We propose two different methods for image segmentation with the objective of marking contaminated regions in images from biochemical tests. The contaminated regions consists of thin hair or fibers and the purpose of this thesis is to eliminate the tedious task of masking the contaminated regions by hand by implementing automatic hair masking. Initially an algorithm based on Morphological Image Processing is presented, followed by solving the problem of pixelwise classification using a Convolutional Neural Network (CNN). Finally, the performance of each implementation is measured by comparing the segmented images with labelled images which are considered to be the ground truth. The result shows that both implementations have strong potential at successfully performing semantic segmentation on the images from the biochemical tests.

Made available in DSpace on 2014-06-11T19:29:40Z (GMT). No. of bitstreams: 0 Previous issue date: 2011-05-10Bitstream added on 2014-06-13T19:38:58Z : No. of bitstreams: 1 rogeri_jg_me_sjrp.pdf: 962940 bytes, checksum: 5f86f6439d28c1cc69d98e55069b9b90 (MD5) / Nos últimos anos, a segurança tornou-se uma preocupação constante da grande maioria das pessoas. Os sistemas biométricos vem ganhando destaque em soluções ligadas à segurança, uma vez que tratam de características físicas e comportamentais para reconhecimento dos indivíduos e permissões de acesso. Este trabalho objetivou a proposição e implementação de um método para reconhecimento de indivíduos por meio de características contidas na região interna da íris com um alto percentual de exatidão no reconhecimento e uma grande diminuição no tempo de processamento, se comparado aos demais métodos encontrados na literatura. No método proposto foram utilizados operadores de morfologia matemática para localização da íris, wavelet de log-Gabor para extração das características e a distância de Hamming para o reconhecimento. Os resultados experimentais obtidos utilizando a base de dados CASIA mostraram que o método é confiável e seguro, além de se destacar com relação ao baixo custo computacional / In the recent years, the security became a constant concern of most people. Biometric systems have been highlighted in solutions related to security, since they deal with physical and behavioral characteristics for individuals recognition and access permissions. This work aims at the implementation of a method for individuals recognition based on the characteristics of the inner region of the iris, seeking a high percentage of accuracy in the recognition and a great reduction in the processing time, as compared to other methods published so far. We use mathematical morphology to search the iris in the image, the log-Gabor wavelet for feature extraction and the Hamming distance for recognition. The experimental results obtained from CASIA database show that the method is safe and reliable, and stand out with regard to the low computational cost

>Magister Scientiae - MSc / This research proposes an approach to recognizing facial expressions in the presence of rotations and partial occlusions of the face. The research is in the context of automatic machine translation of South African Sign Language (SASL) to English. The proposed method is able to accurately recognize frontal facial images at an average accuracy of 75%. It also achieves a high recognition accuracy of 70% for faces rotated to 60◦. It was also shown that the method is able to continue to recognize facial expressions even in the presence of full occlusions of the eyes, mouth and left/right sides of the face. The accuracy was as high as 70% for occlusion of some areas. An additional finding was that both the left and the right sides of the face are required for recognition. As an addition, the foundation was laid for a fully automatic facial expression recognition system that can accurately segment frontal or rotated faces in a video sequence.

Orientador: Aledir Silveira Pereira / Banca: Aparecido Nilceu Marana / Banca: Evandro Luís Linhari Rodrigues / Resumo: Nos últimos anos, a segurança tornou-se uma preocupação constante da grande maioria das pessoas. Os sistemas biométricos vem ganhando destaque em soluções ligadas à segurança, uma vez que tratam de características físicas e comportamentais para reconhecimento dos indivíduos e permissões de acesso. Este trabalho objetivou a proposição e implementação de um método para reconhecimento de indivíduos por meio de características contidas na região interna da íris com um alto percentual de exatidão no reconhecimento e uma grande diminuição no tempo de processamento, se comparado aos demais métodos encontrados na literatura. No método proposto foram utilizados operadores de morfologia matemática para localização da íris, wavelet de log-Gabor para extração das características e a distância de Hamming para o reconhecimento. Os resultados experimentais obtidos utilizando a base de dados CASIA mostraram que o método é confiável e seguro, além de se destacar com relação ao baixo custo computacional / Abstract: In the recent years, the security became a constant concern of most people. Biometric systems have been highlighted in solutions related to security, since they deal with physical and behavioral characteristics for individuals recognition and access permissions. This work aims at the implementation of a method for individuals recognition based on the characteristics of the inner region of the iris, seeking a high percentage of accuracy in the recognition and a great reduction in the processing time, as compared to other methods published so far. We use mathematical morphology to search the iris in the image, the log-Gabor wavelet for feature extraction and the Hamming distance for recognition. The experimental results obtained from CASIA database show that the method is safe and reliable, and stand out with regard to the low computational cost / Mestre

>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.

