Orientador: Carlos Roberto Minussi / Banca: Anna Diva Plasencia Lotufo / Banca: Mara Lúcia Martins Lopes / Banca: Arlan Luiz Bettiol / Banca: Edmárcio Antonio Belati / Resumo: Muitos esforços têm sido despendidos para tentar sanar problemas relacionados com Qualidade da Energia Elétrica (QEE), principalmente na automação de processos e desenvolvimento de equipamentos de monitorização que possibilitem maior desempenho e confiabilidade a todo o Sistema Elétrico. Esta pesquisa apresenta um sistema eficiente de identificador/classificador automático de distúrbios chamado de Rede Neuro-Fuzzy-Wavelet. A estrutura básica dessa rede é composta por três módulos: o módulo de detecção de anomalias onde os sinais com distúrbios são identificados, o módulo de extração de características onde as formas de onda com distúrbio são analisadas, e o módulo de classificação que conta com uma rede neural ARTMAP Fuzzy, a qual indica qual o tipo de distúrbio sofrido pelo sinal. Os tipos de distúrbios incluem os isolados de curto prazo, tais como: afundamento de tensão (sag), elevação de tensão (swell), os distúrbios de longo prazo como distorção harmônica, bem como distúrbios múltiplos simultâneos como afundamento de tensão com distorção harmônica e elevação de tensão com distorção harmônica. A concepção do sistema de inferência (neural wavelet ARTMAP fuzzy) permite realizar a classificação dos referidos distúrbios de forma robusta e com grande rapidez na obtenção das soluções. Testes apontam para o alto desempenho dessa rede na detecção e classificação correta dos tipos de distúrbios de tensão analisados, 100% de acerto. A forma robusta e grande rapidez na obtenção dos resultados, possibilita sua aplicação em tempo real, visto que o esforço computacional, muito pequeno, é alocado, basicamente, na fase de treinamento. Somente uma pequena parcela de tempo computacional é necessária para a efetivação das análises. Além do mais, a metodologia proposta pode ser estendida para a realização de tarefas mais complexas... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: Many efforts have been spent to solve problems related to Power Quality (PQ), principally in process automation and developing monitoring equipments that can provide more reliability and behavior for the electrical system. This research presents an efficient automatic system to identify/classify disturbs by Fuzzy Wavelet Neural Network. The basic structure of this neural network is composed of three modules such as: module for detecting anomalies where the signals with disturbs are identified, module for extracting the characteristics where the wave forms with disturbs are analyzed, and the module of classification that contains a fuzzy ARTMAP neural network that shows the type of disturbs existing in the signal. The types of disturbs include the short term isolated ones which are: voltage dip (sag), voltage increasing (swell); the long term disturbs such as harmonic distortion as well as the multiple simultaneous ones like the voltage dip with harmonic distortion and voltage increasing with harmonic distortion. The inference system (neural wavelet ARTMAP fuzzy) allows executing the classification of the cited disturbs very fast and obtaining reliable results. This neural network provides high performance when classifying and detecting the voltage disturbs very fast with about 100% of accuracy. The speed in obtaining the results allows an application in real time due to a low computational effort, which is basically in the training phase of the neural network. A little time of the computational effort is spent for the analysis. Moreover the proposed methodology can be used for realizing more complex tasks, as for example the localization of the power sources of the voltage disturbs. It is a very important contribution in the power quality, mainly to be a needy activity for solutions on the specialized literature / Doutor

