Esta tese investiga uma abordagem não supervisionada para o problema da detecção de mudanças em imagens multiespectrais e multitemporais de sensoriamento remoto empregando Support Vector Machines (SVM) com o uso dos kernels polinomial e RBF e de uma nova métrica de pertinência de pixels. A proposta metodológica está baseada na diferença das imagens-fração produzidas para cada data. Em imagens de cenas naturais essa diferença nas frações de solo e vegetação tendem a apresentar uma distribuição simétrica próxima à origem. Essa caracteristica pode ser usada para modelar as distribuições normais multivariadas das classes mudança e não-mudança. O algoritmo Expectation-Maximization (EM) é implementado com a finalidade de estimar os parâmetros (vetor de médias, matriz de covariância e probabilidade a priori) associados a essas duas distribuições. A seguir, amostras aleatórias e normalmente distribuidas são extraídas dessas distribuições e rotuladas segundo sua pertinência em uma das classes. Essas amostras são então usadas no treinamento do classificador SVM. A partir desta classificação é estimada uma nova métrica de pertinência de pixels. A metodologia proposta realiza testes com o uso de conjuntos de dados multitemporais de imagens multiespectrais Landsat-TM que cobrem a mesma cena em duas datas diferentes. A métrica de pertinência proposta é validada através de amostras de teste controladas obtidas a partir da técnica Change Vetor Analysis, além disso, os resultados de pertinência obtidos para a imagem original com essa nova métrica são comparados aos resultados de pertinência obtidos para a mesma imagem pela métrica proposta em (Zanotta, 2010). Baseado nos resultados apresentados neste trabalho que mostram que a métrica para determinação de pertinência é válida e também apresenta resultados compatíveis com outra técnica de pertinência publicada na literatura e considerando que para obter esses resultados utilizou-se poucas amostras de treinamento, espera-se que essa métrica deva apresentar melhores resultados que os que seriam apresentados com classificadores paramétricos quando aplicado a imagens multitemporais e hiperespectrais. / This thesis investigates a unsupervised approach to the problem of change detection in multispectral and multitemporal remote sensing images using Support Vector Machines (SVM) with the use of polynomial and RBF kernels and a new metric of pertinence of pixels. The methodology is based on the difference-fraction images produced for each date. In images of natural scenes. This difference in the fractions of bare soil and vegetation tend to have a symmetrical distribution close to the origin. This feature can be used to model the multivariate normal distributions of the classes change and no-change. The Expectation- Maximization algorithm (EM) is implemented in order to estimate the parameters (mean vector, covariance matrix and a priori probability) associated with these two distributions. Then random and normally distributed samples are extracted from these distributions and labeled according to their pertinence to the classes. These samples are then used in the training of SVM classifier. From this classification is estimated a new metric of pertinence of pixel. The proposed methodology performs tests using multitemporal data sets of multispectral Landsat-TM images that cover the same scene at two different dates. The proposed metric of pertinence is validated via controlled test samples obtained from Change Vector Analysis technique. In addition, the results obtained at the original image with the new metric are compared to the results obtained at the same image applying the pertinence metric proposed in (Zanotta, 2010). Based on the results presented here showing that the metric of pertinence is valid, and also provides results consistent with other published in the relevant technical literature, and considering that to obtain these results was used a few training samples, it is expected that the metric proposed should present better results than those that would be presented with parametric classifiers when applied to multitemporal and hyperspectral images.

