Thèses sur le sujet « Space-time features »

The representation (or feature) learning has been an emerging concept in the last years, since it collects a set of techniques that are present in any theoretical or practical methodology referring to artificial intelligence. In computer vision, a very common representation has adopted the form of the well-known Bag of Visual Words. This representation appears implicitly in most approaches where images are described, and is also present in a huge number of areas and domains: image content retrieval, pedestrian detection, human-computer interaction, surveillance, e-health, and social computing, amongst others. The early stages of this dissertation provide an approach for learning visual representations inside evolutionary algorithms, which consists of evolving weighting schemes to improve the BoVW representations for the task of recognizing categories of videos and images. Thus, we demonstrate the applicability of the most common weighting schemes, which are often used in text mining but are less frequently found in computer vision tasks. Beyond learning these visual representations, we provide an approach based on fusion strategies for learning spatiotemporal representations, from multimodal data obtained by depth sensors. Besides, we specially aim at the evolutionary and dynamic modelling, where the temporal factor is present in the nature of the data, such as video sequences of gestures and actions. Indeed, we explore the effects of probabilistic modelling for those approaches based on dynamic programming, so as to handle the temporal deformation and variance amongst video sequences of different categories. Finally, we integrate dynamic programming and generative models into an evolutionary computation framework, with the aim of learning Bags of SubGestures (BoSG) representations and hence to improve the generalization capability of standard gesture recognition approaches. The results obtained in the experimentation demonstrate, first, that evolutionary algorithms are useful for improving the representation of BoVW approaches in several datasets for recognizing categories in still images and video sequences. On the other hand, our experimentation reveals that both, the use of dynamic programming and generative models to align video sequences, and the representations obtained from applying fusion strategies in multimodal data, entail an enhancement on the performance when recognizing some gesture categories. Furthermore, the combination of evolutionary algorithms with models based on dynamic programming and generative approaches results, when aiming at the classification of video categories on large video datasets, in a considerable improvement over standard gesture and action recognition approaches. Finally, we demonstrate the applications of these representations in several domains for human analysis: classification of images where humans may be present, action and gesture recognition for general applications, and in particular for conversational settings within the field of restorative justice.
L’aprenentatge de la representació (o de característiques) ha estat un concepte emergent en els darrers anys, ja que recopila un conjunt de tècniques que són presents en qualsevol metodologia teòrica o pràctica referent a la intel·ligència artifcial. En la visió per computador, una representació molt comuna ha adoptat la forma de la ben coneguda Bossa de Paraules Visuals (BdPV). Aquesta representació apareix implícitament en la majoria d’aproximacions per descriure imatges, i és també present en un enorme nombre d’àrees i dominis: recuperació de contingut en imatges, detecció de vianants, interacció humà-ordinador, vigilància, e-salut, i la computació social, entre d’altres. Les fases inicials d’aquesta dissertació proporcionen una aproximació per aprendre representacions visuals dins d’algorismes evolutius, que consisteix en evolucionar esquemes de pesat per millorar les representacions BdPV en la tasca de reconèixer les categories de vídeos i imatges. Per tant, demostrem l’aplicabilitat dels esquemes de pesat més comuns, que s’usen sovint en la mineria de textos però es troben amb menys freqüència en tasques de visió per computador. Més enllà d’aprendre representacions visuals, proporcionem una aproximació basada en estratègies de fusió per a l’aprenentatge de representacions espai- temporals, a partir de dades multi-modals obtingudes per sensors de profunditat. A més, el nostre objectiu és especialment el modelatge evolutiu i dinàmic, on el factor temporal és present en la naturalesa de les dades, com les seqüències de gestos i accions. De fet, explorem els efectes del modelatge probabilístic per aquelles aproximacions basades en programació dinàmica per a gestionar la deformació temporal i variància entre seqüències de vídeo de categories diferents. Finalment, integrem la programació dinàmica i els models generatius en un marc de computació evolutiva, amb l’objectiu d’aprendre representacions en Bosses de SubGestos i, per tant, millorar la capacitat de generalització de les aproximacions estàndards pel reconeixement de gestos. Els resultats obtinguts en l’experimentació demostra, en primer lloc, que els algorismes evolutius són útils per millorar la representació d’aproximacions BdPV en diverses bases de dades pel reconeixement de categories en imatges fxes i seqüències de vídeo. Per altra banda, la nostra experimentació revela que, tant l’ús de la programació dinàmica i els models generatius per alinear seqüències de vídeos, com les representacions obtingudes d’aplicar estratègies de fusió en dades multi-modals, comporten una millora en el rendiment a l’hora de reconèixer algunes categories de gestos. A més a més, la combinació d’algorismes evolutius amb models basats en programació dinàmica i aproximacions generatives resulten, a l’hora de classifcar categories de vídeos de bases de dades grans, en una millora considerable sobre les aproximacions estàndards de reconeixement de gestos i accions. Finalment, demostrem les aplicacions d’aquestes representacions en varis dominis per a l’anàlisi humà: classifcació d’imatges on els humans poden ser-hi presents, el reconeixement d’accions i gestos per aplicacions en general, i en particular per entorns conversacionals dins del camp de la justícia restaurativa.
