This thesis is a part of a project where a new method, the base descriptor approach, is studied. The purpose of this method is to reduce drift and extract vital information from electronic tongue measurements. Reference solutions, called descriptors, are measured and the measurements are used to find base descriptors. A base descriptor is, in this thesis, a regression vector for prediction of the property that the descriptor represent. The property is in this case the concentration of a chemical substance in the descriptor solution. Measurements from test samples, in this case fruit juices, are projected onto the base descriptors to extract vital and drift-stable information from the test samples. The base descriptors are used to determine the concentrations of the descriptors'chemical substances in the juices and thereby also to classify the different juices. It is assumed that the measurements of samples of juices and descriptors drift the same way. This assumption has to be true in order for the base descriptor approach to work. The base descriptors are calculated by multivariate regression methods like partial least squares regression (PLSR) and principal component regression (PCR). Only two of the descriptors tested in this thesis worked as basis for base descriptors. The base descriptors'predictions of the concentrations of chemical substances in the juices are hard to evaluate since the true concentrations are unknown. Comparing the projections of juice measurements onto the base descriptors with a classification model on the juice measurements performed by principal component analysis (PCA), there is no significant difference in drift of the juice measurements in the results of the two methods. The base descriptors, however, separates the juices for classification somewhat better than the classification of juices performed by PCA.

In pattern recognition, &ldquo / bias-variance&rdquo / trade-off is a challenging issue that the scientists has been working to get better generalization performances over the last decades. Among many learning methods, two-layered homogeneous stacked generalization has been reported to be successful in the literature, in different problem domains such as object recognition and image annotation. The aim of this work is two-folded. First, the problems of stacked generalization are attacked by a proposed novel architecture. Then, a set of success criteria for stacked generalization is studied. A serious drawback of stacked generalization architecture is the sensitivity to curse of dimensionality problem. In order to solve this problem, a new architecture named &ldquo / unanimous decision&rdquo / is designed. The performance of this architecture is shown to be comparably similar to two layered homogeneous stacked generalization architecture in low number of classes while it performs better than stacked generalization architecture in higher number of classes. Additionally, a new success criterion for two layered homogeneous stacked generalization architecture is proposed based on the individual properties of the used descriptors and it is verified in synthetic datasets.

Automatic image annotation is the process of assigning keywords to digital images depending on the content information. In one sense, it is a mapping from the visual content information to the semantic context information. In this thesis, we propose a novel approach for automatic image annotation problem, where the annotation is formulated as a multivariate mapping from a set of independent descriptor spaces, representing a whole image, to a set of words, representing class labels. For this purpose, a hierarchical annotation architecture, named as HANOLISTIC (Hierarchical Image Annotation System Using Holistic Approach), is dened with two layers. At the rst layer, called level-0 annotator, each annotator is fed by a set of distinct descriptor, extracted from the whole image. This enables us to represent the image at each annotator by a dierent visual property of a descriptor. Since, we use the whole image, the problematic segmentation process is avoided. Training of each annotator is accomplished by a supervised learning paradigm, where each word is represented by a class label. Note that, this approach is slightly dierent then the classical training approaches, where each data has a unique label. In the proposed system, since each image has one or more annotating words, we assume that an image belongs to more than one class. The output of the level-0 annotators indicate the membership values of the words in the vocabulary, to belong an image. These membership values from each annotator is, then, aggregated at the second layer by using various rules, to obtain meta-layer annotator. The rules, employed in this study, involves summation and/or weighted summation of the output of layer-0 annotators. Finally, a set of words from the vocabulary is selected based on the ranking of the output of meta-layer. The hierarchical annotation system proposed in this thesis outperforms state of the art annotation systems based on segmental and holistic approaches. The proposed system is examined in-depth and compared to the other systems in the literature by means of using several performance criteria.

Image descriptors play an important role in image representation and analysis. Multi-resolution image descriptors can effectively characterize complex images and extract their hidden information. Wavelets descriptors have been widely used in multi-resolution image analysis. However, making the wavelets transform shift and rotation invariant produces redundancy and requires complex matching processes. As to other multi-resolution descriptors, they usually depend on other theories or information, such as filtering function, prior-domain knowledge, etc.; that not only increases the computation complexity, but also generates errors. We propose a novel multi-resolution scheme that is capable of transforming any kind of image descriptor into its multi-resolution structure with high computation accuracy and efficiency. Our multi-resolution scheme is based on sub-sampling an image into an odd-even image tree. Through applying image descriptors to the odd-even image tree, we get the relative multi-resolution image descriptors. Multi-resolution analysis is based on downsampling expansion with maximum energy extraction followed by upsampling reconstruction. Since the maximum energy usually retained in the lowest frequency coefficients; we do maximum energy extraction through keeping the lowest coefficients from each resolution level. Our multi-resolution scheme can analyze images recursively and effectively without introducing artifacts or changes to the original images, produce multi-resolution representations, obtain higher resolution images only using information from lower resolutions, compress data, filter noise, extract effective image features and be implemented in parallel processing.

