Made available in DSpace on 2016-08-17T14:53:15Z (GMT). No. of bitstreams: 1 Daniel Rodrigues Ericeira.pdf: 2002346 bytes, checksum: df76ac081a5d0e5816a81b5699935561 (MD5) Previous issue date: 2011-03-17 / Mammography is the exam of the breast, used as breast cancer prevention and also as a diagnostic method. This exam, which consists in an X-Ray of the breast, allows cancer detection. The purpose of this work is to use image processing techniques and computer vision to help specialists in detecting suspect regions and masses in digital mammographies. The first stage of the methodology consists in pre-processing the images to make them more suitable to registration, through noise reduction, image segmentation and re-scale. The next stage presents bilateral left and right breast image pairs registration. In order to correct position and compression differences that occur during the exams, rigid registration (followed by optic flow deformable registration) was applied in each image pair. Corresponding pairs of regions were related and their mutual variations were measured through cross-variogram spatial description. On the next stage, a training model for a Support Vector Machine (SVM) was created using as characteristics the cross-variogram values of each pair of regions of 180 cases. This SVM was tested for 100 new cases. The region pairs that contained lesions were classified as suspect regions , and the other regions as non-suspect regions . From the suspect regions, variogram characteristics were extracted as tissue texture descriptors. The regions that contained masses were classified as mass regions and the other regions as non-mass regions . Stepwise linear discriminant analysis was applied to select the most significant characteristics to train the second SVM. Tests with 30 new cases were performed for the trained SVM final classification in mass or non-mass . The best case presented on the final classification: 96% accuracy, 100% sensitivity and 95,34% specificity. The worst case presented: 70% accuracy, 100% sensitivity and 67,56% specificity. On average, the 30 cases presented: 90% accuracy, 100% sensitivity and 85% specificity. / A mamografia é um exame de mama, utilizado de forma preventiva ao câncer de mama e também como método diagnóstico. Este exame, que consiste em uma radiografia das mamas, permite a detecção do câncer. O objetivo deste trabalho é utilizar técnicas de processamento de imagens e visão computacional para auxiliar especialistas na detecção de regiões suspeitas e detecção de massas mamárias em mamografias digitais. A primeira etapa da metodologia consiste em pré-processar as imagens de forma a torná-las mais apropriadas ao registro, através de redução de ruído, segmentação e re-dimensionamento. A etapa seguinte apresenta o registro bilateral de pares de mamas esquerda e direita. Para corrigir as diferenças de posicionamento e compressão ocorridas no momento do exame, o método de registro rígido foi aplicado (seguido do método de registro deformável com fluxo óptico) para cada par de imagens. Pares de regiões correspondentes foram relacionados e suas variações foram medidas através do descritor espacial variograma cruzado. Na etapa seguinte, foi criado um modelo para treinamento de uma Máquina de Vetores de Suporte (MVS) utilizando como características os valores de variograma cruzado de cada par de janelas de 180 casos. Esta MVS foi testada em 100 novos casos. Os pares que continham lesões foram classificados como regiões suspeitas ; as demais, como regiões não-suspeitas . Destas regiões suspeitas, foram extraídas características de variograma como descritores de textura de tecido. As regiões que continham massas foram classificadas como regiões de massa e as demais como regiões de não-massa . Análise linear discriminante stepwise foi aplicada para selecionar as características mais significativas para treinamento de uma segunda MVS. Foram realizados testes com 30 novos casos para a classificação final pela MVS treinada em massa e nãomassa . O melhor resultado apresentou na classificação final: 96% de acurácia, 100% de sensibilidade e 95,34% de especificidade. O pior caso apresentou: 70% de acurácia, 100% de sensibilidade e 67,56% de especificidade. Em média, os 30 casos apresentaram: 90% de acurácia, 100% de sensibilidade e 85% de especificidade.