El aprendizaje de la representación (o de características) ha sido un concepto emergente en los últimos años, ya que recopila un conjunto de técnicas que están presentes en cualquier metodología teórica o práctica referente a la inteligencia artificial. En la visión por computador, una representación muy comuna ha adoptado la forma de la bien conocida Bolsa de Palabras Visuales (BdPV). Esta representación aparece implícitamente en la mayoría de aproximaciones para describir imágenes, y está también presente en un enorme número de áreas y dominios: recuperación de contenido en imágenes, detección de peatones, interacción humano-ordenador, vigilancia, e-salud, y la computación social, entre otras. Las fases iniciales de esta disertación proporcionan una aproximación para aprender representaciones visuales dentro de algoritmos evolutivos, que consisten en evolucionar esquemas de pesado para mejorar las representaciones BdPV en la tarea de reconocer las categorías de vídeos e imágenes. Por lo tanto, demostramos la aplicabilidad de los esquemas de pesado más comunes, que se utilizan a menudo en la minería de textos pero se encuentran con menos frecuencia en tareas de visión por computador. Más allá de aprender representaciones visuales, proporcionamos una aproximación basada en estrategias de fusión para el aprendizaje de representaciones espacio-temporales, a partir de datos multimodales obtenidos por sensores de profundidad. También, nuestro objetivo es especialmente el modelado evolutivo y dinámico, donde el factor temporal está presente en la naturaleza de los datos, como las secuencias de gestos y acciones. De hecho, exploramos los efectos del modelado probabilístico para aquellas aproximaciones basadas en programación dinámica para gestionar la deformación temporal y varianza entre secuencias de vídeo de categorías diferentes. Finalmente, integramos la programación dinámica y los modelos generativos en un marco de computación evolutiva, con el objetivo de aprender representaciones en Bolsas de SubGestos, y por lo tanto mejorar la capacidad de generalización de las aproximaciones estándares para el reconocimiento de gestos. Los resultados obtenidos en la experimentación demuestra, en primer lugar, que los algoritmos evolutivos son útiles para mejorar la representación de aproximaciones BdPV en diversas bases de datos para el reconocimiento de categorías en imágenes fijas y secuencias de vídeo. Por otra parte, nuestra experimentación revela que, tanto el uso de la programación dinámica y los modelos generativos para alinear secuencias de vídeos, como las representaciones obtenidas de aplicar estrategias de fusión en datos multimodales, conllevan una mejora en el rendimiento a la hora de reconocer algunas categorías de gestos. Además, la combinación de algoritmos evolutivos con modelos basados en programación dinámica y aproximaciones generativas resultan, a la hora de clasificar categorías de vídeos de bases de datos grandes, en una mejora considerable sobre las aproximaciones estándares de reconocimiento de gestos y acciones. Finalmente, demostramos las aplicaciones de estas representaciones en varios dominios para el análisis humano: clasificación de imágenes donde los humanos pueden estar presentes, el reconocimiento de acciones y gestos para aplicaciones en general, y en particular para entornos conversacionales dentro del campo de la justicia restaurativa.

Denna studie syftar till att undersöka sambandet mellan kriminalitet, viktimisering och exponering för kriminogena situationer. Självrapporterad data samlades in vid tre tillfällen från 525 Malmöungdomar, varav 320 uppfyllde studiens inkluderingskriterier. Resultaten visar att mycket tid spenderad oövervakad, mycket tid ägnad åt ostrukturerade aktiviteter, mycket tid i sällskap med vänner samt alkoholkonsumtion samvarierar med brottslighet och viktimisering i varierande utsträckning. Sambanden varierar dock i förhållande till de båda utfallsvariablerna och deltagarnas ålder. Livsstils-rutinaktivitetsteorin kan förklara resultaten men behöver i framtiden ta större hänsyn till ålder. Studiens två slutsatser är att (1) brottslighet och viktimisering bör betraktas som två olika men klart relaterade företeelser i förhållande till exponering för kriminogena situationer och att (2) ålder måste tas i beaktande i forskning om exponering för kriminogena situationer eftersom sambanden mellan exponering och de båda utfallsvariablerna varierar från tidiga till sena tonår.