Proteger as linhas de transmissão é uma das tarefas mais importantes dentro dos sistemas elétricos de potência. Faltas em linhas de transmissão devem ser localizadas precisamente e extintas o mais rápido possível. Para tal, o esquema de proteção de linhas utiliza valores amostrados de correntes e tensões para a execução das tarefas de detecção, classificação e localização da falta. Neste esquema, grandezas trifásicas de corrente (IA, IB, IC) e tensão (VA, VB, VC) compõem as entradas do sistema. Após a detecção e classificação da falta, o relé efetua o cálculo da impedância aparente para a verificação da zona de proteção na qual a falta se insere (localização). Dentro deste contexto, a rápida detecção e a correta classificação da falta são passos fundamentais para a lógica de controle de um relé. Para a utilização de sistemas de proteção com alta velocidade de operação, o conjunto detector + classificador deve realizar uma decisão precisa do tipo de falta envolvida em menos de 10 ms após a ocorrência desta. Alguns métodos convencionais têm lidado com este problema, porém os tempos para estimação do tipo de falta são algumas vezes excessivamente longos. Este trabalho apresenta um novo sistema que provê uma rápida e confiável detecção e classificação de faltas através das medidas de valores de correntes trifásicas. O novo método utiliza-se da teoria de Redes Neurais Artificiais, baseada em dois diferentes tipos de redes (MLP e RBF), para a tarefa de detecção e classificação de faltas nos níveis de tempo requeridos para um moderno sistema de proteção. Um estudo comparativo em relação ao desempenho das redes mencionadas também foi realizado. Os testes efetuados para as redes dos tipos MLP e RBF mostraram que o sistema proposto foi capaz de detectar e classificar corretamente 100% dos casos estudados. Deve ainda ser ressaltado, que na maior parte dos casos (93% para a rede MLP e 84% para a rede RBF), o processo de detecção e classificação foi completado com no máximo 5 amostras de pós-falta (5ms). Isto demonstra a rapidez na tarefa de detecção e classificação embutida no método proposto, principalmente levando-se em consideração os tempos apresentados pelos métodos convencionais. / Transmission line protection is one of the major tasks for a power system. Transmission line faults must be located accurately and isolated as fast as possible. In order to perform this task, the power system protection system uses the three-phase currents (IA, IB, IC) and voltages (VA, VB, VC) to detect, classify and locate the fault. After detecting and classifying the fault, the relay calculates the apparent impedance to verify in which protection zone the fault is located. Taking this into account, precise and fast detection and classification methods are fundamental steps for the relay control algorithm. The combination detection + classification must carry out the correct response in less than 10 ms after the fault for a high-speed protection system. Some conventional methods have treated this problem but the time for a correct classification is sometimes excessively long. This work presents a fast and reliable new system for fault detection and classification using the three-phase current measurements. This new system is based on Artificial Neural Networks (RBF and MLP) for the detection and classification tasks. A comparative study involving both types of neural networks was done. Tests showed that the proposed system was able to correctly detect and classify 100% of the studied cases where the majority (93% of the cases for MLP net and 84% for RBF net) of them was done in up to 5 post-fault samples (5 ms). The afore-mentioned demonstrates the high speed of the new method for the detection and classification tasks when compared to the conventional ones.

Tato práce se zabývá a analyzuje možnosti, kterými je možno zjistit geografickou polohu ze snímků nebo videa pouze za použití vizuální informace z obrazu.    Výsledkem práce jsou dvě rozdílné metody geo-lokalizace. První z nich pracuje na principu hledání časů východu a západu Slunce. Hlavní výhodou této metody je její univerzálnost. Funguje s jakoukoliv kamerou umístěnou v externích prostorech a nevyžaduje přítomnost žádných specifických objektů ve scéně. Pro správný výsledek je třeba alespoň celodenní záznam z kamery. Výsledky jsou uspokojivé za každého počasí. Druhá metoda pracuje na základě analýzy stínů ve scéně. Správnou pozici je možno určit, s poměrně velkou přesností, pouze na základě dvou snímků pořízených v různém čase. Tato metoda vyžaduje přítomnost dvou objektů v obraze, které vrhají stín. Přesnosti výsledků navržených metod jsou vyhodnoceny a porovnány. Z výsledků vyplývá, že obě metody lze úspěšně použít pro odhad geografické polohy. Dále byla v rámci práce pořízena rozsáhlá datová sada obrazových sekvencí z volně přístupných webových kamer.

