopping dataset and achieves a competitive result on the KTH dataset.

La planification chirurgicale repose sur l’anatomie du patient, et repose souvent sur l’analyse d’images médicales acquises avant la chirurgie. En particulier, c’est le cas pour les interventions de chirurgie pelvienne en pédiatrie, pour de nombreuses pathologies telles que des tumeurs et des malformations. Dans cette zone anatomique, en raison de sa forte vascularisation et innervation, une bonne planification chirurgicale est extrêmement importante pour éviter des lésions fonctionnelles des organes du patient, qui pourraient nuire a sa qualité de vie. En pratique clinique, la procédure standard repose sur l’analyse visuelle, coupe par coupe, des images de la région pelvienne. Cette tâche, même si elle est facilement accomplie par des radiologues experts, est très complexe et fastidieuse pour les chirurgiens, en raison de la complexité et de la variabilité des structures anatomiques et, par conséquent, de leurs images. De plus, en raison des variations anatomiques selon l’age du patient, toutes ces difficultés sont accentuées en pédiatrie et une compréhension anatomique claire est encore plus importante que pour les adultes. Pour ces raisons, il est important et utile d’être capable de fournir aux chirurgiens des modèles anatomiques 3D spécifiques aux patients, obtenus par traitement et analyse des images IRM.Dans cette thèse, nous proposons un ensemble de méthodes de segmentation d’images IRM de patients pédiatriques. Nous nous concentrons sur trois structures pelviennes importantes : les os du bassin, les vaisseaux sanguins et la vessie. Pour les os, nous proposons une méthode semi-automatique comportant une première étape de recalage de modèles osseux puis une étape de segmentation fine par modèles déformables. La principale contribution de la méthode proposée est l’introduction d’un ensemble de modèles osseux pour différentes tranches d’age, ce qui permet de prendre en compte la variabilité des os pendant la croissance. Pour les vaisseaux, nous proposons une méthode par patchs, apprentissage profond et transfert d’apprentissage, donc ne nécessitant que peu de donnes d’apprentissage. La principale contribution de ce travail est la conception d’une procédure semi-automatique pour l’extraction des patchs, qui permet a l’utilisateur de se focaliser uniquement sur les vaisseaux d’intérêt; et pour la planification chirurgicale. Pour la segmentation de la vessie, nous proposons d’utiliser une approche par modèles déformables, particulièrement robuste aux hétérogénéités de l’image et aux effets de volume partiel, souvent présents dans les images IRM pédiatriques. Toutes les méthodes proposées sont intégrées dans une plateforme logicielle libre pour le traitement d’images médicales, donnant aux chirurgiens des outils performants avec des interfacesutilisateur faciles a utiliser. De plus, nous mettons en place une stratégie de traitement et de portabilité pour la visualisation des modèles 3D du patient, permettant aux chirurgiens de générer, visualiser et partager ces modèles au sein de l’hôpital. En conclusion, les résultats obtenus avec les méthodes proposées sont quantitativement et qualitativement évalués de manière très positive par des chirurgiens pédiatriques, démontrant leurs potentialités pour l’utilisation en pratique clinique dans des procédures de planification chirurgicale. / Surgical planning relies on the patient’s anatomy, and it is often based on medical images acquired before the surgery. This is in particular the case for pelvic surgery on children, for various indications such as malformations or tumors. In this particular anatomical region, due to its high vascularization and innervation, a good surgical planning is extremely important to avoid potential functional damages to the patient’s organs that could strongly affect their quality of life. In clinical practice the standard procedure is still to visually analyze, slice by slice, the images of the pelvic region. This task, even if quite easily performed by the expert radiologists, is difficult and tedious for the surgeons due to the complexity and variability of the anatomical structures and hence their images. Moreover, due to specific anatomy depending on the age of the patient, all the difficulties of the surgical planning are emphasized in the case of children, and a clear anatomical understanding is even more important than for the adults. For these reasons, it is very important and challenging to provide the surgeons with patient-specific 3D reconstructions, obtained from the segmentation of MRI images. In this work we propose a set of segmentation tools for pelvic MRI images of pediatric patients. In particular, we focus on three important pelvic structures: the pelvic bones, the pelvic vessels and the urinary bladder. For pelvic bones, we propose a semi-automatic approach based on template registration and deformable models. The main contribution of the proposed method is the introduction of a set of bones templates for different age ranges, which allows us to take into account the bones variability during growth. For vessels segmentation, we propose a patch-based deep learning approach using transfer learning, thus requiring few training data. The main contribution of this work is the design of a semi-automatic strategy for patches extraction, which allows the user to focus only on the vessels of interest for surgical planning. For bladder segmentation, we propose to use a deformable model approach that is particularly robust to image inhomogeneities and partial volume effects, which are often present in pediatric MRI images. All the developed segmentation methods are integrated in an open-source platform for medical imaging, delivering powerful tools and user-friendly GUIs to the surgeons. Furthermore, we set up a processing and portability workflow for visualization of the 3D patient specific models, allowing surgeons to generate, visualize and share within the hospital the patient specific 3D models. Finally, the results obtained with the proposed methods are quantitatively and qualitatively evaluated by pediatric surgeons, which demonstrates their potentials for clinical use in surgical planning procedures.

The aim of the submitted semestral project is research of 3D scanning object methods. The 3D scans utilization and types of these scans have been mentioned in the first chapter. The scanners are grouped into two basic methods – contact and contactless. The second chapter is dedicated to 3D model measurement principles, which are described in detail in four sub-chapters. In the following chapter is described design with complete realization 3D scanner with the implementation of the application for its controlling. Last part of the thesis concerns the 3D model editing and method evaluating.