DISSERTAÇÃO DE MESTRADO - PROGRAMA DE PÓS-GRADUAÇÃO EM CIÊNCIA DA COMPUTAÇÃO
URI Permanente para esta coleçãohttps://tedebc-teste.ufma.br/handle/tede/1314
Áreas de Concentração e Linhas de Pesquisa:
Automação e Contrôle
Ciência da Computação
Sistemas de Energia Elétrica
Navegar
Navegando DISSERTAÇÃO DE MESTRADO - PROGRAMA DE PÓS-GRADUAÇÃO EM CIÊNCIA DA COMPUTAÇÃO por Data de Publicação
Agora exibindo 1 - 17 de 17
- Resultados por página
- Opções de Ordenação
Item METODOLOGIA DE DETECÇÃO E RECONHECIMENTO DE SEMÁFOROS UTILIZANDO REDES NEURAIS ARTIFICIAIS(Universidade Federal do Maranhão, 2016-03-22) SOARES, Julio Cesar da Silva; ALMEIDA NETO, Areolino de; 279344543-68; http://lattes.cnpq.br/8041675571955870Urban roads are very complex. The increase in the flow of vehicles in the cities has contributed to traffic accidents. Researches for accident reduction show that the traffic lights are effective in reducing accidents. Traffic lights can minimize the occurrence of accidents at intersections and crosswalks. The implementation of traffic light signals shows significant advantages, otherwise reveals some problems such as the failure to detect road signs by drivers on urban roads. This fact is related to excessive visual information, the stress of the drivers and/or eyestrain makes the drivers lose their attention. These reasons motivated researches about intelligent vehicles. This work aims to develop a methodology to detect and recognize traffic lights, to be applied in smart vehicles. This methodology can contribute to the Advanced Driver Support Systems (ADAS), which assists drivers, especially those with partial vision impairment. Image processing techniques are used to develop the detection methodology. Back project and global thresholding are combined to find light points. Local thresholding techniques are applied to calculate the symmetry between the radius and the center of the light points to segment the traffic light body. The first step got an average rate of 99% of detection. The features of the traffic lights are extracted using Haralick texture measures, with the inclusion of color and shape information. The data generated by the feature extraction step were preprocessed using the SMOTE technique to balance the database. The recognition and identification of the traffic lights state are made by an artificial neural network using Multilayer-Perceptron (MLP). The backpropagation learning algorithm are used in the network training. The validation results show an average recognition rate of 98%.Item Diagnóstico de glaucoma a partir de imagens de fundo de olho utilizando índices de diversidade(Universidade Federal do Maranhão, 2018-02-26) ARAUJO, Jose Denes Lima; SILVA, Aristófanes Corrêa; 288745363-72; http://lattes.cnpq.br/2446301582459104; PAIVA, Anselmo Cardoso de; 375523843-87; http://lattes.cnpq.br/6446831084215512; PAIVA, Anselmo Cardoso de; 375523843-87; http://lattes.cnpq.br/6446831084215512; SILVA, Aristófanes Corrêa; 288745363-72; http://lattes.cnpq.br/2446301582459104; BRAZ JUNIOR, Geraldo; http://lattes.cnpq.br/8287861610873629; AIRES, Kelson Romulo Teixeira; http://lattes.cnpq.br/0065931835203045Glaucoma is one of the leading causes of blindness worldwide, and is usually caused by an increase in the intraocular pressure that damages the optic nerve and gradually causes vision loss. It is a disease that has no symptoms in the early stages and its early diagnosis can prevent the vision loss and blindness. Fundus images are used by experts to examine the optic disc in order to identify the changes caused by glaucoma. In addition, image processing and pattern recognition techniques are used for the development of computational tools in order to provide support in the process of analyzing these images. This work proposes a methodology for the glaucoma diagnosis from fundus images using diversity indexes as texture descriptors. After extraction of texture features, genetic algorithms are used to select the best set of features. Finally, the support vector machine is used to perform the classification. The proposed methodology revealed promising results for glaucoma diagnosis, reaching accuracy of 93.41%, sensitivity of 92.36% and specificity of 95.05%, as best resultsItem COEFICIENTES DE CORRELAÇÃO COMO MÉTRICA DE AVALIAÇÃO DAS ESTRATÉGIAS DE CONTROLE INTELIGENTE FEL E MNFEL.