This study aims to examine offending and victimization in relation to situational criminogenic exposure. Self-reported data was collected at three occasions from a sample of 525 adolescents in Malmö, of which 320 fulfilled the study’s inclusion criteria. The results show that spending a lot of time unsupervised, pursuing unstructured activities, spending a lot of time with peers, and alcohol use, are associated with offending and victimization to various extent. However, the associations vary according to outcome and in relation to the participants’ age. Lifestyle-Routine Activities Theory may explain the findings, but needs to consider age as an important factor in the future. The two conclusions from this study are that (1) offending and victimization should be treated as two different, yet related concepts in relation to situational criminogenic exposure, and that (2) it is important to add an age dimension to the study of situational criminogenic exposure because the associations between the exposure variables and the outcome variables vary from early to late adolescence.

In neuropsychiatry, the development of brain imaging and dedicated data analysis for personalized medicine promises to predict both the evolution of diseases and responses of treatments. The ability to estimate the time course of the disease is the first step to understand the response to potential treatments, which implies the development of methods able to capture subject-specific features in addition to the discrimination between pathological conditions. However, methods that effectively characterize the neuronal activity at the whole-brain level are still lacking, and many efforts are currently made in the fields of clinical research and neuroscience to fill this gap. The above is particularly problematic to interpret functional Magnetic Resonance Imaging (fMRI) data, which are indirectly coupled with neuronal activity because of hemodynamics, yielding much slower signals than neuronal activity. We propose a multiscale method that combines a computational whole-brain model with machine learning to solve this issue. In our approach, the model relates the neuronal activity and the fMRI signals in a mechanistic fashion, allowing for access to neuronal activity down to millisecond precision. Specifically, we use a novel methodology that allows the extraction of space-time motifs at different timescales through binned time windows. Then, we use machine learning to study which range of timescales in the modeled neuronal activity is most informative to separate the brain's dynamics during rest, distinguishing subjects, tasks, and neuropsychiatric conditions. Our multiscale computational approach is a further step to study the multiple timescales of brain dynamics and predict the dynamical interactions between brain regions. Overall, this method raises outlooks to detect biomarkers and predict responses of treatments.
En neuropsiquiatría, el desarrollo de imágenes cerebrales y el análisis de datos dedicados a la medicina personalizada prometen predecir tanto la evolución de las enfermedades como las respuestas a los tratamientos. La capacidad de estimar el curso temporal de la enfermedad es el primer paso para comprender la respuesta a posibles tratamientos, lo que implica el desarrollo de métodos capaces de capturar características específicas del sujeto, además de la discriminación entre condiciones patológicas. Sin embargo, todavía faltan métodos que caractericen eficazmente la actividad neuronal a nivel de todo el cerebro, y actualmente se están haciendo muchos esfuerzos en los campos de la investigación clínica y la neurociencia. Lo anterior es particularmente problemático para interpretar los datos funcionales de las imágenes de resonancia magnética (fMRI por sus siglas en inglés), que están acoplados indirectamente con la actividad neuronal debido a la hemodinámica, lo que produce señales mucho más lentas que la actividad neuronal. En este trabajo, proponemos un método multiescala que combina un modelo computacional de cerebro completo con aprendizaje automático para resolver este problema. En nuestro enfoque, el modelo relaciona la actividad neuronal y las señales de resonancia magnética funcional de manera mecanicista, lo que permite el acceso a la actividad neuronal con una precisión de milisegundos. Específicamente, utilizamos una nueva metodología que permite la extracción de patrones espacio-temporales en diferentes escalas temporales a través de ventanas de tiempo. Después, usamos aprendizaje automático para estudiar qué rango de escalas de tiempo en la actividad neuronal modelada es más informativo, para separar la dinámica del cerebro durante el descanso, distinguiendo sujetos, tareas y condiciones neuropsiquiátricas. Nuestro enfoque computacional multiescala es un paso más para estudiar las múltiples escalas de tiempo de la dinámica del cerebro y predecir las interacciones dinámicas entre las regiones del cerebro. En general, este método aumenta las perspectivas para detectar biomarcadores y predecir la respuesta de tratamientos.