abstract: With the rapid development of mobile sensing technologies like GPS, RFID, sensors in smartphones, etc., capturing position data in the form of trajectories has become easy. Moving object trajectory analysis is a growing area of interest these days owing to its applications in various domains such as marketing, security, traffic monitoring and management, etc. To better understand movement behaviors from the raw mobility data, this doctoral work provides analytic models for analyzing trajectory data. As a first contribution, a model is developed to detect changes in trajectories with time. If the taxis moving in a city are viewed as sensors that provide real time information of the traffic in the city, a change in these trajectories with time can reveal that the road network has changed. To detect changes, trajectories are modeled with a Hidden Markov Model (HMM). A modified training algorithm, for parameter estimation in HMM, called m-BaumWelch, is used to develop likelihood estimates under assumed changes and used to detect changes in trajectory data with time. Data from vehicles are used to test the method for change detection. Secondly, sequential pattern mining is used to develop a model to detect changes in frequent patterns occurring in trajectory data. The aim is to answer two questions: Are the frequent patterns still frequent in the new data? If they are frequent, has the time interval distribution in the pattern changed? Two different approaches are considered for change detection, frequency-based approach and distribution-based approach. The methods are illustrated with vehicle trajectory data. Finally, a model is developed for clustering and outlier detection in semantic trajectories. A challenge with clustering semantic trajectories is that both numeric and categorical attributes are present. Another problem to be addressed while clustering is that trajectories can be of different lengths and also have missing values. A tree-based ensemble is used to address these problems. The approach is extended to outlier detection in semantic trajectories. / Dissertation/Thesis / Ph.D. Industrial Engineering 2012

A evolução do sistema Terrestre sempre foi constante, apresentando períodos com instabilidade de grande, média ou pequena magnitude. A zona costeira é um ambiente que está na interface entre o continente e o oceano, na qual as dunas dessa região apresentam intenso dinamismo, o que torna imprescindível o seu constante monitoramento. A técnica de rotação radiométrica controlada por eixo de não mudança (RCEN) possibilita a análise multitemporal de uma paisagem através da avaliação dos padrões de resposta espectral de toda a imagem, sem a necessidade de correção dos efeitos atmosféricos. Os objetivos desta dissertação são analisar a dinâmica espaço-temporal do Parque Nacional dos Lençóis Maranhenses (PNLM) no período de 1984 a 2014, a partir da análise de imagens de satélite e avaliar a técnica RCEN para a identificação e monitoramento das alterações do parque. A área de estudo está localizada na região nordeste do Brasil, litoral oriental do estado do Maranhão, e tornou-se uma área de proteção ambiental por meio do Decreto Federal nº 86.060, de 02 de junho de 1981. Para as análises foram utilizadas imagens dos satélites Landsat 5 - Sensor TM e Landsat 8 - Sensor OLI para os períodos estudados. A borda limite do parque na parte interior do continente foi vetorizada a partir das imagens dos anos de 1984 e de 2014, para se avaliar o avanço e a retração das dunas. Foram elaborados mapas com o deslocamento dunário em um período de 30 anos sendo identificado um maior avanço em relação à retração em toda a área. A variação do comportamento espectral de pontos de monitoramento também foi avaliada, para identificar o período da ocorrência da mudança. A RCEN foi testada na borda interna limítrofe do parque, na qual foi determinado o ângulo de rotação dos eixos radiométricos, que é o principal parâmetro para a obtenção da imagem de detecção de mudanças. Foi possível a identificação das áreas de não mudança (90,67%) e a variação das dunas (5, 20%), vegetação (3,86%) e água (0,27%) em relação à área total. Identificou-se que a expansão das dunas, do litoral em direção ao interior do continente, ocorre no sentido nordeste - sudoeste, seguindo a circulação dos ventos alísios e que a vegetação ocupou os espaçamentos deixados pelas dunas ou pela água das lagoas. A RCEN foi eficaz ao traduzir as alterações na área analisada. Considerando o tipo de mudanças identificadas, entende-se que monitorar o deslocamento dunário no PNLM é relevante para o planejamento do parque, pois este representa um importante ecossistema da região, bem como é um considerável polo turístico para o Brasil. / The Earth system evolution