In this work, we propose a depth estimation system based on image sequence and odometry information. The key idea is that depth estimation is decoupled from pose estimation. Such approach results in multipurpose system applicable on different robot platforms and for different depth estimation related problems. Our implementation uses various filtration techniques, operates real-time and provides appropriate results. Although the system was aimed at and tested on drone platform, it can be well used on any other type of autonomous vehicle that provides odometry information and video output.

The problem of estimating depth from two or more images is a fundamental problem in computer vision, which is commonly referred as to stereo matching. The applications of stereo matching range from 3D reconstruction to autonomous robot navigation. Stereo matching is particularly attractive for applications in real life because of its simplicity and low cost, especially compared to costly laser range finders/scanners, such as for the case of 3D reconstruction. However, stereo matching has its very unique problems like convergence issues in the optimisation methods, and challenges to find matches accurately due to changes in lighting conditions, occluded areas, noisy images, etc. It is precisely because of these challenges that stereo matching continues to be a very active field of research. In this thesis we develop a binocular stereo matching algorithm that works with rectified images (i.e. scan lines in two images are aligned) to find a real valued displacement (i.e. disparity) that best matches two pixels. To accomplish this our research has developed techniques to efficiently explore a 3D space, compare potential matches, and an inference algorithm to assign the optimal disparity to each pixel in the image. The proposed approach is also extended to the trinocular case. In particular, the trinocular extension deals with a binocular set of images captured at the same time and a third image displaced in time. This approach is referred as to t +1 trinocular stereo matching, and poses the challenge of recovering camera motion, which is addressed by a novel technique we call baseline recovery. We have extensively validated our binocular and trinocular algorithms using the well known KITTI and Middlebury data sets. The performance of our algorithms is consistent across different data sets, and its performance is among the top performers in the KITTI and Middlebury datasets.

This report covers the implementation and evaluation of a stereo vision corre- spondence-based depth estimation algorithm on a GPU. The results and feed- back are used for a Multi-view camera system in combination with Jetson TK1 devices for parallelized image processing and the aim of this system is to esti- mate the depth of the scenery in front of it. The performance of the algorithm plays the key role. Alongside the implementation, the objective of this study is to investigate the advantages of parallel acceleration inter alia the differences to the execution on a CPU which are significant for all the function, the imposed overheads particular for a GPU application like memory transfer from the CPU to the GPU and vice versa as well as the challenges for real-time and concurrent execution. The study has been conducted with the aid of CUDA on three NVIDIA GPUs with different characteristics and with the aid of knowledge gained through extensive literature study about different depth estimation algo- rithms but also stereo vision and correspondence as well as CUDA in general. Using the full set of components of the algorithm and expecting (near) real-time execution is utopic in this setup and implementation, the slowing factors are in- ter alia the semi-global matching. Investigating alternatives shows that results for disparity maps of a certain accuracy are also achieved by local methods like the Hamming Distance alone and by a filter that refines the results. Further- more, it is demonstrated that the kernel launch configuration and the usage of GPU memory types like shared memory is crucial for GPU implementations and has an impact on the performance of the algorithm. Just concurrency proves to be a more complicated task, especially in the desired way of realization. For the future work and refinement of the algorithm it is therefore recommended to invest more time into further optimization possibilities in regards of shared memory and into integrating the algorithm into the actual pipeline.