Viruses and their morphology have been detected and studied with electron microscopy (EM) since the end of the 1930s. The technique has been vital for the discovery of new viruses and in establishing the virus taxonomy. Today, electron microscopy is an important technique in clinical diagnostics. It both serves as a routine diagnostic technique as well as an essential tool for detecting infectious agents in new and unusual disease outbreaks. The technique does not depend on virus specific targets and can therefore detect any virus present in the sample. New or reemerging viruses can be detected in EM images while being unrecognizable by molecular methods. One problem with diagnostic EM is its high dependency on experts performing the analysis. Another problematic circumstance is that the EM facilities capable of handling the most dangerous pathogens are few, and decreasing in number. This thesis addresses these shortcomings with diagnostic EM by proposing image analysis methods mimicking the actions of an expert operating the microscope. The methods cover strategies for automatic image acquisition, segmentation of possible virus particles, as well as methods for extracting characteristic properties from the particles enabling virus identification. One discriminative property of viruses is their surface morphology or texture in the EM images. Describing texture in digital images is an important part of this thesis. Viruses show up in an arbitrary orientation in the TEM images, making rotation invariant texture description important. Rotation invariance and noise robustness are evaluated for several texture descriptors in the thesis. Three new texture datasets are introduced to facilitate these evaluations. Invariant features and generalization performance in texture recognition are also addressed in a more general context. The work presented in this thesis has been part of the project Panvirshield, aiming for an automatic diagnostic system for viral pathogens using EM. The work is also part of the miniTEM project where a new desktop low-voltage electron microscope is developed with the aspiration to become an easy to use system reaching high levels of automation for clinical tissue sections, viruses and other nano-sized particles.

Image descriptors play an important role in image representation and analysis. Multi-resolution image descriptors can effectively characterize complex images and extract their hidden information. Wavelet descriptors have been widely used in multi-resolution image analysis. However, making the wavelet transform shift and rotation invariant produces redundancy and requires complex matching processes. As to other multi-resolution descriptors, they usually depend on other methods, such as filtering function, prior-domain knowledge, etc.; that not only increases the computation complexity, but also generates errors. We propose a novel multi-resolution scheme that is capable of transforming any kind of image descriptor into its multi-resolution structure with high computation accuracy and efficiency. Our multi-resolution scheme is based on sub-sampling each image into an odd-even image tree. Through applying image descriptors to the odd-even image tree, we get the relative multi-resolution image descriptors. Multi-resolution analysis is based on downsampling expansion with maximum energy extraction followed by upsampling reconstruction. Since the maximum energy usually retained in the lowest frequency coefficients; we do maximum energy extraction through keeping the lowest coefficients from each resolution level. Our multi-resolution scheme can analyze images recursively and effectively without introducing artifacts or changes to the original images, produce multi-resolution representations, obtain higher resolution images only using information from lower resolutions, compress data, filter noise, extract effective image features and be implemented in parallel processing.