Apresenta-se uma metodologia que tem como base a análise de campos magnéticos no monitoramento da qualidade da energia de sistemas elétricos. Em particular, são avaliados os aspectos referentes à detecção de faltas em sistemas elétricos. Diferente do processo de monitoração tradicional, cujos sensores precisam estar fisicamente conectados aos circuitos analisados, propõe-se estudar a viabilidade da utilização dos sinais provenientes da assinatura magnética resultante no exame do desempenho dos sistemas elétricos. Ressalta-se, assim, a característica não invasiva deste processo. Em uma primeira instância, simulações numéricas e medidas experimentais são usadas para estimar a validade deste método. Com base em valores das correntes de falta fase-terra, relacionados a configurações reais de sistemas de distribuição, provenientes de simulações numéricas e disponibilizadas na literatura, são calculados os campos magnéticos em regiões pré-selecionadas próximas às linhas. A seguir, aplicam-se os conceitos relacionados a wavelets no tratamento dos sinais resultantes. É nesta etapa que, por meio da decomposição da assinatura magnética correspondente, serão obtidos os dados necessários para se correlacionar os componentes dos sinais ao diagnóstico das faltas, nos sistemas elétricos. A Análise de múltirresolução é aplicada. Além destes resultados teóricos, aqueles provenientes de uma bancada experimental são examinados. Algumas configurações canônicas foram pré-selecionadas, visando estudar a eventual influência dos aspectos geométricos nos resultados relacionados à decomposição do sinal em análise. Embora métodos analíticos pudessem ser empregados na determinação da assinatura magnética resultante, os métodos numéricos, tais como o método dos elementos finitos, foram utilizados visando agilizar a obtenção de resultados teóricos a serem avaliados. Da mesma forma, aplicativos já disponibilizados comercialmente foram utilizados na decomposição dos sinais. Esta metodologia foi aplicada, também, para identificar faltas, aplicando-se a análise da variância para os diversos níveis do detalhe wavelet. A validação da metodologia foi feita pela comparação entre os resultados simulados e obtidos experimentalmente. / A methodology based on the analysis of magnetic fields for monitoring the quality of energy in electrical systems is presented herein. Aspects referring to fault detection in electrical systems in particular are evaluated. Contrary to the traditional monitoring process, in which sensors must be physically linked to the circuits under analysis, the results are presented from a feasibility study on the use of signals arising from the resulting magnetic signature by means of the electrical systems analysis. Thus the non-invasive characteristic of this process should be pointed out. First, numerical simulations and experimental measures were used to estimate the validity of this method. Based on values of the current of phase-earth fault related to actual features of the distribution systems and derived from numeric simulations found in the literature, the magnetic fields, in pre-established regions, were calculated. Following this, the concepts related to wavelets in the treatment of resulting signals were applied. It is in this phase that, by means of the decomposition of the corresponding magnetic signature, the data necessary to correlate the signal components for the diagnosis of faults in electrical systems were obtained. A Multiresolution Analysis (MRA) was applied. In addition to these theoretical results, the results from a laboratory workbench were also examined. Some canonical features were pre-selected, aiming to study the influence of geometric aspects on the results related to the signal decomposition analyzed. Although analytical methods could be employed to determine the resulting magnetic signature, numerical methods, such as the finite element method, were used to expedite obtaining the theoretical results to be analyzed. Likewise, commercial software was also used for the decomposition of signals. This methodology was validated by comparing the measured and simulated magnetic flux density. This methodology was also applied to identify and classify faults by means of the variance curve towards the wavelet detail.