Made available in DSpace on 2016-08-17T14:53:21Z (GMT). No. of bitstreams: 1 Peterson.pdf: 1362910 bytes, checksum: 963fec328036941a0790b198cc0d6187 (MD5) Previous issue date: 2012-04-20 / FUNDAÇÃO DE AMPARO À PESQUISA E AO DESENVOLVIMENTO CIENTIFICO E TECNOLÓGICO DO MARANHÃO / Breast cancer is the second most common in the world and which more affects women. In recent years, several Computer Aided Detection/Diagnosis Systems has been developed in order to assist health specialists in the detection and diagnosis of cancer, serving as a second opinion. The aim of this paper is to present a methodology for discrimination and classification of regions extracted from mammograms in mass and non-mass. In this study, Digital Database for Screening Mammography (DDSM) is used. To describe the texture of the region of interest is applied McIntosh Diversity Index, commonly used in ecology. The calculation of this index is proposed in four approaches: through the Histogram, through the Gray Level Co-occurrence Matrix, through the Gray Level Run Length Matrix and through the Gray Level Gap Length Matrix. For the classification of regions in mass and non-mass, is used the supervised classificator Support Vector Machine (SVM). The methodology shows promising results for the classification of masses and non-masses, reaching an accuracy of 93,68%. / O câncer de mama é o segundo tipo de câncer mais frequente no mundo e o que mais acomete as mulheres. Nos últimos anos, vários Sistemas de Detecção e Diagnóstico auxiliados por Computador (Computer Aided Detection/Diagnosis) têm sido desenvolvidos no intuito de auxiliar especialistas da área da saúde na detecção e diagnóstico de câncer, servindo como uma segunda opnião. O objetivo deste trabalho é apresentar uma metodologia de discriminação e classificação de regiões extraídas de mamografias em massa e não massa. Neste estudo, o Digital Database for Screening Mammography (DDSM) é usado. Para descrever a textura da região de interesse é aplicado o Índice de Diversidade de McIntosh, comumente usado em ecologia. O cálculo deste índice é proposto em quatro abordagens: através do Histograma, da Matriz de Co-ocorrência de Níveis de Cinza, da Matriz de Comprimentos de Corrida de Cinza e da Matriz de Comprimentos de Lacuna de Cinza. Para classificação das regiões em massa e não massa, é utilizado o classificador supervisionado Support Vector Machine (SVM). A metodologia apresenta resultados promissores para a classificação de massas e não massas, alcançando uma acurácia de 93,68%.

Made available in DSpace on 2016-08-17T14:53:22Z (GMT). No. of bitstreams: 1 dissertacao Luis Claudio.pdf: 2055680 bytes, checksum: 4e51c93fd4aa6edfad51ab54ad4044f6 (MD5) Previous issue date: 2012-07-24 / This thesis proposes the modeling and implementation of a telediagnostic system for analysis and detection of lesions in mammographic images based on independent component analysis and support vector machine. The system analyzes images from digital mammography sent over the Internet and provides a diagnostic, indicating the presence of suspicious regions, which can be confirmed by a specialist in mammographic images. Besides presenting the methodology for the development of the proposed system, a prototype was developed for testing and to measure its efficiency. The database used for training and testing of the algorithms is the mini-MIAS, and was employed independent component analysis to extract the filters used in segmenting the regions of interest, as well support vector machine to classify regions of interest in normal or suspicious. From tests with the database used, we obtained an average accuracy of 87.8% for images containing lesions. / Este trabalho propõe a modelagem e implementação de um sistema de telediagnóstico para análise e detecção automática de lesões em imagens mamográficas, baseado em análise de componentes independentes e máquina de vetor de suporte. O sistema analisa imagens de mamografia digital enviadas pela Internet e fornece um diagnóstico da imagem, indicando a presença de regiões suspeitas, que podem ser confirmadas por um especialista em imagens mamográficas. Além de apresentar a metodologia para o desenvolvimento do sistema proposto, foi desenvolvido um protótipo para a realização de testes objetivando medir sua eficiência. A base de dados usada para treinamento e teste dos algoritmos foi a mini-MIAS, e foi empregada análise de componentes independentes para extrair os filtros usados na segmentação das regiões de interesse, bem como máquina de vetor de suporte para classificar as regiões de interesse em normais ou suspeitas. A partir de testes realizados com a base de dados utilizada, obteve-se média de acerto de 87,8% para imagens que contém lesões.