This master’s thesis is focused on digital image analysis of mitotic chromosomes. It deals with the design of the processing of digital images - from image preprocessing to clasification of each chromosomes, including testing on a set of images. This work introduces used cytogenetic methods, that are used to visualize chromosomes. In its practical part describes morphology operations and clasification procedure. Classification of the chomosomes was divided into 5 groups (A-G). All algorithms were created in the MATLAB program.

In this master thesis, object's detection in image and tracking these objects in temporal area will be presented. First, theoretical background of the image's preprocessing, image filtration, the foreground extraction, and many others various image's features will be described. Next, design and implementation of detector will be processed. This part of my master thesis containes mainly information about detection of objects on belt conveyer Finally,results, conclusion and many supplementary data such as a photography camera's location will be shown.

[pt] Ondas ultrassônicas guiadas são usadas em larga escala em ensaios não destrutivos (END) e Structural Health Monitoring (SHM), permitindo a inspeção de estruturas e equipamentos de forma não invasiva. A partir da transmissão de um sinal acústico sobre uma estrutura e a captação dos sinais de onda propagados por meio de sensores posicionados estrategicamente, é possível obter informações materiais do objeto inspecionado. Na área de óleo e gás, o uso desse tipo de ondas acústicas é de grande importância no levantamento do perfil da camada de cimento que reveste poços, que tem função de conferir integridade estrutural e isolar a estrutura interna de produção do poço das regiões freáticas do entorno. No processo de desativação e abandono do poço, é fundamental avaliar a qualidade do isolamento hidráulico do cimento, assim como identificar possíveis defeitos. A propagação de ondas guiadas em uma estrutura se dá, em geral, por meio de múltiplos modos e apresenta característica dispersiva, que se traduz numa dependência da velocidade de fase das ondas com a frequência, e uma relação não linear entre número de onda e frequência. A relação de dispersão carrega informações do meio de propagação, tal como constantes elásticas e dimensões, e pode ser visualizada a partir de curvas no plano frequência-número de onda (f-k). Diferentes técnicas vêm sendo exploradas para a obtenção das relações de dispersão a partir de sinais no domínio do tempo captados por sensores ultrassônicos em posições espaciais distintas. Este trabalho explora três métodos distintos para a extração das curvas de dispersão, ou seja, obter os pontos f-k associados aos modos de propagação, a partir de um conjunto de sinais dependentes do espaço-tempo. O primeiro algoritmo se baseia em uma técnica pré-existente que usa uma Transformada de Fourier bidimensional (2-D FT) sobre a matriz de dados de sinais de sensores ultrassônicos no espaço-tempo, gerando uma matriz de amplitudes no plano f-k onde os máximos locais representam pontos pertencentes a curvas de dispersão. A representação da matriz como uma imagem f-k permite a visualização das curvas de dispersão como conjuntos contíguos de pixels de maior claridade. Propõe-se um novo algoritmo baseado em operações morfológicas de processamento de imagem para a identificação de pixels relativos aos pontos das curvas de dispersão na imagem f-k, após um préprocessamento da mesma. A segunda técnica consiste no pré-processamento dessa mesma imagem f-k, obtida pela 2-D FT, e uso de um algoritmo préexistente de detecção de estruturas curvilíneas em imagens para identificar os pontos correspondentes às curvas f-k. O terceiro método é uma adaptação, proposta aqui, de um algoritmo pré-existente para estimar os números de onda das curvas de dispersão relativos a cada frequência através de uma matriz Pencil. Propõe-se também um algoritmo original para a separação dos pontos f-k encontrados pelas três técnicas de extração em curvas distintas. Os algoritmos utilizados para a obtenção das curvas de dispersão têm seu desempenho avaliado em três conjuntos de dados distintos de simulações por elementos finitos, a saber, uma de placa de alumínio fina sob distintos valores de tração axial aplicada paralelamente