FE - DI | Dissertações de Mestrado e Teses de Doutoramento
Permanent URI for this collection
Browse
Browsing FE - DI | Dissertações de Mestrado e Teses de Doutoramento by Field of Science and Technology (FOS) "Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática"
Now showing 1 - 10 of 255
Results Per Page
Sort Options
- 6D Pose Estimation and Object RecognitionPublication . Pereira, Nuno José Matos; Alexandre, Luís Filipe Barbosa de Almeida6D pose estimation is a computer vision task where the objective is to estimate the 3 degrees of freedom of the object’s position (translation vector) and the other 3 degrees of freedom for the object’s orientation (rotation matrix). 6D pose estimation is a hard problem to tackle due to the possible scene cluttering, illumination variability, object truncations, and different shapes, sizes, textures, and similarities between objects. However, 6D pose estimation methods are used in multiple contexts like augmented reality, for example, where badly placed objects into the real-world can break the experience of augmented reality. Another application example is the use of augmented reality in the industry to train new and competent workers where virtual objects need to be placed in the correct positions to look like real objects or simulate their placement in the correct positions. In the context of Industry 4.0, robotic systems require adaptation to handle unconstrained pick-and-place tasks, human-robot interaction and collaboration, and autonomous robot movement. These environments and tasks are dependent on methods that perform object detection, object localization, object segmentation, and object pose estimation. To have accurate robotic manipulation, unconstrained pick-and-place, and scene understanding, accurate object detection and 6D pose estimation methods are needed. This thesis presents methods that were developed to tackle the 6D pose estimation problem as-well as the implementations of proposed pipelines in the real-world. To use the proposed pipelines in the real-world a data set needed to be capture and annotated to train and test the methods. Some controlling robot routines and interfaces were developed in order to be able to control a UR3 robot in the pipelines. The MaskedFusion method, proposed by us, achieves pose estimation accuracy below 6mm in the LineMOD dataset and an AUC score of 93.3% in the challenging YCB-Video dataset. Despite longer training time, MaskedFusion demonstrates low inference time, making it suitable for real-time applications. A study was performed about the effectiveness of employing different color spaces and improved segmentation algorithms to enhance the accuracy of 6D pose estimation methods. Moreover, the proposed MPF6D outperforms other approaches, achieving remarkable accuracy of 99.7% in the LineMOD dataset and 98.06% in the YCB-Video dataset, showcasing its potential for high-precision 6D pose estimation. Additionally, the thesis presents object grasping methods with exceptional accuracy. The first approach, comprising data capture, object detection, 6D pose estimation, grasping detection, robot planning, and motion execution, achieves a 90% success rate in non-controlled environment tests. Leveraging a diverse dataset with varying light conditions proves critical for accurate performance in real-world scenarios. Furthermore, an alternative method demonstrates accurate object grasping without relying on 6D pose estimation, offering faster execution and requiring less computational power. With a remarkable 96% accuracy and an average execution time of 5.59 seconds on a laptop without an NVIDIA GPU, this method demonstrates efficiency and practicality performing unconstrained pick-and-place tasks using a UR3 robot.