(Universidade Federal do Maranhão, 2018-06-26) SANTOS, Alex Newman Veloso dos; RIBEIRO, Paulo Rogério de Almeida; 016.320.583-32; http://lattes.cnpq.br/0035213619257246; OLIVEIRA, Alexandre César Muniz de; 288.350.993-68; OLIVEIRA, Alexandre César Muniz de; 288.350.933-68; http://lattes.cnpq.br/5225588855422632Control systems have been largely used in many fields such as industrial plants, robotics, medicine and so on. Therefore, new techniques are frequently proposed to enhance these systems. Feedback-Error-Learning (FEL) is an intelligent control strategy which applies a neural network alongside a conventional controller, as an example the ProportionalIntegral-Derivative (PID) that is the most used on the industry. The enhanced control is achieved in FEL by the acquisition of the inverse model or the non-linearity compensation. Moreover, Multi-Network-Feedback-Error-Learning (MNFEL), which is based on FEL, uses multiple neural networks that can lead to a better control. FEL and MNFEL works assume that enhanced controls are achieved by adding neural networks, however, there are few works account for the network’s degree of contribution to the control system. A previous research proposed a metric based on Pearson product-moment correlation coefficient (PC). However, this metric assumes working conditions that may not be met in control systems. This works aims to propose two approaches based on Spearman Coefficient (SC) and PC. The evaluation methodology is comprised of two phases. The first phase, placed before the intelligent control strategy insertion, determines the expected SC behavior based on the initial analysis of the correlation between the squared error and the conventional controller. The second phase evaluates the coefficient behavior during the neural network training. Two industrial plants were used in this work: Burner group of a Pelletizing plant and Cooling Coil plant. The results shown: i) the previous work approach using PC may lead to precipitated conclusions about the system in analysis; ii) the proposed approach using SC demonstrated – in both plants – the neural networks’ degree of contribution while enhancing the control; iii) the SC – during the networks’ training – can preview that those networks will or not significantly enhance the control, i.e. indicating that those networks may not contribute in the control system. Thus, the proposed approach, which uses PC and SC, may calculate the contribution of the neural networks during the improvment of the control system with FEL and MNFEL.Item Integrando a Internet das Coisas às Aplicações de TV Digital.(Universidade Federal do Maranhão, 2018-06-27) PEREIRA, Danne Makleyston Gomes; SOARES NETO, Carlos de Salles; 780.057.283-87; http://lattes.cnpq.br/1512846862093142; SILVA, Francisco José da Silva e; 406.504.133-34; http://lattes.cnpq.br/0770343284012942; SILVA, Francisco José da Silva e; 406.504.133-34; http://lattes.cnpq.br/0770343284012942; SOARES NETO, Carlos de Salles; 780.057.283-87; http://lattes.cnpq.br/1512846862093142; COLCHER, Sergio; http://lattes.cnpq.br/1104157433492666; LOPES, Rafael Fernandes; http://lattes.cnpq.br/1972734433460838In comparison to the conventional TV, Digital TV allows not only a better image and sound quality but also interactivity and the offering of new services to viewers. On the other hand, the Internet of Things (IoT) is a combination of ubiquitous computing and the Internet, in which IoT devices (e.g., smart objects) can get and exchange data, cooperating with people and the environment in which they are. Many smart objects located in a home can offer features that can be explored to enrich the experience of watching television. However, the convergence between IoT devices and TV applications has several challenges, such as the dynamic discovery of smart objects and the services they provide; the support for the heterogeneity of communication protocols used by smart objects; the identification of viewers and their profiles; the synchronization of the playing media with the smart objects actuators provided in the physical environment; and the need of providing programming abstractions for the development of TV applications that are aware of the smart objects deployed in the physical environment. This work proposes a software architecture, based on a middleware perspective, which allows TV applications to be aware of the physical environment