Orientador: Janito Vaqueiro Ferreira / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Mecânica / Made available in DSpace on 2018-08-23T16:12:48Z (GMT). No. of bitstreams: 1 ArcilaGago_ManuelFelipe_M.pdf: 7192169 bytes, checksum: 7b23a08769656a07765344e20d1f6ad4 (MD5) Previous issue date: 2013 / Resumo: A variedade de materiais no setor aeronáutico para redução de peso e custo tem se proliferado a um grau intensivo, onde têm sido revisadas diferentes pesquisas para encontrar outros tipos de materiais de fácil maneabilidade para construção de peças que satisfazem as restrições impostas. Assim, existe uma procura constante de soluções para facilitar a produção, e ao mesmo tempo aumentar a segurança das aeronaves levando em consideração pontos importantes como a fadiga e ruptura do material. Um material frequentemente utilizado que atende a estes requisitos devido a suas propriedades de densidade e resistência é o alumínio, e é neste ambiente que existe um processo de manufatura utilizado para a soldagem conhecido como "Friction Stir Welding" (FSW). No presente momento, estudos para criação de modelos que representem características mecânicas utilizadas em projetos em função de parâmetros do processo tem sido pesquisados. Embora este processo seja de difícil modelagem devidos as suas complexidades, tem sido estudado e utilizado diferentes algoritmos que possibilitem o melhoramento da representação do modelo, tais como os relacionados com máquinas de aprendizagem (ML) e suas diferentes otimizações. Neste contexto, a presente pesquisa tem seu foco na obtenção de um modelo baseado no algoritmo de aprendizagem de Máquina de Vetores de Suporte (SVM), e também com outros algoritmos tais como Regressão Polinomial (RP) e Rede Neural Artificial (RNA), buscando encontrar modelos que representem o processo de soldagem por FSW através das propriedades mecânicas obtidas pelos ensaios de tração e por análise de variância (ANOVA), entendendo suas vantagens e, posteriormente, recomendar quais dos algoritmos de aprendizagem tem maior beneficio / Abstract: In the aerospace industry to reduce weight and cost, a great quantity of materials has been used, which has generated research to find types of materials, that have been better maneuverability and to guarantee the properties required to development of pieces for the industry. Thus, the studies look for optimize between production easiness and increase the aircraft safety, taking into consideration important issues such as fatigue and fracture of the materials. One of the most common approach used is aluminum by their mechanical properties (density and strength), although it has many problems to be welding with the traditional methods. Currently, the Friction Stir Welding (FSW) process is used in the industry, as well in the academy. However, the FSW is difficult to model by the complexities in the physical phenomenal occurred during the weld process, as result, has been studied and used different algorithms that allow enhance the model representation. The Machine Learning (ML) is a methodology studied to obtain the model optimized. In this context, the present research focus by to obtain a model-based in learning algorithm using Support Vector Machine (SVM). Although comparisons were made with other algorithms such as Polynomial Regression (PR) and Artificial Neural Network (ANN), searching to find models that represent the FSW process weld using the mechanical properties obtained by tensile tests and analysis of variance (ANOVA). Finally, conclusions to understand the advantages learning algorithms are presented / Mestrado / Mecanica dos Sólidos e Projeto Mecanico / Mestre em Engenharia Mecânica

L'utilisation des technologies haut débit est en train de changer aussi bien les pratiques que le paysage scientifique en microbiologie. D'une part la spectrométrie de masse a d'ores et déjà fait son entrée avec succès dans les laboratoires de microbiologie clinique. D'autre part, l'avancée spectaculaire des technologies de séquençage au cours des dix dernières années permet désormais à moindre coût et dans un temps raisonnable de caractériser la diversité microbienne au sein d'échantillons cliniques complexes. Aussi ces deux technologies sont pressenties comme les piliers de futures solutions de diagnostic. L'objectif de cette thèse est de développer des méthodes d'apprentissage statistique innovantes et versatiles pour exploiter les données fournies par ces technologies haut-débit dans le domaine du diagnostic in vitro en microbiologie. Le domaine de l'apprentissage statistique fait partie intégrante des problématiques mentionnées ci-dessus, au travers notamment des questions de classification d'un spectre de masse ou d'un “read” de séquençage haut-débit dans une taxonomie bactérienne.Sur le plan méthodologique, ces données nécessitent des développements spécifiques afin de tirer au mieux avantage de leur structuration inhérente: une structuration en “entrée” lorsque l'on réalise une prédiction à partir d'un “read” de séquençage caractérisé par sa composition en nucléotides, et un structuration en “sortie” lorsque l'on veut associer un spectre de masse ou d'un “read” de séquençage à une structure hiérarchique de taxonomie bactérienne. / Using high-throughput technologies is changing scientific practices and landscape in microbiology. On one hand, mass spectrometry is already used in clinical microbiology laboratories. On the other hand, the last ten years dramatic progress in sequencing technologies allows cheap and fast characterization of microbial diversity in complex clinical samples. Consequently, the two technologies are approached in future diagnostics solutions. This thesis aims to play a part in new in vitro diagnostics (IVD) systems based on high-throughput technologies, like mass spectrometry or next generation sequencing, and their applications in microbiology.Because of the volume of data generated by these new technologies and the complexity of measured parameters, we develop innovative and versatile statistical learning methods for applications in IVD and microbiology. Statistical learning field is well-suited for tasks relying on high-dimensional raw data that can hardly be used by medical experts, like mass-spectrum classification or affecting a sequencing read to the right organism. Here, we propose to use additional known structures in order to improve quality of the answer. For instance, we convert a sequencing read (raw data) into a vector in a nucleotide composition space and use it as a structuredinput for machine learning approaches. We also add prior information related to the hierarchical structure that organizes the reachable micro-organisms (structured output).