à direção de propagação das ondas; um poço multicamada sem tubing possuindo diferentes tipos de defeito de cimentação-channeling, qualidade de cimento baixa, descolamento interno e externo -, assim como sem defeito; e um pouco multicamada com tubing sob os mesmos defeitos de cimentação e também sem defeito. Compara-se a capacidade dos algoritmos de extração das curvas de dispersão de oferecer informações sobre mudanças materiais entre os casos simulados. Avalia-se também a precisão e custo computacional dos mesmos. / [en] Ultrasonic guided waves are widely used in the fields of Non-Destructive Evaluation (END) and Structural Health Monitoring (SHM), allowing the inspection of structures and pieces of equipment in a non-invasive manner. Through the transmission of an acoustic signal over a given object and the acquisition of the signal from the propagated waves using a group of sensors in predefined positions, it is possible to obtain material information regarding the investigated structure. In the Oil & Gas industry, the use of this type of wave is integral to the logging of the cement layer that outlines the walls of wellbores, which has the purpose of guaranteeing structural support and protecting the well’s internal production structure and the surrounding groundwater from each other. During the deactivation and abandonment of a production well, it is necessary to evaluate the hydraulic isolation of the cement layer, as well as identify possible defects. The propagation of guided waves in a structure is usually multi-modal and of dispersive characteristic. The latter means that the propagating waves phase velocity is dependent on the frequency, translating into a non-linear relationship between wavenumber and frequency. This dispersion relation contains information about the propagating medium, such as elastic constants and dimensions, and can be represented as curves in the frequency-wavenumber (f-k) plane. Different methods are currently being explored for obtaining the dispersion relation from time-domain signals acquired by ultrasonic sensors in different spatial positions. This work explored three different methods for the extraction of the dispersion curves, that is, obtaining the f-k points associated with the modes of propagation, from a dataset composed of space-time signals. The first algorithm is based on a pre-existing technique that uses the bidimensional Fourier Transform (2-D FT) over the matrix containing the space-time signals from the ultrasonic sensors, generating an f-k matrix whose local maximas correspond to points belonging to dispersions curves. The representation of the matrix as an f-k image shows the dispersion curves as contiguous groups of pixels with elevated brightness. A new algorithm is proposed, based on morphological operations from image-processing, to identify the pixels relative to the f-k points of the dispersion curves in the image, after pre-processing is performed. The second technique consists of pre-processing the same fk image, obtained from the 2-D FT, and the use of an existing algorithm for the detection of curvilinear structures in images to identify the points corresponding to the f-k curves. The third method proposes the adaptation of an existing method of estimation of the wavenumbers associated with the dispersion curves for different frequencies, using a matrix Pencil. This work also proposes an original algorithm to separate the f-k points, retrieved by the three techniques, in different curves associated with each mode of propagation. The algorithms used here for the estimation of the dispersion curves are evaluated over three distinct datasets of finite elements simulation: a thin aluminum plate under different values of axial traction parallel to the direction of propagation of the waves; a multilayer wellbore without tubing, with different types of cement defects-channeling, low cement quality, internal and external decoupling-, and without defect; a multilayer wellbore with tubing with the same cement defects and with no defect. Finally, a comparison is drawn over the capacity of the extraction algorithms of providing information regarding changes in the material qualities of the simulated objects. The work also evaluates the precision and computational performance of the aforementioned algorithms.

