[pt] Documentos digitalizados em ambientes de negócios substituíram grandes volumes de papéis. Profissionais autorizados usam carimbos para certificar informações críticas nesses documentos. Muitas empresas precisam verificar o carimbo adequado de documentos de entrada e saída. Na maioria das situações de inspeção, as pessoas realizam inspeção visual para identificar carimbos. Assim sendo, a verificação manual de carimbos é cansativa, suscetível a erros e ineficiente em termos de tempo gasto e resultados esperados. Erros na verificação manual de carimbos podem gerar multas de órgãos reguladores, interrupção de operações e até mesmo comprometer fluxos de trabalho e transações financeiras. Este trabalho propõe dois métodos que combinados podem resolver esse problema, automatizando totalmente a detecção de carimbos em documentos digitalizados do mundo real. Os métodos desenvolvidos podem lidar com conjuntos de dados contendo muitos tipos de carimbos de tamanho de amostra pequena, com múltiplas sobreposições, combinações diferentes por página e dados ausentes. O primeiro método propõe uma arquitetura de rede profunda projetada a partir da relação entre os problemas identificados em carimbos do mundo real e os desafios e soluções da tarefa de detecção de objetos apontados na literatura. O segundo método propõe um novo pipeline de aumento de instâncias de conjuntos de dados de carimbos a partir de dados reais e investiga se é possível detectar tipos de carimbos com amostras insuficientes. Este trabalho avalia os hiperparâmetros da abordagem de aumento de instâncias e os resultados obtidos usando um método Deep Explainability. Foram alcançados resultados de última geração para a tarefa de detecção de carimbos combinando com sucesso esses dois métodos, alcançando 97.3 por cento de precisão e 93.2 por cento de recall. / [en] Scanned documents in business environments have replaced large volumes of papers. Authorized professionals use stamps to certify critical information in these documents. Many companies need to verify the adequate stamping of incoming and outgoing documents. In most inspection situations, people perform a visual inspection to identify stamps. Therefore, manual stamp checking is tiring, susceptible to errors, and inefficient in terms of time spent and expected results. Errors in manual checking for stamps can lead to fines from regulatory bodies, interruption of operations, and even compromise workflows and financial transactions. This work proposes two methods that combined can address this problem, by fully automating stamp detection in real-world scanned documents. The developed methods can handle datasets containing many small sample-sized types of stamps, multiples overlaps, different combinations per page, and missing data. The first method proposes a deep network architecture designed from the relationship between the problems identified in real-world stamps and the challenges and solutions of the object detection task pointed out in the literature. The second method proposes a novel instance augmentation pipeline of stamp datasets from real data to investigate whether it is possible to detect stamp types with insufficient samples. We evaluate the hyperparameters of the instance augmentation approach and the obtained results through a Deep Explainability method. We achieve state-of-the-art results for the stamp detection task by successfully combining these two methods, achieving 97.3 percent of precision and 93.2 percent of recall.