Tato disertační práce se zabývá vylepšením systémů pro rozpoznávání činností člověka. Současný stav vědění v této oblasti jest prezentován. Toto zahrnuje způsoby získávání digitálních obrazů a videí společně se způsoby reprezentace těchto entit za použití počítače. Dále jest prezentováno jak jsou použity extraktory příznakových vektorů a extraktory pros- torově-časových příznakových vektorů a způsoby přípravy těchto dat pro další zpracování. Příkladem následného zpracování jsou klasifikační metody. Pro zpracování se obecně obvykle používají části videa s proměnlivou délkou. Hlavní přínos této práce je vyřčená hypotéza o optimální délce analýzy video sekvence, kdy kvalita řešení je porovnatelná s řešením bez restrikce délky videosekvence. Algoritmus pro ověření této hypotézy jest navržen, implementován a otestován. Hypotéza byla experimentálně ověřena za použití tohoto algoritmu. Při hledání optimální délky bylo též dosaženo jistého zlepšení kvality klasifikace. Experimenty, výsledky a budoucí využití této práce jsou taktéž prezentovány.

Software Product Lines (SPLs) and Software Ecosystems (SECOs) are approaches to capturing families of closely related software systems in terms of common and variable functionality (variability in space). SPLs and especially SECOs are subject to software evolution to adapt to new or changed requirements resulting in different versions of the software family and its variable assets (variability in time). Both dimensions may be interconnected (e.g., through version incompatibilities) and, thus, have to be handled simultaneously as not all customers upgrade their respective products immediately or completely. However, there currently is no integrated approach allowing variant derivation of features in different version combinations. In this thesis, remedy is provided in the form of an integrated approach making contributions in three areas: (1) As variability model, Hyper-Feature Models (HFMs) and a version-aware constraint language are introduced to conceptually capture variability in time as features and feature versions. (2) As variability realization mechanism, delta modeling is extended for variability in time, and a language creation infrastructure is provided to devise suitable delta languages. (3) For the variant derivation procedure, an automatic version selection mechanism is presented as well as a procedure to derive large parts of the application order for delta modules from the structure of the HFM. The presented integrated approach enables derivation of concrete software systems from an SPL or a SECO where both features and feature versions may be configured.

Design space exploration (DSE) is a tool used to evaluate and compare designs as part of the design selection process. While evaluating every possible design in a design space is infeasible, understanding design behavior and response throughout the design space may be accomplished by evaluating a subset of designs and interpolating between them using surrogate models. Surrogate modeling is a technique that uses low-cost calculations to approximate the outcome of more computationally expensive calculations or analyses, such as finite element analysis (FEA). While surrogates make quick predictions, accuracy is not guaranteed and must be considered. This research addressed the need to improve the accuracy of surrogate predictions in order to improve DSE of structural responses. This was accomplished by performing comparative analyses of linear and nonlinear dimensionality-reduction-based radial basis function (RBF) surrogate models for emulating various FEA nodal results. A total of four dimensionality reduction methods were investigated, namely principal component analysis (PCA), kernel principal component analysis (KPCA), isometric feature mapping (ISOMAP), and locally linear embedding (LLE). These methods were used in conjunction with surrogate modeling to predict nodal stresses and coordinates of a compressor blade. The research showed that using an ISOMAP-based dual-RBF surrogate model for predicting nodal stresses decreased the estimated mean error of the surrogate by 35.7% compared to PCA. Using nonlinear dimensionality-reduction-based surrogates did not reduce surrogate error for predicting nodal coordinates. A new metric, the manifold distance ratio (MDR), was introduced to measure the nonlinearity of the data manifolds. When applied to the stress and coordinate data, the stress space was found to be more nonlinear than the coordinate space for this application. The upfront training cost of the nonlinear dimensionality-reduction-based surrogates was larger than that of their linear counterparts but small enough to remain feasible. After training, all the dual-RBF surrogates were capable of making real-time predictions. This same process was repeated for a separate application involving the nodal displacements of mode shapes obtained from a FEA modal analysis. The modal assurance criterion (MAC) calculation was used to compare the predicted mode shapes, as well as their corresponding true mode shapes obtained from FEA, to a set of reference modes. The research showed that two nonlinear techniques, namely LLE and KPCA, resulted in lower surrogate error in the more complex design spaces. Using a RBF kernel, KPCA achieved the largest average reduction in error of 13.57%. The results also showed that surrogate error was greatly affected by mode shape reversal. Four different approaches of identifying reversed mode shapes were explored, all of which resulted in varying amounts of surrogate error. Together, the methods explored in this research were shown to decrease surrogate error when performing DSE of a turbomachine compressor blade. As surrogate accuracy increases, so does the ability to correctly make engineering decisions and judgements throughout the design process. Ultimately, this will help engineers design better turbomachines.