This thesis aims to investigate a fault detection and diagnosis system for automotive cameras using unsupervised learning. 1) Can a front-looking wide-angle camera image dataset be created using Hardware-in-Loop (HIL) simulations? 2) Can an Adversarial Autoencoder (AAE) based unsupervised camera fault detection and diagnosis method be crafted for SPA2 Vehicle Control Unit (VCU) using an image dataset created using Hardware-inLoop? 3) Does using AAE surpass the performance of using Variational Autoencoder (VAE) for the unsupervised automotive camera fault diagnosis model? In the field of camera fault studies, automotive cameras stand out for its complex operational context, particularly in Advanced Driver-Assistance Systems (ADAS) applications. The literature review finds a notable gap in comprehensive image datasets addressing the image artefact spectrum of ADAS-equipped automotive cameras under real-world driving conditions. In this study, normal and fault scenarios for automotive cameras are defined leveraging published and company studies and a fault diagnosis model using unsupervised learning is proposed and examined. The types of image faults defined and included are Lens Flare, Gaussian Noise and Dead Pixels. Along with normal driving images, a balanced fault-injected image dataset is collected using real-time sensor simulation under driving scenario with industrially-recognised HIL setup. An AAE-based unsupervised automotive camera fault diagnosis system using VGG16 as encoder-decoder structure is proposed and experiments on its performance are conducted on both the selfcollected dataset and fault-injected KITTI raw images. For non-processed KITTI dataset, morphological operations are examined and are employed as preprocessing. The performance of the system is discussed in comparison to supervised and unsupervised image partition methods in related works. The research found that the AAE method outperforms popular VAE method, using VGG16 as encoder-decoder structure significantly using 3-layer Convolutional Neural Network (CNN) and ResNet18 and morphological preprocessings significantly ameliorate system performance. The best performing VGG16- AAE model achieves 62.7% accuracy to diagnosis on own dataset, and 86.4% accuracy on double-erosion-processed fault-injected KITTI dataset. In conclusion, this study introduced a novel scheme for collecting automotive sensor data using Hardware-in-Loop, utilised preprocessing techniques that enhance image partitioning and examined the application of unsupervised models for diagnosing faults in automotive cameras. / Denna avhandling syftar till att undersöka ett felupptäcknings- och diagnossystem för bilkameror med hjälp av oövervakad inlärning. De huvudsakliga forskningsfrågorna är om en bilduppsättning från en frontmonterad vidvinkelkamera kan skapas med hjälp av Hardware-in-Loop (HIL)-simulationer, om en Adversarial Autoencoder (AAE)-baserad metod för oövervakad felupptäckt och diagnos för SPA2 Vehicle Control Unit (VCU) kan utformas med en bilduppsättning skapad med Hardware-in-Loop, och om användningen av AAE skulle överträffa prestandan av att använda Variational Autoencoder (VAE) för den oövervakade modellen för felanalys i bilkameror. Befintliga studier om felanalys fokuserar på roterande maskiner, luftbehandlingsenheter och järnvägsfordon. Få studier undersöker definitionen av feltyper i bilkameror och klassificerar normala och felaktiga bilddata från kameror i kommersiella passagerarfordon. I denna studie definieras normala och felaktiga scenarier för bilkameror och en modell för felanalys med oövervakad inlärning föreslås och undersöks. De typer av bildfel som definieras är Lens Flare, Gaussiskt brus och Döda pixlar. Tillsammans med normala bilder samlas en balanserad uppsättning felinjicerade bilder in med hjälp av realtidssensor-simulering under körscenarier med industriellt erkänd HIL-uppsättning. Ett AAE-baserat system för oövervakad felanalys i bilkameror med VGG16 som kodaredekoderstruktur föreslås och experiment på dess prestanda genomförs både på den självinsamlade uppsättningen och felinjicerade KITTI-raw-bilder. För icke-behandlade KITTI-uppsättningar undersöks morfologiska operationer och används som förbehandling. Systemets prestanda diskuteras i jämförelse med övervakade och oövervakade bildpartitioneringsmetoder i relaterade arbeten. Forskningen fann att AAE-metoden överträffar den populära VAEmetoden, genom att använda VGG16 som kodare-dekoderstruktur signifikant med ett 3-lagers konvolutionellt neuralt nätverk (CNN) och ResNet18 och morfologiska förbehandlingar förbättrar systemets prestanda avsevärt. Den bäst presterande VGG16-AAE-modellen uppnår 62,7 % noggrannhet för diagnos på egen uppsättning, och 86,4 % noggrannhet på dubbelerosionsbehandlad felinjicerad KITTI-uppsättning. Sammanfattningsvis introducerade denna studie ett nytt system för insamling av data från bilsensorer med Hardware-in-Loop, utnyttjade förbehandlingstekniker som förbättrar bildpartitionering och undersökte tillämpningen av oövervakade modeller för att diagnostisera fel i bilkameror.