Drug development is a long and expensive process. It is often not until potential drug candidates are administered to humans that accurate quantification of their pharmacokinetic characteristics is achieved. The goal of developing quantitative structure-pharmacokinetic relationships (QSPkRs) is to relate the molecular structure of a chemical entity with its pharmacokinetic characteristics. In this thesis artificial neural networks (ANNs) were used to construct in silico predictive QSPkRs for various pharmacokinetic parameters using different drug data sets. Drug pharmacokinetic data for all studies were taken from the literature. Information for model construction was extracted from drug molecular structure. Numerous theoretical descriptors were generated from drug structure ranging from simple constitutional and functional group counts to complex 3D quantum chemical numbers. Subsets of descriptors were selected which best modeled the target pharmacokinetic parameter(s). Using manual selective pruning, QSPkRs for physiological clearances, volumes of distribution, and fraction bound to plasma proteins were developed for a series of beta-adrenoceptor antagonists. All optimum ANN models had training and cross-validation correlations close to unity, while testing was performed with an independent set of compounds. In most cases the ANN models developed performed better than other published ANN models for the same drug data set. The ability of ANNs to develop QSPkRs with multiple target outputs was investigated for a series of cephalosporins. Multilayer perceptron ANN models were constructed for prediction of half life, volume of distribution, clearances (whole body and renal), fraction excreted in the urine, and fraction bound to plasma proteins. The optimum model was well able to differentiate compounds in a qualitative manner while quantitative predictions were mostly in agreement with observed literature values. The ability to make simultaneous predictions of important pharmacokinetic properties of a compound made this a valuable model. A radial-basis function ANN was employed to construct a quantitative structure-bioavailability relationship for a large, structurally diverse series of compounds. The optimum model contained descriptors encoding constitutional through to conformation dependent solubility characteristics. Prediction of bioavailability for the independent testing set were generally close to observed values. Furthermore, the optimum model provided a good qualitative tool for differentiating between drugs with either low or high experimental bioavailability. QSPkR models constructed with ANNs were compared with multilinear regression models. ANN models were shown to be more effective at selecting a suitable subset of descriptors to model a given pharmacokinetic parameter. They also gave more accurate predictions than multilinear regression equations. This thesis presents work which supports the use of ANNs in pharmacokinetic modeling. Successful QSPkRs were constructed using different combinations of theoretically-derived descriptors and model optimisation techniques. The results demonstrate that ANNs provide a valuable modeling tool that may be useful in drug discovery and development.

A idéia de capacitar uma máquina a reconhecer o ambiente em que atua tem motivado pesquisadores a investir esforços no estudo do mais complexo dos sentidos humanos, a visão. A visão é, antes de tudo, uma tarefa de representação e processamento de informações, sendo portanto adequada ao tratamento computacional. Visto que ainda não se possuem métodos que tenham resultados equivalentes ao que seria obtido com um usuário humano, tem-se estudado intensamente a utilização de feições para um melhor aproveitamento de seu potencial. Dentre estas feições, a forma de um objeto proporciona um poderoso indício de sua identidade e funcionalidade, podendo ser utilizada para seu reconhecimento. Isso distingue a forma de outras feições visuais elementares, como a cor, o movimento ou a textura, que, apesar de igualmente importantes, normalmente não revelam a identidade de um objeto. Assim sendo, a possibilidade de avaliar a robustez e a estabilidade de técnicas alternativas para a representação de forma é vital para prever o desempenho de cada técnica na presença de alguma incerteza ou discrepância. Neste trabalho, alguns descritores de forma descritos na literatura foram implementados e utilizados em estudos de caso para avaliar sua eficácia. Estes estudos de caso foram realizados utilizando-se caracteres, todavia, com finalidades bastante distintas. O primeiro estudo de caso é voltado para aplicações como a robótica móvel, com reconhecimento de comandos localizados no ambiente por parte do robô. Já o estudo de caso principal está direcionado para aplicações de reconhecimento de placas de automóveis, que poderia tanto ser utilizado para monitoramento e controle do fluxo de trânsito, quanto para controle de infrações. Muitas aplicações, incluindo aquelas que envolvem a recuperação e indexação de objetos visuais, são apropriadas para a utilização de feições de forma. Outra característica importante do presente trabalho é a de realçar que a seleção de um bom descritor reduz o esforço necessário na etapa de classificação, o qual é computacionalmente elevado. / The idea of enabling a machine to recognize the environment with which it interacts has motivated researchers to dedicate efforts in studying the most complex of the human senses: vision. Vision is essentially a task of information representation and processing, what makes it suitable for computational treatment. Given that currently there are no methods that perform equivalently to humans, the use of features has been intensively studied in order to improve the performance of the existing methods. Among these features, the shape of an object provides a powerful sign of its identity and functionality, what enables the exploitation of this feature with the purpose of recognition. This evidence distinguishes shape from other visual features, such as color, motion or texture, which, although equally important, normally do not reveal the identity of an object. As a result, the possibility of evaluating the robustness and stability of alternate techniques for shape representation is essential in order to measure the performance of each technique in the presence of uncertainty. In this work, some shape descriptors available in the literature were implemented and used in case studies aiming at evaluating their effectiveness. These case studies were carried out using characters, although, with very different purposes. The first case study is geared towards applications such as mobile robotics, where the robot recognizes commands available in the environment. The main case study is focused on applications of license plate recognition, which could be used both in situations of surveillance and traffic control and in situations of infraction. Many applications, including those that involve the search and indexing of visual objects, are suited for the use of shape features. Another important characteristic of this work is that it emphasizes that the selection of a good shape description reduces the effort during the classification step, which is computationally elevated.