MEDEIROS, Ailana Linhares de Sousa. Avaliação da Política de Cooperação Sul-Sul na UNILAB: percepções da integração sob o olhar dos discentes estrangeiros dos campi do Ceará. 2017. 114f. - Dissertação (Mestrado)- Universidade Federal do Ceará, Programa de Pós-graduação em Avaliação de Políticas Públicas, Fortaleza (CE), 2017. / Submitted by Gustavo Daher (gdaherufc@hotmail.com) on 2017-11-17T19:22:38Z No. of bitstreams: 1 2017_dis_alsmedeiros.pdf: 785682 bytes, checksum: d1627de6a06f8a2cbd4366c294db2427 (MD5) / Approved for entry into archive by Márcia Araújo (marcia_m_bezerra@yahoo.com.br) on 2017-11-20T17:58:02Z (GMT) No. of bitstreams: 1 2017_dis_alsmedeiros.pdf: 785682 bytes, checksum: d1627de6a06f8a2cbd4366c294db2427 (MD5) / Made available in DSpace on 2017-11-20T17:58:02Z (GMT). No. of bitstreams: 1 2017_dis_alsmedeiros.pdf: 785682 bytes, checksum: d1627de6a06f8a2cbd4366c294db2427 (MD5) Previous issue date: 2017 / n this work we evaluate the South-South Cooperation policy of the University of International Integration of Afro-Brazilian Lusophony (UNILAB), from an in-depth perspective (GUSSI, 2008; RODRIGUES, 2008). The assessment is based on the perception of foreign students, who are members of UNILAB. Our main goal is to analyze the South-South Cooperation in practice, within the institutional-academic context. As a methodological and qualitative approach we employ: bibliographic analysis, participant observation and interviews. We discuss the circumstances of the Foreign Brazilian Policy that launched the international educational cooperation, which is linked to such contexts of the foundation of UNILAB. They are the conceptual basis that provide the theoretical-ideological support. UNILAB is born out of the policy of expansion of higher education proposed in the Lula Government (2003-2010) within internalization and internationalization policy. Its main activities focus on the training of human resources that contribute to the integration of Brazil and the nations that make up the Community of Portuguese-speaking countries. From the speech of the individuals inserted in politics, we consider evaluative dimensions that allow us to observe the integration: the trajectory from survival to civil war to higher education; the (un)common Portuguese language; the (dis)knowledge about UNILAB, the city of Redenção, the personal formation and the past; the feeling of gratitude;being in a Brazilian university and, finally, the perceptions about an integration. I realized that, in conclusion, the educational cooperation proposed by Unilab based on South-South Cooperation represents a break of the North-South hegemonic ideology because, even though many adversities arise in the relationships between the students and the local community, the goal of qualify human resources to contribute in their countries of origin is achieved. Moreover, within the context of integration, there are experiences that form and complement life histories which compose the continuous construction of the unilabian institutional trajectory and they end up externalizing, in practice, how the integration in the cooperation happens. / Este estudo trata da avaliação da política de Cooperação Sul-Sul da Universidade da Integração Internacional da Lusofonia Afro-Brasileira, sob a perspectiva da avaliação em profundidade (GUSSI, 2008; RODRIGUES, 2008). Avalia a integração proposta como missão institucional a partir da percepção dos discentes estrangeiros que fazem parte da universidade. O objetivo é avaliar a Cooperação Sul-Sul na prática, dentro do contexto institucional-acadêmico. A abordagem metodológica é qualitativa e adotam-se como procedimentos: a análise bibliográfica, observação participante e entrevistas. Discute-se sobre os contextos da Política Externa Brasileira que propiciou a construção da cooperação educacional internacional, vinculando-se tais contextos à análise da criação da UNILAB, ou seja, a base conceitual que lhe proporciona sustentação teórico-ideológica. A UNILAB nasce a partir da política de expansão do ensino superior proposta no Governo Lula (2003-2010), em um contexto de interiorização e internacionalização, com foco na formação de recursos humanos que contribuam com a integração do Brasil e os países (as nações) que compõem a Comunidade dos Países de Língua Portuguesa. A partir da fala dos sujeitos inseridos na política, identificam-se dimensões avaliativas que permitiram observar a integração, a partir de algumas dimensões que se apresentaram na pesquisa avaliativa: da sobrevivência à guerra ao ensino superior no Timor; a língua portuguesa (in)comum; o (des)conhecimento acerca da UNILAB, a cidade de Redenção, a formação pessoal e o passado; o sentimento de gratidão; o estar em uma universidade no Brasil e, por fim, as percepções sobre a integração. Percebi, finalizando, que a cooperação educacional proposta pela Unilab com base na Cooperação Sul-Sul representa quebra da ideologia hegemônica Norte-Sul, uma vez que embora inúmeras adversidades se apresentem entre as relações de alunos e comunidade local, o fim de qualificar recursos humanos para contribuírem nos seus países de origem é atingido. Além disso, dentro da acepção de integração percorrem experiências que formam e complementam histórias de vida que fazem parte da formação do desenho contínuo da trajetória institucional unilabiana e acabam externalizando, na prática, como acontece a integração na cooperação.