Nesta tese, buscou-se avaliar se a tomografia computadorizada de duplaenergia pós-contraste (TCDE) é capaz de detectar diferenças regionais da perfusão pulmonar em um modelo animal suíno incluindo variações de decúbito, lesão alveolar e oclusão da artéria pulmonar com balão, comparando estes resultados com os obtidos pela perfusão de primeira passagem com a tomografia computadorizada dinâmica (TCD). Dez suínos landrace foram divididos em Grupos A (N = 5, controle) e B (N = 5). Animais do Grupo B foram submetidos ao protocolo de lesão alveolar induzida por ventilação mecânica (LPIV). O volume sanguíneo perfundido e o fluxo sanguíneo pulmonar foram, respectivamente, estimados pela TCDE (%VSPTCDE) e pela TCD (FSPTCD), em diversas condições experimentais: posição supina versus prona, presença versus ausência de LPIV, presença ou ausência de oclusão da artéria pulmonar. A correlação entre %VSPTCDE e FSPTCD foi moderada (R = 0,60) com ampla variabilidade (intervalo 0,35-0,91) entre animais. %VSPTCDE e FSPTCD demonstraram padrões similares de heterogeneidade da perfusão pulmonar nas diferentes condições experimentais. Entretanto, reduções do %VSPTCDE causadas pela oclusão com balão foram em média -29,32 %, enquanto reduções do FSPTCD foram em média -86,78 % (p < 0,001). Estimativas quantitativas do VSPTCDE tiveram um erro médio de +4.3 ml/100g em comparação com o FSPTCD, com limites de concordância de 95 % entre -16,6 ml/100g e 25,1 ml/100g. A TCDE póscontraste é capaz de prover estimativas semiquantitativas que refletem a heterogeneidade regional da perfusão pulmonar causada por mudanças de decúbito, lesão alveolar e oclusão da artéria pulmonar com balão, apresentando moderada correlação com a perfusão de primeira passagem pela TCD / We aimed to evaluate whether contrast-enhanced dual-energy CT (DECT) detects regional pulmonary perfusion changes in a swine model of acute lung injury, with variations in decubitus and transient occlusion of the pulmonary artery, comparing these results with those obtained with dynamic CT perfusion (DynCT). Ten landrace swine were assigned to Groups A (N = 5, control) and B (N = 5). Group B was subjected to ventilator-induced lung injury (VILI). Perfused blood volume and pulmonary blood flow were quantified by DECT (PBVDECT) and DynCT (PBFDynCT), respectively, under different settings: supine versus prone, and with/without balloon occlusion of a pulmonary artery (PA) branch. Correlation of regional PBVDECT versus PBFDynCT was moderate (R = 0.60) with high variability (range 0.35-0.91) among the animals. Regional pulmonary perfusion changes assessed by %PBVDECT agreed with PBFDynCT in response to decubitus changes, lung injury and balloon occlusion in the multivariate analysis. However, reductions in %PBVDECT caused by balloon occlusion were in average -29.32 %, whereas reductions in PBFDynCT were in average -86.78 % (p < 0.001). Quantitative estimates of PBVDECT had a mean bias of +4.3 ml/100g in comparison with PBVDynCT, with 95 % confidence intervals between -16.6 ml/100g and 25.1 ml/100g. Semiquantitative contrastenhanced DECT reflects regional changes in perfusion caused decubitus changes, acute lung injury, and balloon occlusion of the PA, with moderate correlation in comparison with DynCT