- Acessibilidade Aplicação BancáriaPublication . Nunes, Gonçalo Carmona; Inácio, Pedro Ricardo Morais; Brito, João Carlos Nunes da SilvaA unidade curricular de Projeto de Dissertação ou de Estágio Curricular, teve como intuito a planificação seguida da implementação e desenvolvimento do projeto a ser desenvolvido, o qual foi delineado numa primeira fase. A planificação, foi realizada após a integração na empresa ITSector (empresa onde se realizou o estágio). Antes desta fase, foi ainda concretizada uma Academia de Android, fornecida por esta entidade. Finalizada a Academia, foi então, escolhido o tema a ser estudado, preparado e desenvolvido numa aplicação móvel já existente, relacionada com um banco. O presente relatório aborda um tema bastante importante atualmente, que é a acessibilidade, o qual foi estudado e depois implementado na aplicação já desenvolvida, de forma a melhorar e aumentar o nível de acessibilidade da mesma. Com o aumento de pessoas com determinadas incapacidades no mundo, é essencial produzir aplicações que consigam fornecer igualdade de acesso, a qualquer utilizador. Partindo assim deste tema, é realizado inicialmente um enquadramento teórico relativamente às deficiências. Neste documento, as deficiências mais comuns serão as mais evidenciadas. São também enunciadas ferramentas/recursos, e aspetos importantes a ter em conta, antes de qualquer tentativa de criação e desenvolvimento de uma aplicação Android. Posteriormente, é mencionada a planificação que foi definida pormenorizadamente, de como decorreu o desenvolvimento da aplicação. Nesta etapa são, portanto definidas as tarefas realizadas durante o estágio, bem como o respetivo plano a seguir para a execução das mesmas. E por fim, são identificados e discutidos alguns riscos possíveis de ocorrer durante todo o processo de desenvolvimento da aplicação. Além, de estes serem nomeados e detalhados, é também referido o plano de mitigação, sendo constituído pelas soluções a realizar, caso tivesse sido necessário, para minimizar ou mitigar estes riscos. Após a conclusão do estudo e do planeamento do projeto, iniciouse um período de desenvolvimento, onde se implementou a acessibilidade na aplicação móvel. Este capítulo encontrase descrito sucintamente, conforme com as tarefas que foram antes programadas, de modo que fique percetível tudo aquilo que foi aplicado e trabalhado. Após um tempo de longo estudo e investigação para recolha de dados importantes relativamente ao tema deste projeto, e posteriormente implementação, eis que a aplicação atinge os seus resultados e alguns dos quais são apresentados neste relatório. Fazem ainda parte deste relatório, as tecnologias e ferramentas que foram utilizadas no decorrer deste, e sem as quais não seria possível a realização deste projeto. No fim, é ainda feita uma conclusão para evidenciar os pontos fulcrais que ocorreram desde a planificação até à concretização do projeto, e tudo o que poderá ser adicionado à aplicação no futuro.
- Adaptabilidade não Supervisionada Independente da Língua, ao Perfil Linguístico do UtilizadorPublication . Pedro, Ramos Eduardo; Pais, Sebastião Augusto Rodrigues Figueiredo; Cordeiro, João Paulo da CostaVivemos atualmente num período de grande entusiasmo, devido ao exponencial do surgimento de diversas ferramentas associadas às normas e tecnologias agregadas às adaptabilidades não supervisionadas, independentes da língua do utilizador. Apesar disso, o uso dessas ferramentas não está acessível a todos os utilizadores, sendo que se assume como um obstáculo. Assumindo que o sucesso de qualquer tecnologia se encontra dependente da sua aceitação por parte dos seus utilizadores, a presente investigação pretende contribuir para uma exposição de conceitos, normas e tecnologias associadas às sociedades adaptáveis, com métodos e técnicas que permitem uma maior familiarização das interfaces adaptáveis, não supervisionadas ao perfil dos utilizadores. Este trabalho propõe um estudo sobre a análise de uma adaptabilidade não supervisionada independente da língua não supervisionada para um perfil linguístico com uso de métodos e técnicas de processamento de linguagem natural (PLN) usando a similaridade assimétrica com o uso de técnicas de extração de textos e, posteriormente, avaliar as similaridades encontradas. Foram utilizados métodos linguísticos e técnicas uniformes, baseados no perfil linguístico do utilizador, enquanto medida comparada na variedade de línguas com base num vasto leque de línguas potencialmente heterogéneas variáveis, onde se focaliza especificamente uma caraterística metodológica baseada na adaptabilidade não supervisionada independente da língua do perfil linguístico do utilizador. Devido à grande quantidade de informação disponível atualmente no meio eletrónico, uma das tarefas consiste na classificação adaptável do perfil de um utilizador e na extração de termos relevantes e não relevantes que tem vindo a ganhar importância nas pesquisas realizadas nas áreas de extração de termos e na recuperação de informações. A média apresenta melhores resultados do que as combinações baseadas na estratégia de classificação de melhores resultados. O método proposto parece ser útil para desambiguar uma grande percentagem de consultas temporais para os utilizadores.