Modelos geométricos desempenham um papel fundamental em divérsas áreas, desde a indústria do entretenimento até aplicações científicas. Para reduzir o elevado custo de criação de um modelo 3D, a reutilização de modelos existentes é a solução ideal. Recuperação de modelos 3D utilizam técnicas baseadas em conteúdo (do inglês CBR) que auxiliam a busca de modelos desejados em repositórios massivos, muitos disponíveis publicamente na Internet. Pontos principais para técnicas CBR eficientes e eficazes são descritores de forma que capturam com precisão as características de uma forma 3D e são capazes de discriminar entre diferentes formas. Nós apresentamos um descritor com base na distribuição de duas características globais, extraídas de uma forma 3D, depth complexity e thickness, que, respectivamente, capturam aspectos da topologia e da geometria das formas 3D. O descritor final, chamado DCT (depth complexity and thickness histogram), é um histograma 2D invariante a translações, rotações e escalas das formas geométricas. Nós eficientemente implementamos o DCT na GPU, permitindo sua utilização em consultas em tempo real em grandes bases de dados de modelos 3D. Nós validamos o DCT com as Princeton e Toyohashi Forma Benchmarks, contendo 1815 e 10000 modelos respectivamente. Os resultados mostram que DCT pode discriminar classes significativas desses benchmarks, é rápido e robusto contra transformações de forma e diferentes níveis de subdivisão e suavidade dos modelos. / Geometric models play a vital role in several fields, from the entertainment industry to scientific applications. To reduce the high cost of model creation, reusing existing models is the solution of choice. Model reuse is supported by content-based shape retrieval (CBR) techniques that help finding the desired models in massive repositories, many publicly available on the Internet. Key to efficient and effective CBR techniques are shape descriptors that accurately capture the characteristics of a shape and are able to discriminate between different shapes. We present a descriptor based on the distribution of two global features measured on a 3D shape, depth complexity and thickness, which respectively capture aspects of the geometry and topology of 3D shapes. The final descriptor, called DCT (depth complexity and thickness histogram), is a 2D histogram that is invariant to the translation, rotation and scale of geometric shapes. We efficiently implement the DCT on the GPU, allowing its use in real-time queries of large model databases. We validate the DCT with the Princeton and Toyohashi Shape Benchmarks, containing 1815 and 10000 models respectively. Results show that DCT can discriminate meaningful classes of these benchmarks, and is fast to compute and robust against shape transformations and different levels of subdivision and smoothness.

Invasive species are cited as a major threat to native community composition and biodiversity throughout the world. Some recent studies have focused on whether invasive species are the drivers or passengers of change in degraded systems. A strongly-interactive community is supposed to resist invasion by all but the most strong invasive competitors (`drivers') which can establish there without the aid of disturbance and actively reduce the abundance of natives. A weakly-interactive community, impaired by some anthropogenic disruption, is invasible by weaker exotics which are merely `passengers' of the habitat degradation that is more constraining to natives. Though competitive and disturbance-adapted species fit into this model, there is no correlate for species with superior toleration of stress. Systems with high degrees of natural abiotic stress are weakly-interactive and as a result may be invasible by an exotic `tolerator' in the same way that anthropogenically-weak systems are invasible by passengers. Dolomite prairie, differentiated from typic tallgrass prairie by its shallow soils, represents a relatively stressful system. A study of its plant community composition and relationships to environmental variables was done to get a better understanding of the natural abiotic drivers of composition. Compositional patterns were most closely associated with the soil depth gradient. Exotic Poa species were by far the most frequent invaders, a finding more typical of Great Lakes alvars than of typic tallgrass prairie of which the dolomite prairie is a subset. I hypothesized that Poa species dominated dolomite prairie via the tolerator model. In a series of removal treatments, I determined that Poa is neither a driver nor a tolerator, but a passenger of environmental degradation. My results suggest that historic anthropogenic degradation rather than specific competitive ability is a common explanation for exotic dominance on the landscape today. More empirical work needs to be conducted in other stressful (particularly relatively undisturbed) systems to further investigate the tolerator model.