Ionospheric scintillation is a rapid variation in the amplitude and/or phase of radio signals traveling through the ionosphere. This spatial and time-varying phenomenon is of interest because its occurrence may affect the reception quality of satellite signals. Specialized receivers at strategic regions can track multiple variables related to the phenomenon, generating a database of historical observations on the regional behavior of ionospheric scintillation. The analysis of such data is very challenging, since it consists of time-varying measurements of many variables which are heterogeneous in nature and with possibly many missing values, recorded over extensive time periods. There is a need to introduce alternative intuitive strategies that contribute to experts acquiring further knowledge from the ionospheric scintillation data. Such challenges motivated a study on the applicability of visualization techniques to support tasks of identification of relevant attributes in the study of the behavior of phenomena described by multiple time-varying variables, of which the ionospheric scintillation is a good example. In particular, this thesis introduces a visual analytics framework, named TV-MV Analytics, that supports exploratory tasks on time-varying multivariate data and was developed following the requirements of experts on ionospheric scintillation from the Faculty of Science and Technology of UNESP at Presidente Prudente, Brazil. TV-MV Analytics provides an interactive visual exploration loop to analysts inspecting the behavior of multiple variables at different temporal scales, through temporal representations associated with clustering and multidimensional projection techniques. Analysts can also assess how different feature sub-spaces contribute to characterizing a certain behavior, where they may direct the analysis process and include their domain knowledge in the exploratory analysis. We also illustrate the application of TV-MV Analytics on multivariate time-varying data sets from three alternative application domains. Experimental results indicate the proposed solutions show good potential on assisting time-varying multivariate data mining tasks, since it reduces the effort required from experts to gain deeper insight into the historical behavior of the variables describing a phenomenon or domain.
A cintilação ionosférica é uma variação rápida na amplitude e/ou na fase dos sinais de rádio que viajam através da ionosfera. Este fenômeno espacial e variante no tempo é de grande interesse, pois pode afetar a qualidade de recepção dos sinais de satélite. Receptores especializados em regiões estratégicas podem rastrear múltiplas variáveis relacionadas ao fenômeno, gerando um banco de dados de observações históricas sobre o comportamento regional da cintilação. O estudo do comportamento da cintilação é desafiador, uma vez que requer a análise extensiva de dados multivariados e variantes no tempo, coletados por longos períodos. Medições são registradas continuamente, e são de natureza heterogênea, compreendendo múltiplas variáveis de diferentes categorias e possivelmente com muitos valores faltantes. Portanto, existe a necessidade de introduzir estratégias alternativas, eficientes e intuitivas, que contribuam para a adquisição de conhecimento, a partir dos dados, por especialistas que estudam a cintilação ionosférica. Tais desafios motivaram o estudo da aplicabilidade de técnicas de visualização para apoiar tarefas de identificação de atributos relevantes no estudo do comportamento de fenômenos ou domínios que envolvem múltiplas variáveis, como a cintilação. Em particular, esta tese introduz um arcabouço visual, o qual foi denominado TV-MV Analytics, que apoia tarefas de análise exploratória sobre dados multivariados e variáveis no tempo, inspirado em requisitos de especialistas no estudo da cintilação, vinculados à Faculdade de Ciências e Tecnologia da UNESP de Presidente Prudente, Brasil. O TV-MV Analytics fornece aos analistas um ciclo de interativo de exploração que apoia a inspeção do comportamento temporal de múltiplas variáveis, em diferentes escalas temporais, por meio de representações visuais temporais associadas a técnicas de agrupamento e de projeção multidimensional. Também permite avaliar como diferentes sub-espaços de atributos caracterizam um determinado comportamento, podendo direcionar o processo de análise e inserir seu conhecimento do domínio no processo de análise exploratória. As funcionalidades do TV-MV Analytics também são ilustradas em dados variantes no tempo oriundos de outros três domínios de aplicação. Os resultados experimentais indicaram que as soluções propostas têm bom potencial em tarefas de mineração de dados multivariados e variantes no tempo, uma vez que reduz o esforço e contribui para os especialistas obterem informações detalhadas sobre o comportamento histórico das variáveis que descrevem um determinado fenômeno ou domínio.