Os macroinvertebrados bentônicos constituem uma importante comunidade em riachos, pois participam do fluxo de energia, sendo um importante recurso alimentar para níveis tróficos adjacentes e superiores. Formam uma fauna bastante diversificada e a estrutura desta comunidade pode ser influenciada por diversos fatores ambientais, os quais variam no tempo, no espaço e na escala analisada. O presente estudo objetivou: a) analisar a estrutura da comunidade de macroinvertebrados bentônicos em uma bacia hidrográfica gaúcha, a saber bacia do rio Forqueta (RS, Brasil) em nível taxonômico e de grupos funcionais; b) investigar a variação sazonal (verão e inverno) e a variação espacial da estrutura da comunidade em função de diferenças espaciais nas características morfo-fisiográficas dos trechos amostrados; c) identificar a variabilidade da comunidade de macroinvertebrados bentônicos em três escalas espaciais (rio, segmento de rio e mesohábitat), enfatizando quais escalas espaciais melhor explicam a estrutura da comunidade nesta bacia; d) investigar quais os descritores ambientais mensurados influenciam na estrutura da comunidade e e) qual a porcentagem da variabilidade na riqueza de organismos pode ser explicada pelos descritores ambientais locais mensurados. Realizou-se amostragens de macroinvertebrados bentônicos e descritores ambientais nos períodos de inverno de 2007 e verão de 2008 em oito sítios de amostragem na bacia do rio Forqueta (RS, Brasil). A estrutura da comunidade de macroinvertebrados bentônicos foi caracterizada por uma baixa diversidade, presença de muitas famílias raras e poucas abundantes. Houve predomínio do grupo funcional de coletores nos ambientes analisados. Observou-se maior diversidade nos locais amostrados localizados na unidade geomorfológica Serra Geral, denotando a importância da geomorfologia na estruturação das comunidades aquáticas. Para a análise da estrutura da comunidade em escalas espaciais, as amostragens seguiram um delineamento amostral hierárquico. Foram analisados oito segmentos de rio, considerando os mesohábitats: corredeira e remanso, no verão de 2008. A análise nested Anova mostrou que a comunidade de macroinvertebrados bentônicos varia nas escalas analisadas, sendo que a comunidade está estruturada principalmente de acordo com a escala de mesohábitat. 46,5% da variação na riqueza foi explicada pelas diferenças entre os mesohábitats. Nesta escala houve maior variação na estrutura da comunidade, segundo a Permanova. A Análise de Redundância parcial (pRDA) evidenciou pH, largura do leito do rio, velocidade da corrente e alcalinidade como os descritores que mais contribuiram para explicar a estrutura espacial da comunidade. A partilha da variância mostrou que 12,5% da variabilidade da comunidade foi explicada puramente pelos descritores ambientais. Os resultados mostraram a correspondência entre a distribuição das comunidades de macroinvertebrados bentônicos e os descritores ambientais, demonstrando a importância das variações em mesoescala para o estudo da distribuição destes organismos nesta bacia. / The benthic macroinvertebrate constitute an important community in streams, as part of the flow of energy, being an important food resource for adjacent and higher trophic levels. Form a very diverse fauna and the structure of this community can be influenced by several environmental factors, which vary in time, in space and on the scale considered. This study aimed to: a) analyze the community structure of benthic macroinvertebrate in Forqueta river basin (RS, Brazil) in level taxonomic and functional groups, during winter 2007 and summer 2008, b) investigate the seasonal variation (summer and winter) and the spatial variation of community structure due to spatial differences in morpho-physiographic, d) identify the variability of the macroinvertebrate community at three spatial scales (river, river segment and mesohabitat), emphasizing spatial scales which best explain the community structure in this basin, e) investigate the descriptors which measured environmental influences on community structure and f) what percentage of the variability in the richness of organisms can be explained by local environmental descriptors. The structure of the benthic macroinvertebrate community was characterized by a low diversity, the presence of many rare and few abundant families. There was a higher of the functional group of collectors in the environments analyzed. There was greater diversity in Serra Geral geomorphological unit, emphasizing the importance of geomorphology in structuring aquatic communities. For the analysis of community structure in spatial scales, the sampling followed a hierarchical sampling design. We analyzed the eight segments of the river, considering the mesohabitats: riffles and pools, in the summer of 2008. The nested ANOVA showed that the benthic macroinvertebrate community changes in scales, and that the community is organized mainly according to the scale of mesohabitat. 46.5% of the variation in richness was explained by differences between the mesohabitats. This scale was greater variation in community structure, according to Permanova. Partial Redundancy Analysis (pRDA) showed pH, width of the river bed, flow and alkalinity as the descriptors that contributed most to explain the spatial structure of the community. The partition of variance showed that 12.5% of the variability of community was explained by purely environmental descriptors. The results showed the correlation between the distribution of benthic macroinvertebrate communities and environmental descriptors, demonstrating the importance of variations in mesoscale to study the distribution of these organisms in this basin.