context and interact with it. It also allows the identification of the viewers and their profiles, also providing multi-modal interactions with them and the application running in the digital TV. The proposed software architecture is based on a conceptual model called IoTTV-Ont, also developed as part of this work, that allows the description of a household with smart objects and the people present in it. A prototype that uses a particular implementation of the proposed software architecture was evaluated through use cases, exploring all the requirements specified in this work.Item Mobcons-AT: uma ferramenta de autoria de restrições de mobilidade baseada na transformação de modelos(Universidade Federal do Maranhão, 2018-07-30) AZEVEDO JÚNIOR, Adalberto Teixeira; SILVA, Francisco José da Silva e; 406504133-34; http://lattes.cnpq.br/0770343284012942; COUTINHO, Luciano Reis; 504488203-82; http://lattes.cnpq.br/590156473265585; COUTINHO, Luciano Reis; 504488203-82; http://lattes.cnpq.br/590156473265585; SILVA, Francisco José da Silva e; 406504133-34; http://lattes.cnpq.br/0770343284012942; VALE, Samyr Beliche; http://lattes.cnpq.br/1531971102610447; COSTA, Fabio Moreira; http://lattes.cnpq.br/0925150626762308There are many situations in which there is a need to monitor in realtime the location and behavior of people and/or vehicles in order to detect possible irregularities and control where they are located and how they move, such as in companies, public transportation and public security. In this dissertation, Mobility Constraints Authoring Tool (MobCons-AT) is presented, an authoring tool for enduser programming that allows the specification of mobility restrictions rules that must be followed by mobile devices. The rules are specified by means of a domain specific language called Mobility Constraints Specification Language (MobCons-SL), the central part of the tool. Once specified in MobCons-SL, these rules are automatically transformed into artifacts that allow real-time detection of constraint violations by using Complex Event Processing (CEP). This approach allows the reduction in the delivery time of solution to the customer and necessary cost for the development of customized software solutions for real-time detection of mobility constraints. This dissertation also describes the use of MobCons-AT in five case studies, showing its applicability for diverse scenarios.Item Cosmo: Um ambiente virtual de aprendizado com foco na introdução de algoritmos(Universidade Federal do Maranhão, 2018-07-31) RABÊLO JÚNIOR, Dilson José Lins; SOARES NETO, Carlos de Salles; 780057283-87; http://lattes.cnpq.br/1512846862093142; SOARES NETO, Carlos de Salles; 780057283-87; http://lattes.cnpq.br/1512846862093142; VALE, Samyr Béliche; http://lattes.cnpq.br/1531971102610447; FRANÇA, Rômulo Martins; http://lattes.cnpq.br/8153351143565398The Cosmo is a multitasking teaching platform extensible by plugins, focused in activities related to the study of algorithms. This platform has a layer of Logs management, which stores various information. The proposal of this work is to analyze the Logs of the Cosmo in search of factors that may help platform requirements and identify the differentiated student learning curve. In the experiment it is analyzed the progress of a class of 45 undergraduate students in a course of Introduction to Algorithms. An individualized qualitative assessment and a class analysis were carried out, showing that the collection of the Cosmo data can support the progress evaluation of learners.Item INavigS: uma infraestrutura de software ciente de contexto para navegação indoor(Universidade Federal do Maranhão, 2018-08-27) MEDEIROS, Allinger Lima; VALE, Samyr Béliche; 711523003-00; http://lattes.cnpq.br/1531971102610447; VALE, Samyr Béliche; 711523003-00; http://lattes.cnpq.br/1531971102610447; SILVA, Francisco José da Silva e; http://lattes.cnpq.br/0770343284012942; LEÃO, Erico Meneses; http://lattes.cnpq.br/1117954290627743Context-aware computing is a research area of ubiquitous computing that aims to provide services and/or information without taking the user’s focus away from the current task. In turn, location-based services are categorized as a special type of context-aware applications, where localization assumes a primary role. Because of the evolution of positioning technologies and the fact that people spend most of