Avec l'augmentation considérable d'acquisitions de données temporelles dans les dernières décennies comme les systèmes GPS, les séquences vidéo ou les suivis médicaux de pathologies ; le besoin en algorithmes de traitement et d'analyse efficaces d'acquisition longitudinales n'a fait qu'augmenter. Dans cette thèse, nous proposons une extension du formalisme mean-shift, classiquement utilisé en traitement d'images, pour le groupement de séries temporelles multidimensionnelles. Nous proposons aussi un algorithme de groupement hiérarchique des séries temporelles basé sur la mesure de dynamic time warping afin de prendre en compte les déphasages temporels. Ces choix ont été motivés par la nécessité d'analyser des images acquises en imagerie par résonance magnétique sur des patients atteints de sclérose en plaques. Cette maladie est encore très méconnue tant dans sa genèse que sur les causes des handicaps qu'elle peut induire. De plus aucun traitement efficace n'est connu à l'heure actuelle. Le besoin de valider des hypothèses sur les lésions de sclérose en plaque nous a conduit à proposer des méthodes de groupement de séries temporelles ne nécessitant pas d'a priori sur le résultat final, méthodes encore peu développées en traitement d'images
Due to the dramatic increase of longitudinal acquisitions in the past decades such as video sequences, global positioning system (GPS) tracking or medical follow-up, many applications for time-series data mining have been developed. Thus, unsupervised time-series data mining has become highly relevant with the aim to automatically detect and identify similar temporal patterns between time-series. In this work, we propose a new spatio-temporal filtering scheme based on the mean-shift procedure, a state of the art approach in the field of image processing, which clusters multivariate spatio-temporal data. We also propose a hierarchical time-series clustering algorithm based on the dynamic time warping measure that identifies similar but asynchronous temporal patterns. Our choices have been motivated by the need to analyse magnetic resonance images acquired on people affected by multiple sclerosis. The genetics and environmental factors triggering and governing the disease evolution, as well as the occurrence and evolution of individual lesions, are still mostly unknown and under intense investigation. Therefore, there is a strong need to develop new methods allowing automatic extraction and quantification of lesion characteristics. This has motivated our work on time-series clustering methods, which are not widely used in image processing yet and allow to process image sequences without prior knowledge on the final results

In automatic speech recognition systems (ASRs), training is a critical phase to the system?s success. Communication media, either analog (such as analog landline phones) or digital (VoIP) distort the speaker?s speech signal often in very complex ways: linear distortion occurs in all channels, either in the magnitude or phase spectrum. Non-linear but time-invariant distortion will always appear in all real systems. In digital systems we also have network effects which will produce packet losses and delays and repeated packets. Finally, one cannot really assert what path a signal will take, and so having error or distortion in between is almost a certainty. The channel introduces an acoustical mismatch between the speaker's signal and the trained data in the ASR, which results in poor recognition performance. The approach so far, has been to try to undo the havoc produced by the channels, i.e. compensate for the channel's behavior. In this thesis, we try to characterize the effects of different transmission media and use that as an inexpensive and repeatable way to train ASR systems.

This thesis deals with various aspects of the analysis of video sequences. The first problem this work deals with is the detection of known categories of events. The second problem this work addresses is the retrieval of events of interest regardless of their specific nature. Finally we propose a method that can leverage both semantic high level features and low level image features to reduce the disk space and bandwidth needed.