has always been constant, presenting large, medium or small magnitude instability. The coastal zone is an environment that occurs at the interface between the continent and the ocean, where the dunes have an intense dynamic, requiring their constant monitoring. The radiometric rotation controlled by no change axis (RCEN) technique allow multi-temporal analysis of a landscape by evaluating the spectral response patterns of the entire image, without the need for correction of atmospheric effects. The aim of this study was to analyze the spatio-temporal dynamics of the Parque Nacional dos Lençóis Maranhenses (PNLM) during 1984 to 2014, from satellite imagery analysis and evaluate the RCEN technique for the identification and monitoring of these changes. The study area is located in northeastern Brazil, eastern coast of the state of Maranhão, and has become an environmental protection area through the Federal Decree No. 86060 of June 2, 1981. For these analyses were used satellite images from the Landsat 5-TM and Landsat 8-OLI sensors. The park edge boundary in the inner part of the continent was vectored from the images of the years 1984 and 2014, to assess the advancement and retraction of the dunes. Were made maps with the dunes changes over a period of 30 years and were identified a bigger advancement in relation of the shrinkage among throughout the area. The spectral pattern variation in some monitoring points was also evaluated to identify the timing of the change occurrence. The RCEN was tested in the adjacent inner border of the park in where the angle of rotation of radiometric axis was determined, considering that this angle is the main parameter for obtaining the change detection image. Were identified areas without changes (90.67%) and the variation of the dunes (5, 20%), vegetation (3.86%) and water (0.27%) in relation to the total area. It was identified that the expansion of the dunes, occurs from the coast line towards the interior of the continent, with the northeast – southwest direction, following the circulation of trade winds, while the vegetation occupied the gaps left by the dunes or water ponds. The RCEN was effective in identify the changes in the study area. Considering those changes detected, monitoring the dunes movement in PNLM is relevant to the park planning, once it represents an important ecosystem of the region and is a great tourist hub for Brazil.

Esta tese investiga uma abordagem supervisionada para o problema da detecção de mudanças em imagens multitemporais de sensoriamento remoto empregando Support Vector Machines (SVM) com o uso dos kernels polinomial e gaussiano (RBF). A proposta metodológica está baseada na diferença das imagens-fração produzidas para cada data. Em imagens de cenas naturais a diferença nas frações de solo e vegetação tendem a apresentar uma distribuição simétrica em torno da origem. Esse fato pode ser usado para modelar duas distribuições normais multivariadas: mudança e não-mudança. O algoritmo Expectation-Maximization (EM) é implementado para estimar os parâmetros (vetor de médias, matriz de covariância e probabilidade a priori) associados a essas duas distribuições. Amostras aleatórias são extraídas dessas distribuições e usadas para treinar o classificador SVM nesta abordagem supervisionada. A metodologia proposta realiza testes com o uso de conjuntos de dados multitemporais de imagens multiespectrais TM-Landsat, que cobrem a mesma cena em duas datas diferentes. Os resultados são comparados com outros procedimentos, incluindo trabalhos anteriores, um conjunto de dados sintéticos e o classificador SVM One-Class. / In this thesis, we investigate a supervised approach to change detection in remote sensing multi-temporal image data by applying Support Vector Machines (SVM) technique using polynomial kernel and Gaussian kernel (RBF). The methodology is based on the difference-fraction images produced for two dates. In natural scenes, the difference in the fractions such as vegetation and bare soil occurring in two different dates tend to present a distribution symmetric around the origin of the coordinate system. This fact can be used to model two normal multivariate distributions: class change and no-change. The Expectation-Maximization algorithm (EM) is implemented to estimate the parameters (mean vector, covariance matrix and a priori probability) associated with these two distributions. Random samples are drawn from these distributions and used to train the SVM classifier in this supervised approach.The proposed methodology performs tests using multi-temporal TMLandsat multispectral image data covering the same scene in two different dates. The results are compared to other procedures including previous work, a synthetic data set and SVM One-Class.