their time in the indoor environment, the demand for indoor location-based services keeps growing. However, the development of these applications involves overcoming several challenges. In particular, for applications using indoor navigation services, one of the most important categories, the challenges include: obtaining user location and orientation, representing indoor space, generation custom routes, generation of instructions along the route, power management, and non-existence of APIs. In view of these challenges and in order to reduce the effort required for the development of these applications, this work proposes a context-sensitive and mobile-supported software infrastructure that provides transparent indoor positioning and navigation services.Item Aprendizagem Profunda Aplicada ao Diagnóstico de Melanoma(Universidade Federal do Maranhão, 2019-02-14) MAIA, Lucas Bezerra; PAIVA, Anselmo Cardoso de; 375523843-87; http://lattes.cnpq.br/6446831084215512; BRAZ JÚNIOR, Geraldo; 000520303-18; http://lattes.cnpq.br/8287861610873629; BRAZ JÚNIOR, Geraldo; 000520303-18; http://lattes.cnpq.br/8287861610873629; PAIVA, Anselmo Cardoso de; 375523843-87; http://lattes.cnpq.br/6446831084215512; ALMEIDA, João Dallyson Sousa de; http://lattes.cnpq.br/6047330108382641; CARVALHO FILHO, Antonio Oseas de; http://lattes.cnpq.br/7913655222849728Melanoma is the most lethal type of cancer when compared to others skin diseases. However, when the diagnosis is made in its initial stage, patients have high rates of recovery. Several approaches to automatic detection and diagnosis of melanoma have been explored by different authors in order to provide an auxiliary opinion to specialists. Training models with the existing data sets have been a difficult task due to the problem of imbalanced data. This work aims to evaluate to the evaluation the performance of machine learning algorithms combined with imbalanced learning technique, regarding the task of melanoma diagnosis. The architectures of Convolutional Neural Networks VGG16, VGG19, Inception, and ResNet were used along with ABCD rule to extract patterns of skin lesions in a set of 200 dermatoscopic images. The Random Forest classifier reached a sensitivity of 92.5 % and a kappa index of 77.15 % after the use of attribute selection with Greedy Stepwise and balancing the training data with Synthetically Minority Oversampling TEchnique (SMOTE) and the Edited Nearest Neighbor (ENN) rule.Item Segmentação automática da próstata em imagens de ressonância magnética utilizando redes neurais convolucionais, mapa probabilístico e treinamento adversário(Universidade Federal do Maranhão, 2019-02-15) FERREIRA, Jonnison Lima; PAIVA, Anselmo Cardoso de; 375523843-87; http://lattes.cnpq.br/6446831084215512; SILVA, Aristófanes Corrêa; 288745363-72; http://lattes.cnpq.br/2446301582459104; SILVA, Aristófanes Corrêa; 288745363-72; http://lattes.cnpq.br/2446301582459104; PAIVA, Anselmo Cardoso de; 375523843-87; http://lattes.cnpq.br/6446831084215512; CAVALCANTE, André Borges; http://lattes.cnpq.br/3885279033465023; BRAZ JÚNIOR, Geraldo; http://lattes.cnpq.br/8287861610873629; CARVALHO FILHO, Antonio Oseas de; http://lattes.cnpq.br/7913655222849728Prostate cancer is the second most common cancer among men, being the second most deadly. Early detection is a strategy to find the tumor at an early stage and thus provide a better chance of treatment. Currently the prostate gland imaging test has grown for prevention, diagnosis and treatment. The manual segmentation of the prostate is delayed and the propensity to variability among those expected, due to work, alternatives such as computational systems that use image processing and the identification of more advanced and exploited patterns for the early diagnosis of this disease, providing a second opinion for the specialist and increase the process. In this work, several automatic tasks are provided for the segmentation of the prostate from magnetic resonance imaging using a deep learning technique, probabilistic mapping and adversarial training of neural networks. The proposed methodology was tested on two public imaging databases, the Prostate 3T prostate and the PROMISE12, resulting in an average Dice of 89%.Item Aprendizagem Profunda Aplicada ao Diagnóstico do Glaucoma(Universidade Federal do Maranhão, 2019-02-19) LIMA, Alan Carlos de Moura; ALMEIDA, João