Nesta tese, buscou-se avaliar se a tomografia computadorizada de duplaenergia pós-contraste (TCDE) é capaz de detectar diferenças regionais da perfusão pulmonar em um modelo animal suíno incluindo variações de decúbito, lesão alveolar e oclusão da artéria pulmonar com balão, comparando estes resultados com os obtidos pela perfusão de primeira passagem com a tomografia computadorizada dinâmica (TCD). Dez suínos landrace foram divididos em Grupos A (N = 5, controle) e B (N = 5). Animais do Grupo B foram submetidos ao protocolo de lesão alveolar induzida por ventilação mecânica (LPIV). O volume sanguíneo perfundido e o fluxo sanguíneo pulmonar foram, respectivamente, estimados pela TCDE (%VSPTCDE) e pela TCD (FSPTCD), em diversas condições experimentais: posição supina versus prona, presença versus ausência de LPIV, presença ou ausência de oclusão da artéria pulmonar. A correlação entre %VSPTCDE e FSPTCD foi moderada (R = 0,60) com ampla variabilidade (intervalo 0,35-0,91) entre animais. %VSPTCDE e FSPTCD demonstraram padrões similares de heterogeneidade da perfusão pulmonar nas diferentes condições experimentais. Entretanto, reduções do %VSPTCDE causadas pela oclusão com balão foram em média -29,32 %, enquanto reduções do FSPTCD foram em média -86,78 % (p < 0,001). Estimativas quantitativas do VSPTCDE tiveram um erro médio de +4.3 ml/100g em comparação com o FSPTCD, com limites de concordância de 95 % entre -16,6 ml/100g e 25,1 ml/100g. A TCDE póscontraste é capaz de prover estimativas semiquantitativas que refletem a heterogeneidade regional da perfusão pulmonar causada por mudanças de decúbito, lesão alveolar e oclusão da artéria pulmonar com balão, apresentando moderada correlação com a perfusão de primeira passagem pela TCD / We aimed to evaluate whether contrast-enhanced dual-energy CT (DECT) detects regional pulmonary perfusion changes in a swine model of acute lung injury, with variations in decubitus and transient occlusion of the pulmonary artery, comparing these results with those obtained with dynamic CT perfusion (DynCT). Ten landrace swine were assigned to Groups A (N = 5, control) and B (N = 5). Group B was subjected to ventilator-induced lung injury (VILI). Perfused blood volume and pulmonary blood flow were quantified by DECT (PBVDECT) and DynCT (PBFDynCT), respectively, under different settings: supine versus prone, and with/without balloon occlusion of a pulmonary artery (PA) branch. Correlation of regional PBVDECT versus PBFDynCT was moderate (R = 0.60) with high variability (range 0.35-0.91) among the animals. Regional pulmonary perfusion changes assessed by %PBVDECT agreed with PBFDynCT in response to decubitus changes, lung injury and balloon occlusion in the multivariate analysis. However, reductions in %PBVDECT caused by balloon occlusion were in average -29.32 %, whereas reductions in PBFDynCT were in average -86.78 % (p < 0.001). Quantitative estimates of PBVDECT had a mean bias of +4.3 ml/100g in comparison with PBVDynCT, with 95 % confidence intervals between -16.6 ml/100g and 25.1 ml/100g. Semiquantitative contrastenhanced DECT reflects regional changes in perfusion caused decubitus changes, acute lung injury, and balloon occlusion of the PA, with moderate correlation in comparison with DynCT

The "Mark V" Holographic Video Monitor ("HoloMonitor") is a continuation of effort and accomplishments to produce a low-cost device capable of reproducing true full-color horizontal-parallax-only computer-generated holograms at typical video frame-rates. While other devices around the world may have greater capabilities, these devices are currently confined to laboratory settings due to their sheer complexity and expense. The aim of this project is to provide researchers and "tinkerers" with a device capable of recreating holographic effects in full color, respectable resolution, in real time, and at a comparatively low cost. The "Mark V" HoloMonitor is a closer representation of a consumer product than any other device of the MIT/BYU series of HoloMonitors. In this thesis, I discuss the complete design and construction of all of the optic (sans modulator) electronic subsystems that compose this device, along with explaining and providing working code needed to drive it in various modes of operation. The main objective of this thesis is to sufficiently instruct undergraduate and graduate colleagues so that they can replicate and build upon this work.

In dem von der Deutschen Forschungsgemeinschaft geförderten Projekt PROBADO (http://www.probado.de) werden von drei universitären Informatik-Instituten und zwei Bibliotheken innovative Verfahren für nicht-textuelle Dokumente entwickelt und prototypisch für die Bereiche 3D-Architekturmodelle und Musikdokumente in bibliothekarische Arbeitsabläufe integriert. Tonaufnahmen und Musikalien stehen dabei im Mittelpunkt des Teilprojekts Probado-Musik, das von der Bayerischen  Staatsbibliothek und dem Institut für Informatik III der Universität Bonn (Arbeitsgruppe Professor Clausen) betreut wird. Die drei wichtigsten Ziele dieses Teilprojektes sind die weitgehende Automatisierung der Erschließung und Indexierung von Musikdokumenten, die Anwendung neuartiger inhaltsbasierter Suchverfahren für Musik und der Aufbau nutzerfreundlicher Oberflächen zur komfortablen Suche und Bereitstellung von diesen Musikdokumenten. Hierfür wurde ein Musik-Repository aufgebaut, das Musikdokumente in verschiedenen Formaten enthält. Dazu gehören eingescannte Notendrucke im TIFF- und JPEG-Format, Audio-Dateien im WAV- und MP3-Format sowie symbolische Musikformate wie Music-XML und die Ausgabeformate von OMR-Programmen („Optical Music Recognition“). Das Musik-Repository enthält mittlerweile ca. 100.000 Seiten von eingescannten Notendrucken und Audio-Dateien von mehreren Hundert CDs. Die dazugehörigen Metadaten werden aus dem Katalog der Bayerischen Staatsbibliothek und externen Datenquellen gewonnen und in einer FRBR-basierten Datenbank abgelegt. (Katrin Braun, Münchener Digitalisierungszentrum / Digitale Bibliothek, Bayerische Staatsbibliothek)