- Adaptive Methods for Color Vision Impaired UsersPublication . Ribeiro, Maria Madalena Gonçalves; Gomes, Abel João PadrãoColor plays a key role in the understanding of the information in computer environments. It happens that about 5% of the world population is affected by color vision deficiency (CVD), also called color blindness. This visual impairment hampers the color perception, ending up by limiting the overall perception that CVD people have about the surrounding environment, no matter it is real or virtual. In fact, a CVD individual may not distinguish between two different colors, what often originates confusion or a biased understanding of the reality, including web environments, whose web pages are plenty of media elements like text, still images, video, sprites, and so on. Aware of the difficulties that color-blind people may face in interpreting colored contents, a significant number of recoloring algorithms have been proposed in the literature with the purpose of improving the visual perception of those people somehow. However, most of those algorithms lack a systematic study of subjective assessment, what undermines their validity, not to say usefulness. Thus, in the sequel of the research work behind this Ph.D. thesis, the central question that needs to be answered is whether recoloring algorithms are of any usefulness and help for colorblind people or not. With this in mind, we conceived a few preliminary recoloring algorithms that were published in conference proceedings elsewhere. Except the algorithm detailed in Chapter 3, these conference algorithms are not described in this thesis, though they have been important to engender those presented here. The first algorithm (Chapter 3) was designed and implemented for people with dichromacy to improve their color perception. The idea is to project the reddish hues onto other hues that are perceived more regularly by dichromat people. The second algorithm (Chapter 4) is also intended for people with dichromacy to improve their perception of color, but its applicability covers the adaptation of text and image, in HTML5- compliant web environments. This enhancement of color contrast of text and imaging in web pages is done while keeping the naturalness of color as much as possible. Also, to the best of our knowledge, this is the first web recoloring approach targeted to dichromat people that takes into consideration both text and image recoloring in an integrated manner. The third algorithm (Chapter 5) primarily focuses on the enhancement of some of the object contours in still images, instead of recoloring the pixels of the regions bounded by such contours. Enhancing contours is particularly suited to increase contrast in images, where we find adjacent regions that are color indistinguishable from dichromat’s point of view. To our best knowledge, this is one of the first algorithms that take advantage of image analysis and processing techniques for region contours. After accurate subjective assessment studies for color-blind people, we concluded that the CVD adaptation methods are useful in general. Nevertheless, each method is not efficient enough to adapt all sorts of images, that is, the adequacy of each method depends on the type of image (photo-images, graphical representations, etc.). Furthermore, we noted that the experience-based perceptual learning of colorblind people throughout their lives determines their visual perception. That is, color adaptation algorithms must satisfy requirements such as color naturalness and consistency, to ensure that dichromat people improve their visual perception without artifacts. On the other hand, CVD adaptation algorithms should be object-oriented, instead of pixel-oriented (as typically done), to select judiciously pixels that should be adapted. This perspective opens an opportunity window for future research in color accessibility in the field of in human-computer interaction (HCI).
- Algorithms for the assessment of tremors for diagnostic support of neurodegenerative diseasesPublication . Raposo, João Miguel Vilelas; Santos, Nuno Manuel Garcia dos; Oliveira, José António Martinez Souto deAs doenças neurodegenerativas costumam incluir sintomas como tremores frequentemente visto em movimento dos membros. No entanto, esses movimentos podem ser invisíveis par ao olho humano e pode variar de paciente para paciente, variando também com a hora do dia. O preciso medição dos tremores pode provar ser uma ferramenta importante para o apoio no diagnóstico e a terapia para essas doenças. Com base no conhecimento prévio, esta dissertação irá desenvolver um algorimo (ou conjunto de algoritmos) para medir e fornecer informações a um médico sobre o estado do sintoma específico.