Dallyson Sousa de; 003998573-38; http://lattes.cnpq.br/6047330108382641; BRAZ JÚNIOR, Geraldo; 000520303-18; http://lattes.cnpq.br/8287861610873629; BRAZ JÚNIOR, Geraldo; 000520303-18; http://lattes.cnpq.br/8287861610873629; ALMEIDA, João Dallyson Sousa de; http://lattes.cnpq.br/6047330108382641; PAIVA, Anselmo Cardoso de; http://lattes.cnpq.br/6446831084215512; VERAS, Rodrigo de Melo Souza; http://lattes.cnpq.br/2634254790193199Glaucoma is a cluster of ocular diseases that cause damage to the eye’s optic nerve and cause successive narrowing of the visual field in affected patients, due to an increase in intraocular pressure, which can lead the patient to blindness at an advanced stage without clinical reversal. For several years, from techniques of manual analysis of the internal structures of the eye to the use of deep learning with convolutional neural networks (CNNs) were successfully used in the diagnosis of glaucoma. However, building a deep learning network requires a lot of effort that in many situations is not always able to achieve satisfactory results due to the amount of parameters that need to be configured to adapt the CNN architecture to the problem in question. The objective of this work is to use a hyperparameter search technic to select the tuned parameters of a genetic algorithm (GA) to select the best CNN architecture through evolutionary techniques and to be able to aid in the accurate diagnosis of glaucoma, in eye fund images. The proposed methodology was applied in 455 images from RIM-ONE dataset, in its version 2 (r2), with resized images to 96x96 pixels in the RGB color model. The selected CNN by AG, after its training, achieved for the diagnosis of glaucoma the results of 96.63% for accuracy, 94.87% for sensitivity, 98.00% for specificity, 97.37% for precision and 96.10% for f-score.Item Brain-to-brain mapping: an approach to share neural Information on Ratslam(Universidade Federal do Maranhão, 2019-03-22) MENEZES, Matheus Chaves; OLIVEIRA, Alexandre César Muniz de; 288350933-68; http://lattes.cnpq.br/5225588855422632; OLIVEIRA, Alexandre César Muniz de; 288350933-68; http://lattes.cnpq.br/5225588855422632; RIBEIRO, Paulo Rogério de Almeida; http://lattes.cnpq.br/0035213619257246; ALMEIDA NETO, Areolino de; http://lattes.cnpq.br/8041675571955870; FREITAS, Edison Pignaton de; http://lattes.cnpq.br/2154028088891512Diversas aplicações robóticas são melhor executadas por sistemas com vários robôs em vez de apenas um, por exemplo, para explorar grandes áreas em missões críticas de busca e resgate em cenários de pós-desastre. Essas vantagens podem ser devidas à divisão de atividades, redução de custos e tempo. A Localização e Mapeamento Simultâneos (SLAM) desempenha um papel central na exploração de ambientes desconhecidos. O RatSLAM, que é baseado no sistema de navegação presente no hipocampo do cérebro dos roedores, tem sido amplamente utilizado em aplicações SLAM baseadas em vídeo. No RatSLAM, a informação neural é definida como experiências, que associam características do ambiente e movimento em uma representação única no mapa. Este trabalho apresenta uma abordagem para compartilhar informações neurais no RatSLAM, chamado brain-tobrain mapping, no qual a experiência de mapas parciais é compartilhada por vários robôs para construir cooperativamente um mapa de todo o ambiente. O primeiro passo para compartilhar informações neurais é conectar diferentes instâncias do RatSLAM através de um mecanismo de fusão, específico para o RatSLAM. Para realizar a fusão, é necessário que os robôs passem pelo menos um lugar comum entre eles e adquiram a mesma experiência sobre o lugar comum. A fusão permite que todos os robôs saibam sobre suas experiências (pose cells, local view cells e experience map) em uma estrutura compartilhada. Assim, os robôs de exploração podem reutilizar experiências aprendidas sobre o ambiente para melhorar o seu procedimento de mapeamento, como por exemplo: um robô pode corrigir parte do mapa de outro robô, enquanto usa informações compartilhadas para melhorar seu próprio mapa fechando loops. Três experimentos de diferentes ambientes foram realizados para validar a nova abordagem: um ambiente simulado, um laboratório de pesquisa e um dataset usado para validar o trabalho original do RatSLAM. Os resultados mostraram que o mapa final construído por robôs com experiência compartilhada é