A presença de distúrbios na energia elétrica fornecida aos consumidores pode causar a diminuição no tempo de vida útil dos equipamentos, mal funcionamento ou até mesmo sua perda. Desse modo, ferramentas capazes de realizar a detecção, localização, classificação, compressão e o armazenamento de sinais de forma automática e organizada são essenciais para garantir um processo de monitoramento adequado ao sistema elétrico de potência como um todo. Dentre as ferramentas comumente aplicadas às tarefas supramencionadas, pode-se destacar a Transformada Wavelet (TW) e as Redes Neurais Artificiais (RNAs). Contudo, ainda não foi estabelecida uma metodologia para obtenção e validação da TW e seu nível de decomposição, bem como da arquitetura e da topologia de RNAs mais apropriadas às tarefas supracitadas. O principal fato que levou a esta constatação deve-se à análise da literatura correlata, onde é possível notar o uso de distintas TW e RNAs. Neste contexto, a primeira contribuição desta pesquisa foi o projeto e desenvolvimento de um método eficiente de segmentação de sinais com distúrbios associados à Qualidade da Energia Elétrica (QEE). O método desenvolvido se beneficia das propriedades da TW de identificação temporal de descontinuidades em sinais. A segunda contribuição é o desenvolvimento de um algoritmo automático que, por meio do método de segmentação desenvolvido e de classificação por RNAs, indique as melhores ferramentas (Wavelets e RNAs) para as tarefas de segmentação, extração de características e classificação de distúrbios de QEE. Esse algoritmo foi desenvolvido com base nos recursos dos Algoritmos Evolutivos (AEs) e adotou RNAs do tipo Perceptron Multicamadas, pois, esta arquitetura pode ser considerada consagrada no que se refere à classificação de padrões. Por fim, a terceira contribuição é relativa ao desenvolvimento de um procedimentos baseados em AEs, a fim de se aprimorar métodos de compressão de dados que preservem as informações relevantes nos sinais de QEE. Assim, é importante mencionar que os resultados dessa pesquisa poderão determinar mecanismos automáticos a serem utilizados no processo de registro, tratamento e armazenamento de informações que serão importantes para se manter um banco de dados (histórico) atualizado nas concessionárias de energia, a partir do qual, índices e um melhor mapeamento e entendimento de todos os distúrbios relacionados à QEE poderão ser melhor entendidos e solucionados. / The presence of disturbances in the electrical power supplied to consumers can decrease the lifetime of the equipment, cause malfunction or even their breakdown. Thus, tools able to perform detection, localization, classification, compression and storage of signals automatically and organized manner are essential to ensure adequate monitoring process to electric power systems as a whole. Among the tools commonly applied to the tasks mentioned above, one can highlight the Wavelet Transform (WT) and Artificial Neural Networks (ANN). However, the WT has not been established yet and nor its level of decomposition, as well as the most appropriate ANN architecture and topology to the tasks already mentioned. The main fact that has led to this finding is due to the review of related literature, where it is possible to note the use of distinct WT and ANN. Therefore, the first contribution of this research was the design and development of an efficient method of segmentation of signals associate to Power Quality (PQ) disturbances. The developed method take advantage of WT properties of temporal identification of signal discontinuities. The second contribution is the development of an automatic algorithm that, through the segmentation method developed and classification by ANN, indicates the best tools (Wavelets and ANN) for the tasks of segmentation, extraction of characteristics and classification of QEE disturbances. This algorithm was developed based on the resources of the Evolutionary Algorithms and it adopts Multi-layered Perceptron type ANN, once this architecture can be considered consecrated with regard to the pattenrs classification. Finally, the third contribution is related to the development of EA based procedures in order to improve data compression methods that preserve the relevant information in the PQ signals. Thus, it is important to mention that the results of this research may determine automatic mechanisms to be used in the process of recording, processing and storing information that will be important in order to maintain an up-to-date (historical) database in the utilities, from which , indexes and a better mapping and understanding of all PQ related disturbances can be better understood and solved.

A presença de distúrbios na energia elétrica fornecida aos consumidores pode causar a diminuição no tempo de vida útil dos equipamentos, mal funcionamento ou até mesmo sua perda. Desse modo, ferramentas capazes de realizar a detecção, localização, classificação, compressão e o armazenamento de sinais de forma automática e organizada são essenciais para garantir um processo de monitoramento adequado ao sistema elétrico de potência como um todo. Dentre as ferramentas comumente aplicadas às tarefas supramencionadas, pode-se destacar a Transformada Wavelet (TW) e as Redes Neurais Artificiais (RNAs). Contudo, ainda não foi estabelecida uma metodologia para obtenção e validação da TW e seu nível de decomposição, bem como da arquitetura e da topologia de RNAs mais apropriadas às tarefas supracitadas. O principal fato que levou a esta constatação deve-se à análise da literatura correlata, onde é possível notar o uso de distintas TW e RNAs. Neste contexto, a primeira contribuição desta pesquisa foi o projeto e desenvolvimento de um método eficiente de segmentação de sinais com distúrbios associados à Qualidade da Energia Elétrica (QEE). O método desenvolvido se beneficia das propriedades da TW de identificação temporal de descontinuidades