- Alocação de Máquinas Virtuais em Ambientes de Computação em Nuvem Baseada em Requisitos de Service Level AgreementPublication . Oliveira, Raysa da Luz; Freire, Mário MarquesA computação em nuvem teve um avanço considerável nos últimos anos, trazendo grandes benefícios incluindo escalabilidade, flexibilidade, acessibilidade global, melhor utilização de recursos e redução de custos, entre outros. Apesar de todos os benefícios, esta adesão e crescimento trás consigo grandes desafios como otimização do uso de recursos computacionais, redução de custos, garantia da qualidade de serviço (Quality of Service (QoS)), segurança, etc. As garantias da qualidade de serviço são estabelecidas através de Service Level Agreements (SLAs), que são contratos estabelecidos entre o cliente e o fornecedor do serviço de computação em nuvem, visando especificar de forma mensurável as metas de nível de serviço a serem cumpridas, além dos papéis e responsabilidades das partes envolvidas. Este trabalho apresenta um estudo sobre cumprimento de SLAs por algoritmos de alocação de máquinas virtuais em ambientes de computação em nuvem. O trabalho tem em consideração métricas como disponibilidade, custo, tempo de conclusão de uma aplicação (task completion time) e nível de tolerância a faltas, avaliando o cumprimento de tais métricas em diferentes cenários. O estudo é realizado utilizando o framework CloudSim Plus para modelação e execução de simulações de computação em nuvem. São introduzidos dois módulos no framework visando: (i) especificação de SLAs e templates de máquinas virtuais em formato JavaScript Object Notation (JSON), seguindo padrões do Amazon Elastic Compute Cloud (Amazon EC2); (ii) injeção de faltas aleatórias, permitindo avaliar como os SLAs são afetados perante o surgimento de faltas nos servidores. Por fim, o trabalho apresenta uma proposta para automação da criação e alocação de máquinas virtuais, visando cumprir os SLAs e libertar o cliente da necessidade de especificar a quantidade mínima de máquinas virtuais para atendimento dos níveis de serviço exigidos. Mesmo com todo o nível de automação que os fornecedores de computação em nuvem possam oferecer, os resultados obtidos mostram que é possível melhorar a automação destes serviços, reduzindo a necessidade de intervenção do cliente e as violações de SLA devido a uma inadequada configuração de máquinas virtuais realizada pelo cliente.
- An ambient assisted living solution for mobile environmentsPublication . Horta, Edgar Tavares da; Rodrigues, Joel José Puga CoelhoAn Ambient Assisted Living (AAL) mobile health application solution with biofeedback based on body sensors is very useful to perform a data collection for diagnosis in patients whose clinical conditions are not favourable. This system allows comfort, mobility, and efficiency in all the process of data collection providing more confidence and operability. A physical fall may be considered something natural in the life span of a human being from birth to death. In a perfect scenario it would be possible to predict when a fall will occur in order to avoid it. Falls represent a high risk for senior people health. Those falls can cause fractures or injuries causing great dependence and debilitation to the elderly and even death in extreme cases. Falls can be detected by the accelerometer included in most of the available mobile phones or portable digital assistants (PDAs). To reverse this tendency, it can be obtained more accurate data for patients monitoring from the body sensors attached to the human body (such as, electrocardiogram (ECG), electromyography (EMG), blood volume pulse (BVP), electro dermal activity (EDA), and galvanic skin response (GSR)). Then, this dissertation reviews the related literature on this topic and introduces a mobile solution for falls prevention, detection, and biofeedback monitoring. The proposed system collects sensed data that is sent to a smartphone or tablet through Bluetooth. Mobile devices are used to process and display information graphically to users. The falls prevention system uses collected data from sensors in order to control and advice the patient or even to give instructions to treat an abnormal condition to reduce the falls risk. In cases of symptoms that last more time it can even detect a possible disease. The signal processing algorithms plays a key role in the fall prevention system. These algorithms in real time, through the capture of biofeedback data, are needed to extract relevant information from the signals detected to warn the patient. Monitoring and processing data from sensors is realized by a smartphone or tablet that will send warnings to users. All the process is performed in real time. These mobile devices are also used as a gateway to send the collected data to a Web service, which subsequently allows data storage and consultation. The proposed system is evaluated, demonstrated, and validated through a prototype and it is ready for use.