visualmente semelhante (mas não idêntico) a um construído por um robô realizando a mesma tarefa de mapeamento individualmente, ou seja, sem compartilhar informações.Item Monitoramento ubíquo da saúde mental: detectando padrões de sociabilidade enriquecidos por contexto através do processamento de eventos complexos(Universidade Federal do Maranhão, 2020-02-10) MOURA, Ivan Rodrigues de; COUTINHO, Luciano Reis; 504488203-82; http://lattes.cnpq.br/5901564732655853; SILVA, Francisco José da Silva e; 406504133-34; http://lattes.cnpq.br/0770343284012942; SILVA, Francisco José da Silva e; 406504133-34; http://lattes.cnpq.br/0770343284012942; COUTINHO, Luciano Reis; 504488203-82; http://lattes.cnpq.br/5901564732655853; SOARES NETO, Carlos de Salles; http://lattes.cnpq.br/1512846862093142; CAMARGO, Raphael Yokoingawa de; http://lattes.cnpq.br/5519687175393434Traditionally, the monitoring of individuals with mental disorders is conducted through face-to-face meetings with professionals specialized in mental health. Today, however, computational methods can use ubiquitous devices (e.g., smartphones and wearable technologies) to monitor social behaviors related to mental health rather than relying on self-reports. These devices represent a valuable source of contextual data that allows the identification of the social activities experienced by individuals in their daily routine. Therefore, the use of these technologies to identify social activities habit enables the recognition of abnormal social behaviors that may be mental disorders indicative. For this reason, this study presents a new approach to monitoring mental health through social situation awareness. This work introduces an algorithm capable of detecting sociability patterns, i.e., it characterizes the periods of the day when the individual socializes habitually. The recognition of social routine is performed under different context conditions (e.g., workday and weekend), which allows differentiating abnormal social behaviors from changes in social habits expected in certain situations. The solution presented is also able to identify changes in social patterns that may be indicative of mental disorders presence. The implementation of the proposed algorithm used the combination of the Frequent Patterns Mining approach with the Complex Events Processing, which allows the realization of the continuous social data stream processing. The evaluation performed demonstrated that context-based recognition provides a better understanding of the social routine, also indicating that the proposed solution is capable of detecting sociability patterns similar to a batch algorithm. Additionally, it was validated the performance of the social behavior change detection mechanism.Item Monitoramento ubíquo da saúde mental: detectando padrões de sociabilidade enriquecidos por contexto através do processamento de eventos complexos.(Universidade Federal do Maranhão, 2020-02-17) MOURA, Ivan Rodrigues de; COUTINHO, Luciano Reis; http://lattes.cnpq.br/5901564732655853; SILVA, Francisco José da Silva e; http://lattes.cnpq.br/0770343284012942; SILVA, Francisco José da Silva e; http://lattes.cnpq.br/0770343284012942; COUTINHO, Luciano Reis; http://lattes.cnpq.br/5901564732655853; SOARES NETO, Carlos de Salles; http://lattes.cnpq.br/1512846862093142; CAMARGO, Raphael Yokoingawa de; http://lattes.cnpq.br/5519687175393434Traditionally, the monitoring of individuals with mental disorders is conducted through face-to-face meetings with professionals specialized in mental health. Today, however, computational methods can use ubiquitous devices (e.g., smartphones and wearable technologies) to monitor social behaviors related to mental health rather than relying on self-reports. These devices represent a valuable source of contextual data that allows the identification of the social activities experienced by individuals in their daily routine. Therefore, the use of these technologies to identify social activities habit enables the recognition of abnormal social behaviors that may be mental disorders indicative. For this reason, this study presents a new approach to monitoring mental health through social situation awareness. This work introduces an algorithm capable of detecting sociability patterns, i.e., it characterizes the periods of the day when the individual socializes habitually. The recognition of social