Esta tese investiga uma abordagem não supervisionada para o problema da detecção de mudanças em imagens multiespectrais e multitemporais de sensoriamento remoto empregando Support Vector Machines (SVM) com o uso dos kernels polinomial e RBF e de uma nova métrica de pertinência de pixels. A proposta metodológica está baseada na diferença das imagens-fração produzidas para cada data. Em imagens de cenas naturais essa diferença nas frações de solo e vegetação tendem a apresentar uma distribuição simétrica próxima à origem. Essa caracteristica pode ser usada para modelar as distribuições normais multivariadas das classes mudança e não-mudança. O algoritmo Expectation-Maximization (EM) é implementado com a finalidade de estimar os parâmetros (vetor de médias, matriz de covariância e probabilidade a priori) associados a essas duas distribuições. A seguir, amostras aleatórias e normalmente distribuidas são extraídas dessas distribuições e rotuladas segundo sua pertinência em uma das classes. Essas amostras são então usadas no treinamento do classificador SVM. A partir desta classificação é estimada uma nova métrica de pertinência de pixels. A metodologia proposta realiza testes com o uso de conjuntos de dados multitemporais de imagens multiespectrais Landsat-TM que cobrem a mesma cena em duas datas diferentes. A métrica de pertinência proposta é validada através de amostras de teste controladas obtidas a partir da técnica Change Vetor Analysis, além disso, os resultados de pertinência obtidos para a imagem original com essa nova métrica são comparados aos resultados de pertinência obtidos para a mesma imagem pela métrica proposta em (Zanotta, 2010). Baseado nos resultados apresentados neste trabalho que mostram que a métrica para determinação de pertinência é válida e também apresenta resultados compatíveis com outra técnica de pertinência publicada na literatura e considerando que para obter esses resultados utilizou-se poucas amostras de treinamento, espera-se que essa métrica deva apresentar melhores resultados que os que seriam apresentados com classificadores paramétricos quando aplicado a imagens multitemporais e hiperespectrais. / This thesis investigates a unsupervised approach to the problem of change detection in multispectral and multitemporal remote sensing images using Support Vector Machines (SVM) with the use of polynomial and RBF kernels and a new metric of pertinence of pixels. The methodology is based on the difference-fraction images produced for each date. In images of natural scenes. This difference in the fractions of bare soil and vegetation tend to have a symmetrical distribution close to the origin. This feature can be used to model the multivariate normal distributions of the classes change and no-change. The Expectation- Maximization algorithm (EM) is implemented in order to estimate the parameters (mean vector, covariance matrix and a priori probability) associated with these two distributions. Then random and normally distributed samples are extracted from these distributions and labeled according to their pertinence to the classes. These samples are then used in the training of SVM classifier. From this classification is estimated a new metric of pertinence of pixel. The proposed methodology performs tests using multitemporal data sets of multispectral Landsat-TM images that cover the same scene at two different dates. The proposed metric of pertinence is validated via controlled test samples obtained from Change Vector Analysis technique. In addition, the results obtained at the original image with the new metric are compared to the results obtained at the same image applying the pertinence metric proposed in (Zanotta, 2010). Based on the results presented here showing that the metric of pertinence is valid, and also provides results consistent with other published in the relevant technical literature, and considering that to obtain these results was used a few training samples, it is expected that the metric proposed should present better results than those that would be presented with parametric classifiers when applied to multitemporal and hyperspectral images.