em sinais. A segunda contribuição é o desenvolvimento de um algoritmo automático que, por meio do método de segmentação desenvolvido e de classificação por RNAs, indique as melhores ferramentas (Wavelets e RNAs) para as tarefas de segmentação, extração de características e classificação de distúrbios de QEE. Esse algoritmo foi desenvolvido com base nos recursos dos Algoritmos Evolutivos (AEs) e adotou RNAs do tipo Perceptron Multicamadas, pois, esta arquitetura pode ser considerada consagrada no que se refere à classificação de padrões. Por fim, a terceira contribuição é relativa ao desenvolvimento de um procedimentos baseados em AEs, a fim de se aprimorar métodos de compressão de dados que preservem as informações relevantes nos sinais de QEE. Assim, é importante mencionar que os resultados dessa pesquisa poderão determinar mecanismos automáticos a serem utilizados no processo de registro, tratamento e armazenamento de informações que serão importantes para se manter um banco de dados (histórico) atualizado nas concessionárias de energia, a partir do qual, índices e um melhor mapeamento e entendimento de todos os distúrbios relacionados à QEE poderão ser melhor entendidos e solucionados. / The presence of disturbances in the electrical power supplied to consumers can decrease the lifetime of the equipment, cause malfunction or even their breakdown. Thus, tools able to perform detection, localization, classification, compression and storage of signals automatically and organized manner are essential to ensure adequate monitoring process to electric power systems as a whole. Among the tools commonly applied to the tasks mentioned above, one can highlight the Wavelet Transform (WT) and Artificial Neural Networks (ANN). However, the WT has not been established yet and nor its level of decomposition, as well as the most appropriate ANN architecture and topology to the tasks already mentioned. The main fact that has led to this finding is due to the review of related literature, where it is possible to note the use of distinct WT and ANN. Therefore, the first contribution of this research was the design and development of an efficient method of segmentation of signals associate to Power Quality (PQ) disturbances. The developed method take advantage of WT properties of temporal identification of signal discontinuities. The second contribution is the development of an automatic algorithm that, through the segmentation method developed and classification by ANN, indicates the best tools (Wavelets and ANN) for the tasks of segmentation, extraction of characteristics and classification of QEE disturbances. This algorithm was developed based on the resources of the Evolutionary Algorithms and it adopts Multi-layered Perceptron type ANN, once this architecture can be considered consecrated with regard to the pattenrs classification. Finally, the third contribution is related to the development of EA based procedures in order to improve data compression methods that preserve the relevant information in the PQ signals. Thus, it is important to mention that the results of this research may determine automatic mechanisms to be used in the process of recording, processing and storing information that will be important in order to maintain an up-to-date (historical) database in the utilities, from which , indexes and a better mapping and understanding of all PQ related disturbances can be better understood and solved.

Apresenta-se uma metodologia que tem como base a análise de campos magnéticos no monitoramento da qualidade da energia de sistemas elétricos. Em particular, são avaliados os aspectos referentes à detecção de faltas em sistemas elétricos. Diferente do processo de monitoração tradicional, cujos sensores precisam estar fisicamente conectados aos circuitos analisados, propõe-se estudar a viabilidade da utilização dos sinais provenientes da assinatura magnética resultante no exame do desempenho dos sistemas elétricos. Ressalta-se, assim, a característica não invasiva deste processo. Em uma primeira instância, simulações numéricas e medidas experimentais são usadas para estimar a validade deste método. Com base em valores das correntes de falta fase-terra, relacionados a configurações reais de sistemas de distribuição, provenientes de simulações numéricas e disponibilizadas na literatura, são calculados os campos magnéticos em regiões pré-selecionadas próximas às linhas. A seguir, aplicam-se os conceitos relacionados a wavelets no tratamento dos sinais resultantes. É nesta etapa que, por meio da decomposição da assinatura magnética correspondente, serão obtidos os dados necessários para se correlacionar os componentes dos sinais ao diagnóstico das faltas, nos sistemas elétricos. A Análise de múltirresolução é aplicada. Além destes resultados teóricos, aqueles provenientes de uma bancada experimental são examinados. Algumas configurações canônicas foram pré-selecionadas, visando estudar a eventual influência dos aspectos geométricos nos resultados relacionados à decomposição do sinal em análise. Embora métodos analíticos pudessem ser empregados na determinação da assinatura magnética resultante, os métodos numéricos, tais como o método dos elementos finitos, foram utilizados visando agilizar a obtenção de resultados teóricos a serem avaliados. Da mesma forma, aplicativos já disponibilizados comercialmente foram utilizados na decomposição dos sinais. Esta metodologia foi aplicada, também, para identificar faltas, aplicando-se a análise da variância para os diversos níveis do detalhe wavelet. A validação da metodologia foi feita pela comparação entre os resultados simulados e obtidos experimentalmente. / A methodology based on the analysis of magnetic fields for monitoring the quality of energy in electrical systems is presented herein. Aspects referring to fault detection in electrical systems in particular are evaluated. Contrary to the traditional monitoring