No description available.

Doctor of Philosophy(PhD) / This thesis is concerned with the theoretical and practical development of a surface-based mapping algorithm for reliable and robust localization and mapping in prior unknown and unstructured environments. A surface-based map consists of a set of compressed surfaces, processed and represented without geometrical modelling. Each surface in the surface-based map represents an object in the environment. The ability to represent the exact shapes of objects via individual surfaces during the mapping process makes the surface-based mapping algorithm valuable in a number of navigation applications, such as mapping of prior unknown indoor and outdoor unstructured environments, target tracking, path planning and collision avoidance. The ability to unify representations of the same object taken from different viewpoints into a single surface makes the algorithm capable of working in multi-robot mapping applications. A surface-based map of the environment is build incrementally by acquiring the 3D range image of the scene, extracting the objects' surfaces from the 3D range image, aligning the set of extracted surfaces relative to the map and unifying the aligned set of surfaces with surfaces in the map. In the surface unification process the surfaces representing the same object are unified to make a single surface. The thesis introduces the following new methods which are used in the surface-based mapping algorithm: the extraction of surfaces from 3D range images based on a scanned surface continuity check; homogenization of the representation of the non-homogenously sampled surfaces; the alignment of the surface set relative to a large set of surfaces based on surface-based alignment algorithm; evaluating the correspondence between two surfaces based on the overlap area between surfaces; unification of the two surfaces belonging to the same object; and surface unification for a large set of surfaces. The theoretical contributions of this thesis are demonstrated with a series of practical implementations in different outdoor environments.

Doctor of Philosophy(PhD) / This thesis is concerned with the theoretical and practical development of a surface-based mapping algorithm for reliable and robust localization and mapping in prior unknown and unstructured environments. A surface-based map consists of a set of compressed surfaces, processed and represented without geometrical modelling. Each surface in the surface-based map represents an object in the environment. The ability to represent the exact shapes of objects via individual surfaces during the mapping process makes the surface-based mapping algorithm valuable in a number of navigation applications, such as mapping of prior unknown indoor and outdoor unstructured environments, target tracking, path planning and collision avoidance. The ability to unify representations of the same object taken from different viewpoints into a single surface makes the algorithm capable of working in multi-robot mapping applications. A surface-based map of the environment is build incrementally by acquiring the 3D range image of the scene, extracting the objects' surfaces from the 3D range image, aligning the set of extracted surfaces relative to the map and unifying the aligned set of surfaces with surfaces in the map. In the surface unification process the surfaces representing the same object are unified to make a single surface. The thesis introduces the following new methods which are used in the surface-based mapping algorithm: the extraction of surfaces from 3D range images based on a scanned surface continuity check; homogenization of the representation of the non-homogenously sampled surfaces; the alignment of the surface set relative to a large set of surfaces based on surface-based alignment algorithm; evaluating the correspondence between two surfaces based on the overlap area between surfaces; unification of the two surfaces belonging to the same object; and surface unification for a large set of surfaces. The theoretical contributions of this thesis are demonstrated with a series of practical implementations in different outdoor environments.