Data warehousing is a generally accepted method of providing corporate decision support. Today, the majority of information in these warehouses originates from sources within a company, although changes often occur from the outside. Companies need to look outside their enterprises for valuable information, increasing their knowledge of customers, suppliers, competitors etc. The largest and most frequently accessed information source today is the Web, which holds more and more useful business information. Today, the Web primarily relies on HTML, making mechanical extraction of information a difficult task. In the near future, XML is expected to replace HTML as the language of the Web, bringing more structure and content focus. One problem when considering XML-sources in a data warehouse context is their lack of update awareness capabilities, which restricts eligible data warehouse maintenance policies. In this work, we wrap XML-sources in order to provide update awareness capabilities. We have implemented a wrapper prototype that provides update awareness capabilities for autonomous XML-sources, especially change awareness, change activeness, and delta awareness. The prototype wrapper complies with recommendations and working drafts proposed by W3C, thereby being compliant with most off-the-shelf XML tools. In particular, change information produced by the wrapper is based on methods defined by the DOM, implying that any DOM-compliant software, including most off-the-shelf XML processing tools, can be used to incorporate identified changes in a source into an older version of it. For the delta awareness capability we have investigated the possibility of using change detection algorithms proposed for semi-structured data. We have identified similarities and differences between XML and semi-structured data, which affect delta awareness for XML-sources. As a result of this effort, we propose an algorithm for change detection in XML-sources. We also propose matching criteria for XML-documents, to which the documents have to conform to be subject to change awareness extension.

The objectives of my PhD were to identify the spatial and the temporal dynamics of the brain areas involved in tactile change detection as well as the neural mechanisms responsible for the processing of tactile change detection. To that aim, three specific MEG studies were performed; each of them is addressing specific research aims.The first study investigated the spatiotemporal dynamics of the multilevel cortical processing of tactile change detection in human healthy subjects. This study disclosed a hierarchical organization from unimodal early tactile change detection at secondary somatosensory cortex to multi modal complex processing at bilateral temporo-parietal junctions, posterior parietal cortex and supplementary motor areas. The second study aimed at discriminating between debated neural mechanisms responsible for the genesis of the somatosensory mismatch negativity (sMMN). To do so, we manipulated the predictability of the deviant stimuli and the response to omissions in different kind of oddballs, the response to deviant stimuli paired with standards and occurring alone. We found out that mechanisms for early tactile change detection reflected by the sMMN were better explained by the predictive coding theory compared to the adaptation and adjustment theories. Finally we sought to characterize the alterations in early cortical tactile change detection in Friedreich Ataxia (FRDA); a neurological disorder characterized by somatosensory and cerebellar pathways degeneration. The aim of this work was to study the role of the cerebellum in the genesis of sMMN and its potential selectivity for somatosensory change detection compared to auditory. This study demonstrated that, in FRDA, both tactile and auditory pathways are affected at the level of primary sensory neurons and dorsal root/spiral ganglia in a genetically determined. By contrasts, early cortical sensory change detection in FRDA was impaired only in the tactile modality in line with the sMMN impairment described in patients with acquired cerebellar lesions or during cerebellar inhibition by trans cranial magnetic stimulation. These data brought novel empirical evidence supporting the contribution of spinocerebellar tracts in sMMN genesis at cSII cortex.In conclusion, this PhD contributed to identify the network responsible for tactile change detection that involves cuneocerebellar spinocerebellar tract and cSII cortex as somatosensory specific areas and TPJ, SMA & PPC as multimodal brain areas. We further provided evidence that early change detection mechanisms at SII cortex fall under the predictive coding framework and that change detection is hierarchically organized with inputs from low level areas for genesis of an adequate generative model of our environment and conscious representation of our body. / Doctorat en Sciences médicales (Médecine) / info:eu-repo/semantics/nonPublished

Rapid urban growth and unprecedented rural to urban transition, along with a huge population growth are new phenomena for both high and low income countries, which started in the mid-20th century. However, urban growth rates and patterns are different in developed countries and developing ones. In less developed countries, urbanization and rural to urban transition usually takes place in an unmanaged way and they are associated with a series of socioeconomical and environmental issues and problems. Identification of the city growth trends in past decades can help urban planners and managers to minimize these negative impacts. In this research, urban growth in the city of Isfahan, Iran, is the subject of study. Isfahan the third largest city in Iran has experienced a huge urban growth and population boom during the last three decades. This transition led to the destruction of natural and agricultural lands and environmental pollutions. Historical and recent remotely sensed data, along with different remote sensing techniques and methods have been used by researchers for urban land use and land cover change detection. In this study three Landsat TM and ETM+ images of the study site, acquired in 1985, 2000 and 2009 are used. Before starting processing, radiometric normalization is done to minimize the atmospheric effects. Then, processing methods including principal component analysis (PCA), vegetation indices and supervised classification are implemented on the images. Accuracy assessment of the PCA method showed that the first PC was responsible for more than 81% of the total variance, and therefore used for analysis of PCA differencing. ΔPC1t1-t2 shows the amount of changes in land use and land cover during the period of study. In this study ten vegetation indices were selected to be applied to the 1985 image. Accuracy assessments showed that Transformed Differencing Vegetation Index (TDVI) is the most sensitive and accurate index for mapping vegetation in arid and semi-arid urban areas. Hence, TDVI was applied to the 2000 and 2009 images. ΔTDVIt1-t2 showed the changes in land use and land cover especially the land use transformation from vegetation cover into the urban class. Supervised classification is the last method applied to the images. Training sites were assigned for the selected classes and accuracy was monitored during the process of training site selection. The results of classification show the expansion of urban class and diminishment in natural and agricultural lands.