process, in which sensors must be physically linked to the circuits under analysis, the results are presented from a feasibility study on the use of signals arising from the resulting magnetic signature by means of the electrical systems analysis. Thus the non-invasive characteristic of this process should be pointed out. First, numerical simulations and experimental measures were used to estimate the validity of this method. Based on values of the current of phase-earth fault related to actual features of the distribution systems and derived from numeric simulations found in the literature, the magnetic fields, in pre-established regions, were calculated. Following this, the concepts related to wavelets in the treatment of resulting signals were applied. It is in this phase that, by means of the decomposition of the corresponding magnetic signature, the data necessary to correlate the signal components for the diagnosis of faults in electrical systems were obtained. A Multiresolution Analysis (MRA) was applied. In addition to these theoretical results, the results from a laboratory workbench were also examined. Some canonical features were pre-selected, aiming to study the influence of geometric aspects on the results related to the signal decomposition analyzed. Although analytical methods could be employed to determine the resulting magnetic signature, numerical methods, such as the finite element method, were used to expedite obtaining the theoretical results to be analyzed. Likewise, commercial software was also used for the decomposition of signals. This methodology was validated by comparing the measured and simulated magnetic flux density. This methodology was also applied to identify and classify faults by means of the variance curve towards the wavelet detail.

This thesis presents the development of a Sensor Tower Platform (STP) comprised of an array of Passive Infra-Red (PIR) sensors along with a classification algorithm that enables the STP to distinguish between human intrusion, animal intrusion and clutter arising from wind-blown vegetative movement in an outdoor environment. The research was motivated by the aim of exploring the potential use of wireless sensor networks (WSNs) as an early-warning system to help mitigate human-wildlife conflicts occurring at the edge of a forest. While PIR sensors are in commonplace use in indoor settings, their use in an outdoor environment is hampered by the fact that they are prone to false alarms arising from wind-blown vegetation. Every PIR sensor is made up of one or more pairs of pyroelectric pixels arranged in a plane, and the orientation of interest in this thesis is one in which this plane is a vertical plane, i.e., a plane perpendicular to the ground plane. The intersection of the Field Of View (FOV) of the PIR sensor with a second vertical plane that lies within the FOV of the PIR sensor, is called the virtual pixel array (VPA). The structure of the VPA corresponding to the plane along which intruder motion takes place determines the form of the signal generated by the PIR sensor. The STP developed in this thesis employs an array of PIR sensors designed so as to result in a VPA that makes it easier to discriminate between human and animal intrusion while keeping to a small level false alarms arising from vegetative motion. The design was carried out in iterative fashion, with each successive iteration separated by a lengthy testing phase. There were a total of 5 design iterations spanning a total period of 14 months. Given the inherent challenges involved in gathering data corresponding to animal intrusion, the testing of the SP was carried out both using real-world data and through simulation. Simulation was carried out by developing a tool that employed animation software to simulate intruder and animal motion as well as some limited models of wind-blown vegetation. More specifically, the simulation tool employed 3-dimensional models of intruder and shrub motion that were developed using the popular animation software Blender. The simulated output signal of the PIR sensor was then generated by calculating the area of the 3-dimensional intruder when projected onto the VPA of the STP. An algorithm for efficiently calculating this to a good degree of approximation was implemented in Open Graphics Library (OpenGL). The simulation tool was useful both for evaluating various competing design alternatives as well as for developing an intuition for the kind of signals the SP would generate without the need for time-consuming and challenging animal-motion data collection. Real-world data corresponding to human motion was gathered on the campus of the Indian Institute of Science (IISc), while animal data was recorded at a dog-trainer facility in Kengeri as well as the Bannerghatta Biological Park, both located in the outskirts of Bengaluru. The array of PIR sensors was designed so as to result in a VPA that had good spatial resolution. The spatial resolution capabilities of the STP permitted distinguishing between human and animal motion with good accuracy based on low-complexity, signal-energy computations. Rejecting false alarms arising from vegetative movement proved to be more challenging. While the inherent spatial resolution of the STP was very helpful, an alternative approach turned out to have much higher accuracy, although it is computationally more intensive. Under this approach, the intruder signal, either human or animal, was modelled as a chirp waveform. When the intruder moves along a circular arc surrounding the STP, the resulting signal is periodic with constant frequency. However, when the intruder moves along a more likely straight-line path, the resultant signal has a strong chirp component. Clutter signals arising from vegetative motion does not exhibit this chirp behavior and an algorithm that exploited this difference turned in a classification accuracy in excess of 97%.