Dans la thèse on introduit et on étudie une généralisation spatiale sur $\R^d$ du quantile réel usuel sous la forme d'une surface quantile via des formes $\phi$ et d'un point d'observation $O$. Notre point de départ est de simplement admettre la subjectivité due à l'absence de relation d'ordre totale dans $\R^d$ et donc de développer une vision locale et directionnelle des données. Ainsi, les observations seront ordonnées du point de vue d'un observateur se trouvant à un point $O \in \R^d$. Dans le chapitre 2, on introduit la notion du quantile vue d'un observateur $O$ dans la direction $u \in \Sd$ et de niveau $\alpha$ via des des demi-espaces orthogonaux à chaque direction d'observation. Ce choix de classe implique que les résultats de convergence ne dépendent pas du choix de $O$. Sous des hypothèses minimales de régularité, l'ensemble des points quantile vue de $O$ définit une surface fermée. Sous hypothèses minimales, on établit pour les surfaces quantile empiriques associées les théorèmes limites uniformément en le niveau de quantile et la direction d'observation, avec vitesses asymptotiques et bornes d'approximation non-asymptotiques. Principalement la LGNU, la LLI, le TCLU, le principe d'invariance fort uniforme puis enfin l'approximation du type Bahadur-Kiefer uniforme, et avec vitesse d'approximation. Dans le chapitre 3, on étend les résultats du chapitre précédent au cas où les formes $\phi$ sont prises dans une classe plus générale (fonctions, surfaces, projections géodésiques, etc) que des demi-espaces qui correspondent à des projections orthogonales par direction. Dans ce cadre plus général, les résultats dépendent fortement du choix de $O$, et c'est ce qui permet de tirer des interprétations statistiques. Dans le chapitre 4, des conséquences méthodologiques en statistique inférentielle sont tirées. Tout d'abord on introduit une nouvelle notion de champ de profondeurs directionnelles baptisée champ d'altitude. Ensuite, on définit une notion de distance entre lois de probabilité, basée sur la comparaison des deux collections de surfaces quantile du type Gini-Lorrentz. La convergence avec vitesse des mesures empiriques pour cette distance quantile, permet de construire différents tests en contrôlant leurs niveaux et leurs puissances. Enfin, on donne une version des résultats dans le cas où une information auxiliaire est disponible sur une ou plusieurs coordonnées sous la forme de la connaissance exacte de la loi sur une partition finie. / The main issue of the thesis is the development of spatial generalizations on $\R^d$ of the usual real quantile. Facing the usual fact that $\R^d$ is not naturally ordered, our idea is to simply admit subjectivity and thus to define a local viewpoint rather than a global one, anchored at some point of reference $O$ and arbitrary shape $\phi$ with the motivation of crossing information gathered by changing viewpoint $O$, shape $\phi$ and $\alpha$-th order of quantile. In Chapter 2, we study the spatial quantile points seen from an observer $O$ in a direction $u \in \Sd$ of level $\alpha$ through the class of the half-spaces orthogonal to the direction $u$. This choice implies that the convergence theorems do not depend on the choice of $O$. Under minimal regularity assumptions, the set of all quantile points seen from $O$ is a closed surface. Under minimal assumptions, we establish for the associated empirical quantile surfaces the convergence theorems uniformly on the quantile level and the observation direction with the asymptotic speed and non-asymptotic bounds of approximation. Mainly, we establish the ULLN, the ULIL, the UCLT, the uniform strong invariance principle and finally the Bahadur-Kiefer type embedding, with the approximation speed rate. In Chapter 3, all the results of the previous chapter are extended to the case where the shapes $ \phi $ are taken in a class more general (functions, surfaces, geodesic projections, etc) than orthogonal projections (half-spaces). In this general setting, the results depend strongly on the choice of $ O $. It is this dependence which permit to draw statistical interpretations: modes detection, mass localization, etc. In Chapter 4, some methodological consequences in inferential statistics are drawn. First we introduce a new concept of directional depth fields called altitude fields. In a second application is defined a new distances between probability distributions, based on the comparison of two collections of quantile surfaces, which are indexes of the type Gini-Lorrentz. The convergence with speed of the empirical quantile measures for these distances, can build different tests with control of their level and their power. A third use of the quantile surfaces is for the case where $ \alpha = 1/2$. Finally, we give a version of our theorems in the case where auxiliary information is available on one or more coordinates of the random variable. By assuming known the probability of the elements of a finite partition, the asymptotic variance of the limiting process decreases and the simulations with few points clearly shows the reframe of the estimated surfaces to the real ones.