Current Ambient Intelligence and Intelligent Environment research focuses on the interpretation of a subject’s behaviour at the activity level by logging the Activity of Daily Living (ADL) such as eating, cooking, etc. In general, the sensors employed (e.g. PIR sensors, contact sensors) provide low resolution information. Meanwhile, the expansion of ubiquitous computing allows researchers to gather additional information from different types of sensor which is possible to improve activity analysis. Based on the previous research about sitting posture detection, this research attempts to further analyses human sitting activity. The aim of this research is to use non-intrusive low cost pressure sensor embedded chair system to recognize a subject’s activity by using their detected postures. There are three steps for this research, the first step is to find a hardware solution for low cost sitting posture detection, second step is to find a suitable strategy of sitting posture detection and the last step is to correlate the time-ordered sitting posture sequences with sitting activity. The author initiated a prototype type of sensing system called IntelliChair for sitting posture detection. Two experiments are proceeded in order to determine the hardware architecture of IntelliChair system. The prototype looks at the sensor selection and integration of various sensor and indicates the best for a low cost, non-intrusive system. Subsequently, this research implements signal process theory to explore the frequency feature of sitting posture, for the purpose of determining a suitable sampling rate for IntelliChair system. For second and third step, ten subjects are recruited for the sitting posture data and sitting activity data collection. The former dataset is collected byasking subjects to perform certain pre-defined sitting postures on IntelliChair and it is used for posture recognition experiment. The latter dataset is collected by asking the subjects to perform their normal sitting activity routine on IntelliChair for four hours, and the dataset is used for activity modelling and recognition experiment. For the posture recognition experiment, two Support Vector Machine (SVM) based classifiers are trained (one for spine postures and the other one for leg postures), and their performance evaluated. Hidden Markov Model is utilized for sitting activity modelling and recognition in order to establish the selected sitting activities from sitting posture sequences.2. After experimenting with possible sensors, Force Sensing Resistor (FSR) is selected as the pressure sensing unit for IntelliChair. Eight FSRs are mounted on the seat and back of a chair to gather haptic (i.e., touch-based) posture information. Furthermore, the research explores the possibility of using alternative non-intrusive sensing technology (i.e. vision based Kinect Sensor from Microsoft) and find out the Kinect sensor is not reliable for sitting posture detection due to the joint drifting problem. A suitable sampling rate for IntelliChair is determined according to the experiment result which is 6 Hz. The posture classification performance shows that the SVM based classifier is robust to “familiar” subject data (accuracy is 99.8% with spine postures and 99.9% with leg postures). When dealing with “unfamiliar” subject data, the accuracy is 80.7% for spine posture classification and 42.3% for leg posture classification. The result of activity recognition achieves 41.27% accuracy among four selected activities (i.e. relax, play game, working with PC and watching video). The result of this thesis shows that different individual body characteristics and sitting habits influence both sitting posture and sitting activity recognition. In this case, it suggests that IntelliChair is suitable for individual usage but a training stage is required.

>Magister Scientiae - MSc / The SASL project is in the process of developing a machine translation system that can translate fully-fledged phrases between SASL and English in real-time. To-date, several systems have been developed by the project focusing on facial expression, hand shape, hand motion, hand orientation and hand location recognition and estimation. Achmed developed a highly accurate upper body pose recognition and estimation system. The system is capable of recognizing and estimating the location of the arms from a twodimensional video captured from a monocular view at an accuracy of 88%. The system operates at well below real-time speeds. This research aims to investigate the use of optimizations and parallel processing techniques using the CUDA framework on Achmed’s algorithm to achieve real-time upper body pose recognition and estimation. A detailed analysis of Achmed’s algorithm identified potential improvements to the algorithm. Are- implementation of Achmed’s algorithm on the CUDA framework, coupled with these improvements culminated in an enhanced upper body pose recognition and estimation system that operates in real-time with an increased accuracy.