Die vorliegende Arbeit beschäftigt sich mit der Konzeptionierung und Entwicklung eines neuartigen Fahrerassistenzsystems für Nutzfahrzeuge, welches die lichte Höhe von vor dem Fahrzeug befindlichen Hindernissen berechnet und über einen Abgleich mit der einstellbaren Fahrzeughöhe die Passierbarkeit bestimmt. Dabei werden die von einer Monokamera aufgenommenen Bildsequenzen genutzt, um durch indirekte und direkte Rekonstruktionsverfahren ein 3D-Abbild der Fahrumgebung zu erschaffen. Unter Hinzunahme einer Radodometrie-basierten Eigenbewegungsschätzung wird die erstellte 3D-Repräsentation skaliert und eine Prädiktion der longitudinalen und lateralen Fahrzeugbewegung ermittelt. Basierend auf dem vertikalen Höhenplan der Straßenoberfläche, welcher über die Aneinanderreihung mehrerer Ebenen modelliert wird, erfolgt die Klassifizierung des 3D-Raums in Fahruntergrund, Struktur und potentielle Hindernisse. Die innerhalb des Fahrschlauchs liegenden Hindernisse werden hinsichtlich ihrer Entfernung und Höhe bewertet. Ein daraus abgeleitetes Warnkonzept dient der optisch-akustischen Signalisierung des Hindernisses im Kombiinstrument des Fahrzeugs. Erfolgt keine entsprechende Reaktion durch den Fahrer, so wird bei kritischen Hindernishöhen eine Notbremsung durchgeführt. Die geschätzte Eigenbewegung und berechneten Hindernisparameter werden mithilfe von Referenzsensorik bewertet. Dabei kommt eine dGPS-gestützte Inertialplattform sowie ein terrestrischer und mobiler Laserscanner zum Einsatz. Im Rahmen der Arbeit werden verschiedene Umgebungssituationen und Hindernistypen im urbanen und ländlichen Raum untersucht und Aussagen zur Genauigkeit und Zuverlässigkeit des Verfahrens getroffen. Ein wesentlicher Einflussfaktor auf die Dichte und Genauigkeit der 3D-Rekonstruktion ist eine gleichmäßige Umgebungsbeleuchtung innerhalb der Bildsequenzaufnahme. Es wird in diesem Zusammenhang zwingend auf den Einsatz einer Automotive-tauglichen Kamera verwiesen. Die durch die Radodometrie bestimmte Eigenbewegung eignet sich im langsamen Geschwindigkeitsbereich zur Skalierung des 3D-Punktraums. Dieser wiederum sollte durch eine Kombination aus indirektem und direktem Punktrekonstruktionsverfahren erstellt werden. Der indirekte Anteil stützt dabei die Initialisierung des Verfahrens zum Start der Funktion und ermöglicht eine robuste Kameraschätzung. Das direkte Verfahren ermöglicht die Rekonstruktion einer hohen Anzahl an 3D-Punkten auf den Hindernisumrissen, welche zumeist die Unterkante beinhalten. Die Unterkante kann in einer Entfernung bis zu 20 m detektiert und verfolgt werden. Der größte Einflussfaktor auf die Genauigkeit der Berechnung der lichten Höhe von Hindernissen ist die Modellierung des Fahruntergrunds. Zur Reduktion von Ausreißern in der Höhenberechnung eignet sich die Stabilisierung des Verfahrens durch die Nutzung von zeitlich vorher zur Verfügung stehenden Berechnungen. Als weitere Maßnahme zur Stabilisierung wird zudem empfohlen die Hindernisausgabe an den Fahrer und den automatischen Notbremsassistenten mittels einer Hysterese zu stützen. Das hier vorgestellte System eignet sich für Park- und Rangiervorgänge und ist als kostengünstiges Fahrerassistenzsystem interessant für Pkw mit Aufbauten und leichte Nutzfahrzeuge. / The present work deals with the conception and development of a novel advanced driver assistance system for commercial vehicles, which estimates the clearance height of obstacles in front of the vehicle and determines the passability by comparison with the adjustable vehicle height. The image sequences captured by a mono camera are used to create a 3D representation of the driving environment using indirect and direct reconstruction methods. The 3D representation is scaled and a prediction of the longitudinal and lateral movement of the vehicle is determined with the aid of a wheel odometry-based estimation of the vehicle's