Software Product Lines (SPLs) and Software Ecosystems (SECOs) are approaches to capturing families of closely related software systems in terms of common and variable functionality (variability in space). SPLs and especially SECOs are subject to software evolution to adapt to new or changed requirements resulting in different versions of the software family and its variable assets (variability in time). Both dimensions may be interconnected (e.g., through version incompatibilities) and, thus, have to be handled simultaneously as not all customers upgrade their respective products immediately or completely. However, there currently is no integrated approach allowing variant derivation of features in different version combinations. In this thesis, remedy is provided in the form of an integrated approach making contributions in three areas: (1) As variability model, Hyper-Feature Models (HFMs) and a version-aware constraint language are introduced to conceptually capture variability in time as features and feature versions. (2) As variability realization mechanism, delta modeling is extended for variability in time, and a language creation infrastructure is provided to devise suitable delta languages. (3) For the variant derivation procedure, an automatic version selection mechanism is presented as well as a procedure to derive large parts of the application order for delta modules from the structure of the HFM. The presented integrated approach enables derivation of concrete software systems from an SPL or a SECO where both features and feature versions may be configured.:I. Context and Preliminaries 1. The Configurable TurtleBot Driver as Running Example 1.1. TurtleBot: A Domestic Service Robot 1.2. Configurable Driver Functionality 1.3. Software Realization Artifacts 1.4. Development History of the Driver Software 2. Families of Variable Software Systems 2.1. Variability 2.1.1. Variability in Space and Time 2.1.2. Internal and External Variability 2.2. Manifestations of Configuration Knowledge 2.2.1. Variability Models 2.2.2. Variability Realization Mechanisms 2.2.3. Variability in Realization Assets 2.3. Types of Software Families 2.3.1. Software Product Lines 2.3.2. Software Ecosystems 2.3.3. Comparison of Software Product Lines and Software Ecosystems 3. Fundamental Approaches and Technologies of the Thesis 3.1. Model-Driven Software Development 3.1.1. Metamodeling Levels 3.1.2. Utilizing Models in Generative Approaches 3.1.3. Representation of Languages using Metamodels 3.1.4. Changing the Model-Representation of Artifacts 3.1.5. Suitability of Model-Driven Software Development 3.2. Fundamental Variability Management Techniques of the Thesis 3.2.1. Feature Models as Variability Models 3.2.2. Delta Modeling as Variability Realization Mechanism 3.2.3. Variant Derivation Process of Delta Modeling with Feature Models 3.3. Constraint Satisfaction Problems 3.4. Scope 3.4.1. Problem Statement 3.4.2. Requirements 3.4.3. Assumptions and Boundaries II. Integrated Management of Variability in Space and Time 4. Capturing Variability in Space and Time with Hyper-Feature Models 4.1. Feature Models Cannot Capture Variability in Time 4.2. Formal Definition of Feature Models 4.3. Definition of Hyper-Feature Models 4.4. Creation of Hyper-Feature Model Versions 4.5. Version-Aware Constraints to Represent Version Dependencies and Incompatibilities 4.6. Hyper-Feature Models are a True Extension to Feature Models 4.7. Case Study 4.8. Demarcation from Related Work 4.9. Chapter Summary 5. Creating Delta Languages Suitable for Variability in Space and Time 5.1. Current Delta Languages are not Suitable for Variability in Time 5.2. Software Fault Trees as Example of a Source Language 5.3. Evolution Delta Modules as Manifestation of Variability in Time 5.4. Automating Delta Language Generation 5.4.1. Standard Delta Operations Realize Usual Functionality 5.4.2. Custom Delta Operations Realize Specialized Functionality 5.5. Delta Language Creation Infrastructure 5.5.1. The Common Base Delta Language Provides Shared Functionality for all Delta Languages 5.5.2. Delta Dialects Define Delta Operations for Custom Delta Languages 5.5.3. Custom Delta Languages Enable Variability in Source Languages 5.6. Case Study 5.7. Demarcation from Related Work 5.8. Chapter Summary 6. Deriving Variants with Variability in Space and Time 6.1. Variant Derivation Cannot Handle Variability in Time 6.2. Associating Features and Feature Versions with Delta Modules 6.3. Automatically Select Versions to Ease Configuration 6.4. Application Order and Implicitly Required Delta Modules 6.4.1. Determining Relevant Delta Modules 6.4.2. Forming a Dependency Graph of Delta Modules 6.4.3. Performing a Topological Sorting of Delta Modules 6.5. Generating Variants with Versions of Variable Assets 6.6. Case Study 6.7. Demarcation from Related Work 6.8. Chapter Summary III. Realization and Application 7. Realization as Tool Suite