routine is performed under different context conditions (e.g., workday and weekend), which allows differentiating abnormal social behaviors from changes in social habits expected in certain situations. The solution presented is also able to identify changes in social patterns that may be indicative of mental disorders presence. The implementation of the proposed algorithm used the combination of the Frequent Patterns Mining approach with the Complex Events Processing, which allows the realization of the continuous social data stream processing. The evaluation performed demonstrated that context-based recognition provides a better understanding of the social routine, also indicating that the proposed solution is capable of detecting sociability patterns similar to a batch algorithm. Additionally, it was validated the performance of the social behavior change detection mechanism.Item IndoorDSL: uma abordagem baseada em MDE para o desenvolvimento de mapas de representação do espaço indoor(Universidade Federal do Maranhão, 2020-09-24) ANDRADE, Evaldo da Silva; VALE, Samyr Béliche; 71152300300; http://lattes.cnpq.br/1531971102610447; VALE, Samyr Béliche; 71152300300; http://lattes.cnpq.br/1531971102610447; SOARES NETO, Carlos de Salles; http://lattes.cnpq.br/1512846862093142; MOURA, Raimundo Santos; http://lattes.cnpq.br/7902070751612416Indoor space representation models can provide important information about architectural and semantic aspects of a given building. Which enables the use of this information to encourage the development of applications aimed at navigation and positioning within this type of environment, applications that are supported through indoor positioning systems that appear to optimize navigation, or even increase productivity, or aid people with disabilities, among other solutions. However, information about the environment and tools that enable the elaboration of the indoor environment representation model that contain the information that can be used to feed the navigation and positioning systems within those environments are not always available. Either when available they become prohibitive due to the logistical apparatus necessary for the design of the model, or they demand a significant effort through editing tools that combine geographic and architectural information through manual and repetitive editing of primitives that make up this environment, making the task become quite error prone. Therefore, this work presents the proposal of creating indoor space representation through Domain Specific Language, that through a graphical syntax with aspects of the domain, provides the possibility of indoor environment modeling at a high level of abstraction in order to enable users with none or regular knowledge about the domain to create their own models and perform transformations between different models. Therefore its development takes place within the scope of INavigS, a software infrastructure focused on indoor positioning and navigation. Despite providing numerous features, still presents some gaps, as in the case of the elaboration of the indoor space representation model. When applying the language in the context of INavigS, feedback is obtained from users regarding the use of language, identifying it as a satisfactory tool for maps creation which may represent the indoor environment.Item Redes neurais multinível para classificação do ângulo da câmara anterior utilizando Imagens OCT-SA(Universidade Federal do Maranhão, 2021-03-18) FERREIRA, Marcos Melo; BRAZ JUNIOR, Geraldo; http://lattes.cnpq.br/8287861610873629; BRAZ JUNIOR, Geraldo; http://lattes.cnpq.br/8287861610873629; PAIVA, Anselmo Cardoso de; http://lattes.cnpq.br/6446831084215512; ALMEIDA, João Dallyson Sousa de; http://lattes.cnpq.br/6047330108382641; ARAÚJO, Sidnei Alves de; http://lattes.cnpq.br/2542529753132844Glaucoma is identified as one of the main causes of visual impairment, and the main cause of irreversible blindness. The main forms of the disease are primary open-angle glaucoma and primary angle-closure glaucoma. In people with angle-closure glaucoma, the anterior chamber angle narrows, consequently causing an increase in intraocular pressure causing damage to the optic nerve, causing partial or total vision loss. As the damage is irreversible, an early diagnosis is essential, but it is hampered due to the fact that the