- An Embryo Quality Assessment Application for ART LaboratoriesPublication . Silva, João Pedro Ribeiro Pessoa da; Fazendeiro, Paulo André PaisIn this work it is described a system that aims to meet the needs of information management in Assisted Reproductive Technologies (ART) laboratories; it is also proposed an evaluation model of embryos based on a triangular norm. One of the most common undesired consequences of ART treatments is the occurrence of multiple pregnancies. Therefore reducing the number of embryos per transfer is of paramount importance. In order to improve pregnancy rates while reducing the incidence of multiple births an accurate embryo quality evaluation is essential. The main objectives of the synthesized system are twofold: (i) the gathering and management of all the information needed to perform the embryo evaluation and (ii) the development of a tool that can serve as aid to carry out that evaluation with precision and consistency. The application herein presented accomplishes these two objectives. The analysis of the requirements of the assessment process has resulted in a flexible data model, used in the presented prototype, supporting the selective embryo transfer decision-making process.
- An Internet of Things (IoT) Ecosystem for Detection and Removal of Radon GasPublication . Ferreira, Sérgio Moreira; Silva, Bruno Miguel Correia da; Soares, SandraThe region that served as the basis for this work is characterized by the abundance of different types of granite. In some places, the predominance of certain uraniferous minerals, the existence of faults and mines may even be responsible for the existence of very high levels of natural radioactivity, especially with regard to radon. The main motivation for studying the effects of population exposure to radon is based on the fact that this gas, radioactive, odorless, colorless and tasteless, has been recognized since 1988 by the World Health Organization as the 2nd leading cause of death from lung cancer for the general population, after tobacco. Approximately 80 % of the background radiation to which populations are exposed daily comes from natural sources, which can include radioactive materials of natural origin, NORM, present in the earth’s crust, in food, in some drinks and, including certain building materials. In December 2018, a new legislation was approved in Portugal, DL no 108/2018, transposed from European Directive 59/2013 / Euratom, which establishes uniform basic safety standards for the protection of the health of people subject to exposure professional, public and medical professionals against the dangers arising from ionizing radiation. The radon isotope, 222Rn, which results from radio disintegration, 226Ra, and one of the uranium decay products, 238U, is the main radionuclide of radiological interest in terms human health. This radioactive gas is spontaneously released from some rocks and, being slightly denser than air, it tends to accumulate in ground areas of closed spaces and is responsible for the majority of the radiation dose received by the population. Some traditional solutions have been applied in order to solve the problem, however, after the detection of the radon they do not use the data collected in order to combat and prevent the gas from accumulating in the interior spaces. In Portugal it is a problem that has not been given due attention. Taking into account the possible harmful effects of radon on the population, the main motivation of this work was the development of an IoT (Internet of Things) system with the objective of detecting, mitigating and predicting the levels of radon gas inside homes. Based on this assumption, it will be possible to create a healthier environment that will reduce health risks in closed environments. To achieve this goal, an IoT system was developed, consisting of a radon sensor and an automated switch, which allows the fans installed and / or switched off to be installed in the residence, as well as predicting and making decisions in the face of a potentially dangerous situation based on the records provided by the sensors. This system also includes a web application that allows the user to consult the statistical data related to the average radon concentration. On the other hand, through the web application, it is possible to not only activate the fans manually, but also to edit the limit at which the fans are activated. This system was tested in a real context and, therefore, it is possible to compare two radon gas mitigation methods and identify which would be the most effective. Once the method to be used was determined and after stabilizing the data obtained by the sensor, the results obtained by the system were analyzed, which allowed us to conclude that the system reduced considerably the levels of radon in the house. On average, there was a 93% reduction in the concentration of radon gas. This result demonstrates that the implementation of the system was successful and allows us to conclude that it can be applied on a larger scale.
- An off-the-shelf platform for automatic and interactive text messaging using Short Message ServicePublication . Oliveira, Daniel Sabugueiro; Santos, Nuno Manuel Garcia dos; Esgalhado, Maria da Graça ProençaThis dissertation presents the design and construction of a platform for the implementation of an automatic and interactive message exchanging system using Short Message Service (SMS), also known as texting, created to support a research on the Psychology field. The research included researchers from the Psychology and Education Department and the Computer Science Department from University of Beira Interior making it a multidisciplinary study. Its goal was to assess if the exposure to motivational, persuasive and informative SMS texts could improve the self-efficacy of mathematics students from that university. The platform was devised to use low cost off-the-shelf parts, yet allowing the design of an efficient and robust system. The source code for the SMS system is publicly available at the Assisted Living Computing and Telecommunications (ALLab) website.