DeltaEcore 7.1. Creating Delta Languages 7.1.1. Shared Base Metamodel 7.1.2. Common Base Delta Language 7.1.3. Delta Dialects 7.2. Specifying a Software Family with Variability in Space and Time 7.2.1. Hyper-Feature Models 7.2.2. Version-Aware Constraints 7.2.3. Delta Modules 7.2.4. Application-Order Constraints 7.2.5. Mapping Models 7.3. Deriving Variants 7.3.1. Creating a Configuration 7.3.2. Collecting Delta Modules 7.3.3. Ordering Delta Modules 7.3.4. Applying Delta Modules 8. Evaluation 8.1. Configurable TurtleBot Driver Software 8.1.1. Variability in Space 8.1.2. Variability in Time 8.1.3. Integrated Management of Variability in Space and Time 8.2. Metamodel Family for Role-Based Modeling and Programming Languages 8.2.1. Variability in Space 8.2.2. Variability in Time 8.2.3. Integrated Management of Variability in Space and Time 8.3. A Software Product Line of Feature Modeling Notations and Constraint Languages 8.3.1. Variability in Space 8.3.2. Variability in Time 8.3.3. Integrated Management of Variability in Space and Time 8.4. Results and Discussion 8.4.1. Results and Discussion of RQ1: Variability Model 8.4.2. Results and Discussion of RQ2: Variability Realization Mechanism 8.4.3. Results and Discussion of RQ3: Variant Derivation Procedure 9. Conclusion 9.1. Discussion 9.1.1. Supported Evolutionary Changes 9.1.2. Conceptual Representation of Variability in Time 9.1.3. Perception of Versions as Incremental 9.1.4. Version Numbering Schemes 9.1.5. Created Delta Languages 9.1.6. Scalability of Approach 9.2. Possible Future Application Areas 9.2.1. Extend to Full Software Ecosystem Feature Model 9.2.2. Model Software Ecosystems 9.2.3. Extract Hyper-Feature Model Versions and Record Delta Modules 9.2.4. Introduce Metaevolution Delta Modules 9.2.5. Support Incremental Reconfiguration 9.2.6. Apply for Evolution Analysis and Planning 9.2.7. Enable Evolution of Variable Safety-Critical Systems 9.3. Contribution 9.3.1. Individual Contributions 9.3.2. Handling Updater Stereotypes IV. Appendix A. Delta Operation Generation Algorithm B. Delta Dialects B.1. Delta Dialect for Java B.2. Delta Dialect for Eclipse Projects B.3. Delta Dialect for DocBook Markup B.4. Delta Dialect for Software Fault Trees B.5. Delta Dialect for Component Fault Diagrams B.6. Delta Dialect for Checklists B.7. Delta Dialect for the Goal Structuring Notation B.8. Delta Dialect for EMF Ecore B.9. Delta Dialect for EMFText Concrete Syntax Files

碩士
國立東華大學
自然資源與環境學系
103
The disaster occurrences are likely the results of the interaction between the human and environment. By comprehending and organizing systematical characteristics of past disasters which would guide the future for developing a more effective way of management in land use and human behaviors. Those sensitivity disaster-prone areas are impacting socio-economic environment of the Coorong. Therefore, it is imperatively important that the correlation should be recognized between the associated characteristics of disasters and the human. Taiwan has been suffering the most frequent disasters in the world due to its geographical location. In the recent years, disasters had increasingly resulted casualties and economic losses which mainly due to heavy rainfalls from the typhoon. Approximately, a 22 percent of typhoons per year were directly attacking to the eastern part of Taiwan. Hualien areas are impacted by the disasters and resulted in a considerable loss. Ruisui Township is located in the eastern basin of Hsiukuluan River where the main population gathering area of southern Hualien is. This research utilized the historical disaster records from 1953 to 2013 to investigate the interaction and relationship between human and environment. The data analysis tools are Arc GIS and Excel software. The data was analyzed and offered the effective perspectives which might reduce damages from the disasters. Moreover, disaster’s space-time trend and feature were also analyzed. In addition, this study investigated the frequent human activities through indicators to assess the terrain locations in order to identify the association between various regional disasters and the highest sensitivity regions against hazards. The results showed that the disaster occurrences have been extending to the eastward of Ruisui Township since the year of 1980. The highest density area of the disasters is located at Hualien County Road 64 where the Erosion-type of disaster occurred between Ruigane road between Wuhe platform northernmost Maliyun tribe. The highest to the least sensitivity locations appeared to be on fan valley, followed by alluvial terraces, and then the plains. In conclusion, this study found that the primary relationship between Ruisui Township and human activities caused disasters from the houses and roads construction in environmental impacts. In the future, it is recommended that the land development of choices should be avoided in highly sensitive areas.