The present thesis employed event-related potentials, the minute responses of the brain, to examine the differences in processing of increases and decreases in auditory intensity. The manner in which intensity was manipulated (i.e., whether it represented physical or psychological change) varied across the studies of the thesis. Study 1 investigated the processing of physical intensity change during wakefulness and natural sleep. An oddball paradigm (80 dB standard, 90 dB increment, 60 dB decrement) was presented to subjects during the waking state and during sleep. The increment elicited a larger deviant-related negativity and P3a than the decrement in the waking state. During sleep, only the increment deviant continued to elicit ERPs related to the detection of change. The waking and sleeping findings support the notion that increases in intensity are more salient to an observer. Studies 2 and 3 of this thesis determined the degree to which this differential salience could be attributed to the fact that intensity increments result in increased activation of the change and transient detection systems while intensity decrements result in greater activation of only the change detection system. In order to address this question, an alternating intensity pattern was employed (HLHLHLHL) with deviants created by the repetition of a tone in the sequence (HLHLHHHL) that violated the expectancy for a higher (psychological decrements) or lower intensity tone (psychological increments). Because deviant stimuli were physically identical to preceding standards, this manipulation should not have led to increased output of the transient detection system (N1 enhancement), permitting isolation of the output of the change detection system (Mismatch Negativity, MMN). The findings of these studies indicated that psychological increments resulted in shorter latency and larger amplitude MMNs than psychological decrements and that these differences could not be explained by the physical differences between deviant stimuli or temporal integration. This thesis provides convincing evidence that stimuli representing increments in intensity result in faster and more robust change detection. Further, the increased salience of increment stimuli cannot be solely explained by the contribution of transient detector activation, as it persists even when deviance-related processing is isolated to the change detection system.

Change point detection (CPD) is the method of detecting sudden changes in timeseries, and its importance is great concerning network traffic. With increased knowledge of occurring changes in data logs due to updates in networking equipment,a deeper understanding is allowed for interactions between the updates and theoperational resource usage. In a data log that reflects the amount of network traffic, there are large variations in the time series because of reasons such as connectioncount or external changes to the system. To circumvent these unwanted variationchanges and assort the deliberate variation changes is a challenge. In this thesis, we utilize data logs retrieved from a network equipment vendor to detect changes, then compare the detected changes to when firmware/signature updates were applied, configuration changes were made, etc. with the goal to achieve a deeper understanding of any interaction between firmware/signature/configuration changes and operational resource usage. Challenges in the data quality and data processing are addressed through data manipulation to counteract anomalies and unwanted variation, as well as experimentation with parameters to achieve the most ideal settings. Results are produced through experiments to test the accuracy of the various change pointdetection methods, and for investigation of various parameter settings. Through trial and error, a satisfactory configuration is achieved and used in large scale log detection experiments. The results from the experiments conclude that additional information about how changes in variation arises is required to derive the desired understanding.