Made available in DSpace on 2015-03-26T13:23:54Z (GMT). No. of bitstreams: 1 texto completo.pdf: 1451062 bytes, checksum: 39d55bd85f54e95a546d07249a16f7ab (MD5) Previous issue date: 2013-07-19 / Conselho Nacional de Desenvolvimento Científico e Tecnológico / The importance of coffee agriculture in Brazil is notorious, especially for the State of Minas Gerais, which is the Brazilian state that accounts for most of the national production. In the Southern region and in Zona da Mata they are concentrated most of the crops of the State; also, small holdings are predominant in the region and cultivation is done in mountain region, what makes it difficult mapping the crops by automated methods. The application of Artificial Neural Networks (ANNs), having as input variables Haralick s descriptors, has shown a promising approach for the discrimination of higher complexity classes. In this context, it aimed to develop a system to identify areas where coffee is cultivated using ANNs, having as input variables Haralick s descriptors. The studied area is located at the city of Araponga, where 59 fields were selected with coffee plantations for data collection. The software used for processing and sorting the images was MATLAB; and for evaluating the sorting performance, Arcgis was used. The methodology for the development of ANN consisted in two steps: in the first step, the ANN was trained with representative samples of each class of interest (coffee, forest, water, bare soil, and grassland, and urban area), thus verifying the potential to discriminate output classes; in the second step the objective was to classify the coffee plantations accordingly with the age. Kappa index was used for evaluating the performance of ANN; the usage of this coefficient is satisfactory for assessing the accuracy of a thematic class. Kappa index for discriminating the coffee region of the other class of interest was 65,18%, what can be considered a good index. To classify the coffee plantations accordingly with the age, Kappa index was variable (0.675 to 0.4783), being very good for the Itatiaia farm and reasonable to the Pedra Redonda farm. / A importância da cafeicultura para o Brasil é notória, em especial para o estado de Minas Gerais que é o estado brasileiro responsável pela maior parte da produção nacional. Nas regiões sul e da zona da mata onde estão concentradas a maior parte da lavoura no estado de Minas Gerias, há a predominância de pequenas propriedades e o cultivo é feito em região de montanha o que acaba dificultando o mapeamento por métodos automatizados. A aplicação de Redes Neurais Artificiais (RNAs) tendo como variáveis de entrada os descritores de Haralick tem se mostrado uma abordagem promissora na discriminação de classes de maior complexidade. Neste contexto objetivou-se desenvolver um sistema para identificar áreas cultivadas com café utilizando RNAs tendo como variáveis de entrada os descritores de Haralick. A área de estudo está localizada no município de Araponga, onde foram selecionados 59 talhões com plantios de café, sendo levantados dados relativos à idade e data de recepa. O software utilizado para o processamento e classificação da imagem foi o MATLAB, e para avaliar o desempenho da classificação foi o Arcgis. A metodologia para o desenvolvimento da RNA consistiu em duas etapas: na primeira a RNA foi treinada com amostras representativas de cada classe de interesse (café, mata, água, solo exposto e pastagem e área urbana), verificando assim o potencial em discriminar entre as classes de saída; na segunda etapa o objetivo foi classificar as plantações de café de acordo com a idade e com a data de recepa. Utilizou-se o índice Kappa para avaliar o desempenho da RNA, uma vez que o uso desse coeficiente é satisfatório na avaliação da precisão de uma classe temática. O índice Kappa para discriminar a região cafeeira das outras classes temáticas foi de 65,18%, o que pode ser considerado um índice bom. Para classificar os plantios de café em função da idade e data de recepa o índice Kappa foi variável (0,675 a 0,4783), sendo considerado muito bom para a fazenda Itatiaia e razoável para a fazenda Pedra Redonda.