disease is asymptomatic in early stages. For early detection of the disease, routine imaging tests are recommended, one of which is Anterior Segment Optical Coherence Tomography, which allows an angle classification, which is essential for diagnosis. An analysis of this type of image requires a degree of interpretation on the part of specialists, because of this, the evaluation of many images requires a lot of time, which can lead to professional fatigue. The use of automated methods to assist in the interpretation of images would contribute to get diagnoses more quickly. In this work, an automated method is proposed to classify the anterior chamber angle, present in Anterior Segment images, based on deep learning, using convolutional neural networks. Initially, five pre-trained models of convolutional networks were adjusted to perform feature extraction and classify images. Next, the models were combined in a multilevel architecture, with the objective of increasing the classification capacity. As best result achieved an AUC value (Area Under the Curve) of 0.999.Item Composição de objetos de aprendizagem multimídia através de sumarizadores automáticos de texto baseados em modelos deep learning(Universidade Federal do Maranhão, 2022-09-16) OLIVEIRA, Leandro Massetti Ribeiro; SOARES NETO, Carlos de Salles; http://lattes.cnpq.br/1556965324419871; SOARES NETO, Carlos de Salles; http://lattes.cnpq.br/1556965324419871; OLIVEIRA, Alexandre César Muniz de; http://lattes.cnpq.br/5225588855422632; CARVALHO, Windson Viana de; http://lattes.cnpq.br/1744732999336375A Learning Object (LO) is a digital resource that can be used and reused or referenced during a process of technological support for teaching and learning. Despite being mostly multimedia, with audio, video, text and images synchronized with each other, some digital education resources have texts as one of their main elements in the teaching process, such as websites, texts, video classes, seminars, and the summarization of these texts can be a way of composing multimedia LOs. However, text summarization is a costly process in time and effort, creating the need to seek new ways to generate this content. The present work show a solution for the composition of multimedia LOs through automatic text summarizers based on Deep Learning Transformers models from two experiments: The first one composing LOs from educational texts in Portuguese using translators and text summarizers, in this experiment the results presented were positive and allow comparing the performance of summaries as generators of LO in text format; The second experiment presents an educational video summarization solution using the same Deep Learning models for subtitle summarization, the tests were performed using the EDUVSUM dataset in which it was possible to improve the results of the original article reaching 26.53% accuracy in a multi-class problem and average absolute error of 1.49 per video frame and 1.45 per video segment.Item SysIoTML: uma técnica para modelagem de aplicações no contexto de IoT(Universidade Federal do Maranhão, 2023-05-11) SIQUEIRA, Rodrigo do Nascimento; SANTOS, Davi Viana dos; http://lattes.cnpq.br/9297257833779277; SANTOS, Davi Viana dos; http://lattes.cnpq.br/9297257833779277; TELES, Ariel Soares; http://lattes.cnpq.br/5012476998883237; MARQUES, Anna Beatriz dos Santos; http://lattes.cnpq.br/5522150204610320The Internet of Things (IoT) is a concept that connects smart objects equipped with sensors, networks, and processing technologies that work together to provide an environment in which smart services are brought to users. Systems modeling should be conducted to create IoT Systems and ensure the implementation of a good system. IoT increases the complexity of systems modeling due to novel concepts that need to be addressed. However, there are no established techniques for systems modeling for this specific context. This paper presents the development of a new technique for IoT systems modeling, the SysIoTML, an extension of SysML. Such techniques consider specific aspects of IoT Systems (behavior and interactivity). We proposed the SysIoTML and conducted a concept-proof to analyze the technical feasibility. The developed technique proved useful, and the participants were able to model the proposed problem. The main contribution is to advance IS modeling through a new technique.