No contexto de imagens geradas por computador, efeitos multi-fragmento são aqueles que determinam a cor do pixel baseados em informações computadas a partir de mais de um fragmento. Nesse tipo de efeito, a contribuição de cada fragmento é extraída de sua visibilidade com respeito a um determinado ponto de vista. Observando uma sequencia de fragmentos vista através de um pixel, a visibilidade de um fragmento depende da sua relação espacial com os demais fragmentos. Essa relação pode ser reduzida ao problema de ordenação de múltiplos fragmentos. Portanto, ordenação é essencial para correta avaliação de efeitos multi-fragmento. A pesquisa desta tese foca em dois problemas multi-fragmento clássicos: transparência independente de ordem e anti-aliasing de fragmentos transparentes. Enquanto o efeito de transparência necessita de ordenação de fragmentos ao longo do raio de visualização do pixel, anti-aliasing aumenta a complexidade do problema ao adicionar informação espacial do fragmento com respeito à área do pixel. A contribuição desta tese é o desenvolvimento de uma solução para visibilidade de fragmentos que pode tirar proveito do pipeline de transformação e iluminação, implementando nas GPUs de hoje. Nós descrevemos ambos os problemas de transparência e anti-aliasing, discutindo soluções anteriores, além de classificá-las e compará-las. Nossa análise associa soluções a implementações específicas, comparando uso de memória, desempenho e qualidade de imagem. Os documentos resultantes fornecem uma visão geral das áreas abordadas, contendo: qual é o estado-da-arte atualmente, o que ele é capaz de fazer e quais são suas limitações, ou seja, onde melhorias são possíveis. Como parte integrante desta tese, nós propomos duas novas técnicas para processar transparência independente de ordem. Nós mostramos como obter o menor consumo de memória para cálculo exato de transparência, em um número finito de passos de geometria; permitindo aumento da complexidade das cenas representadas e da resolução da imagem final, em relação aos métodos anteriores, dada uma determinada configuração de hardware. Adicionalmente, demonstramos que, para a maior parte dos casos, os fragmentos mais próximos ao observador tem maior impacto sobre a cor final do pixel. Também mostramos como esta perspectiva sobre o problema inspirou novas técnicas. A pesquisa também inclui a investigação de uma nova abordagem para anti-aliasing para fragmentos transparentes. Através do uso de uma única amostra por fragmento, nosso objetivo é reduzir o consumo de memória enquanto melhoramos desempenho e qualidade. Experimentos preliminares apresentam resultados promissores em comparação com a técnica mais usada para anti-aliasing. / Multi-fragment effects, in the computer-generated imagery context, are effects that determine pixel color based on information computed from more than one fragment. In such effects, the contribution of each fragment is extracted from its visibility with respect to a point of view. Seen through a pixel’s point of view, the visibility of one fragment depends on its spatial relationship with other fragments. This relationship can be reduced to the problem of sorting multiple fragments. Therefore, sorting is the key to multi-fragment evaluation. The research on this dissertation is focused on two classical multi-fragment effects: order-independent transparency and anti-aliasing of transparent fragments. While transparency rendering requires sorting of fragments along the view ray of a pixel, anti-aliasing increases the problem complexity by adding spatial information of fragments with respect to the pixel area. This dissertation contribution relies on the work towards the development of a solution for the visibility of fragments that can take advantage of the transformation and lighting pipeline implemented in current GPUs. We describe both transparency and aliasing problems, for which we discuss existing solutions, analyzing, classifying and comparing them. The analysis associates solutions to specific applications, comparing memory usage, performance, and quality. The result is a general view of each field: which are the current state-of-the-art capabilities and in which direction significant improvements can be made. As part of this dissertation, we proposed two novel techniques for order-independent transparency rendering. We show how to achieve the minimum memory footprint for computing exact transparency in a bounded number of geometry passes; allowing increasing scene complexity and image resolution to be feasible within current hardware capabilities. Additionally, we demonstrate that, for most scenarios, the front-most fragments have the greatest impact on the pixel color. We also show how the perspective we propose has inspired recent transparency techniques. The research includes the investigation of a novel anti-aliasing approach for transparent fragments. Through the use of a single sample per fragment, we aim at reducing memory footprint while improving performance and quality. Preliminary experiments show promising results, in comparison with a well established and largely used anti-aliasing technique.

The relation between zooplankton biomass and size structure and the trophic state and depth of shallow lake ecosystems was studied. Presented dataset describes monthly sampling within the experimental season 2011. Generally, total zooplankton biomass clearly correlated with high nutrient concentration mesocosms.

The aim of the thesis is to extend knowledge in terms of the application of modern seeding machinery for soil preparation for sowing in the cultivation of cereals. In the literary section of the report a brief overview of the use of seed drills for minimum tillage. Load small-plot experiment on the selected station and simultaneously perform evaluation activities seeders work to prepare the ground in terms of quality seed, seed depth, coming in plants (growth phase), or the occurrence of weeds. The thesis will brief overview of the analysis of operating costs. The results evaluate and propose recommendations for possible use in agricultural practice.