Magister Scientiae - MSc / Research has shown that five fundamental parameters are required to recognize any sign language gesture: hand shape, hand motion, hand location, hand orientation and facial expressions. The South African Sign Language (SASL) research group at the University of the Western Cape (UWC) has created several systems to recognize sign language gestures using single parameters. These systems are, however, limited to a vocabulary size of 20 – 23 signs, beyond which the recognition accuracy is expected to decrease. The first aim of this research is to investigate the use of two parameters – hand motion and hand shape – to recognise a larger vocabulary of SASL gestures at a high accuracy. Also, the majority of related work in the field of sign language gesture recognition using these two parameters makes use of Hidden Markov Models (HMMs) to classify gestures. Hidden Markov Support Vector Machines (HM-SVMs) are a relatively new technique that make use of Support Vector Machines (SVMs) to simulate the functions of HMMs. Research indicates that HM-SVMs may perform better than HMMs in some applications. To our knowledge, they have not been applied to the field of sign language gesture recognition. This research compares the use of these two techniques in the context of SASL gesture recognition. The results indicate that, using two parameters results in a 15% increase in accuracy over the use of a single parameter. Also, it is shown that HM-SVMs are a more accurate technique than HMMs, generally performing better or at least as good as HMMs.

The radar transmitter identification problem involves the identification of a specific radar transmitter based on a received pulse. The radar transmitters are of identical make and model. This makes the problem challenging since the differences between radars of identical make and model will be solely due to component tolerances and variation. Radar pulses also vary in time and frequency which means that the problem is non-stationary. Because of this fact, time-frequency representations such as shift-invariant quadratic time-frequency representations (Cohen’s class) and wavelets were used. A model for a radar transmitter was developed. This consisted of an analytical solution to a pulse-forming network and a linear model of an oscillator. Three signal classification algorithms were developed. A signal classifier was developed that used a radially Gaussian Cohen’s class transform. This time-frequency representation was refined to increase the classification accuracy. The classification was performed with a support vector machine classifier. The second signal classifier used a wavelet packet transform to calculate the feature values. The classification was performed using a support vector machine. The third signal classifier also used the wavelet packet transform to calculate the feature values but used a Universum type classifier for classification. This classifier uses signals from the same domain to increase the classification accuracy. The classifiers were compared against each other on a cubic and exponential chirp test problem and the radar transmitter model. The classifier based on the Cohen’s class transform achieved the best classification accuracy. The classifier based on the wavelet packet transform achieved excellent results on an Electroencephalography (EEG) test dataset. The complexity of the wavelet packet classifier is significantly lower than the Cohen’s class classifier. Copyright / Dissertation (MEng)--University of Pretoria, 2010. / Electrical, Electronic and Computer Engineering / unrestricted

Canadian remote areas are characterized by a minimal human footprint, restricted accessibility, ubiquitous lichen/snow cover (e.g. Arctic) or continuous forest with water bodies (e.g. Sub-Arctic). Effective mapping of earth surface cover and land cover changes using free medium-resolution Landsat images in remote environments is a challenge due to the presence of spectrally mixed pixels, restricted field sampling and ground truthing, and the often relatively homogenous cover in some areas. This thesis investigates how remote sensing methods can be applied to improve the capability of Landsat images for mapping earth surface features and land cover changes in Canadian remote areas. The investigation is conducted from the following four perspectives: 1) determining the continuity of Landsat-8 images for mapping surficial materials, 2) selecting classification algorithms that best address challenges involving mixed pixels, 3) applying advanced image fusion algorithms to improve Landsat spatial resolution while maintaining spectral fidelity and reducing the effects of mixed pixels on image classification and change detection, and, 4) examining different change detection techniques, including post-classification comparisons and threshold-based methods employing PCA(Principal Components Analysis)-fused multi-temporal Landsat images to detect changes in Canadian remote areas. Three typical landscapes in Canadian remote areas are chosen in this research. The first is located in the Canadian Arctic and is characterized by ubiquitous lichen and snow cover. The second is located in the Canadian sub-Arctic and is characterized by well-defined land features such as highlands, ponds, and wetlands. The last is located in a forested highlands region with minimal built-environment features. The thesis research demonstrates that the newly available Landsat-8 images can be a major data source for mapping Canadian geological information in Arctic areas when Landsat-7 is decommissioned. In addition, advanced classification techniques such as a Support-Vector-Machine (SVM) can generate satisfactory classification results in the context of mixed training data and minimal field sampling and truthing. This thesis research provides a systematic investigation on how geostatistical image fusion can be used to improve the performance of Landsat images in identifying surface features. Finally, SVM-based post-classified multi-temporal, and threshold-based PCA-fused bi-temporal Landsat images are shown to be effective in detecting different aspects of vegetation change in a remote forested region in Ontario. This research provides a comprehensive methodology to employ free Landsat images for image classification and change detection in Canadian remote regions.