own movement. Based on the vertical elevation plan of the road surface, which is modelled by attaching several surfaces together, the 3D space is classified into driving surface, structure and potential obstacles. The obstacles within the predicted driving tube are evaluated with regard to their distance and height. A warning concept derived from this serves to visually and acoustically signal the obstacle in the vehicle's instrument cluster. If the driver does not respond accordingly, emergency braking will be applied at critical obstacle heights. The estimated vehicle movement and calculated obstacle parameters are evaluated with the aid of reference sensors. A dGPS-supported inertial measurement unit and a terrestrial as well as a mobile laser scanner are used. Within the scope of the work, different environmental situations and obstacle types in urban and rural areas are investigated and statements on the accuracy and reliability of the implemented function are made. A major factor influencing the density and accuracy of 3D reconstruction is uniform ambient lighting within the image sequence. In this context, the use of an automotive camera is mandatory. The inherent motion determined by wheel odometry is suitable for scaling the 3D point space in the slow speed range. The 3D representation however, should be created by a combination of indirect and direct point reconstruction methods. The indirect part supports the initialization phase of the function and enables a robust camera estimation. The direct method enables the reconstruction of a large number of 3D points on the obstacle outlines, which usually contain the lower edge. The lower edge can be detected and tracked up to 20 m away. The biggest factor influencing the accuracy of the calculation of the clearance height of obstacles is the modelling of the driving surface. To reduce outliers in the height calculation, the method can be stabilized by using calculations from older time steps. As a further stabilization measure, it is also recommended to support the obstacle output to the driver and the automatic emergency brake assistant by means of hysteresis. The system presented here is suitable for parking and maneuvering operations and is interesting as a cost-effective driver assistance system for cars with superstructures and light commercial vehicles.

<p>Deep Learning (DL)-based edge intelligence has garnered significant attention in recent years due to the rapid proliferation of the Internet of Things (IoT), embedded, and intelligent systems, collectively termed edge devices. Sensor data streams acquired by these edge devices are processed by a Deep Neural Network (DNN) application that runs on the device itself or in the cloud. However, the high computational complexity and energy consumption of processing DNNs often limit their deployment on these edge inference systems due to limited compute, memory and energy resources. Furthermore, high costs, strict application latency demands, data privacy, security constraints, and the absence of reliable edge-cloud network connectivity heavily impact edge application efficiency in the case of cloud-assisted DNN inference. Inevitably, performance and energy efficiency are of utmost importance in these edge inference systems, aside from the accuracy of the application. To facilitate energy- efficient edge inference systems running computationally complex DNNs, this dissertation makes three key contributions.</p> <p><br></p> <p>The first contribution adopts a full-system approach to Approximate Computing, a design paradigm that trades off a small degradation in application quality for significant energy savings. Within this context, we present the foundational concepts of AxIS, the first approximate edge inference system that jointly optimizes the constituent subsystems leading to substantial energy benefits compared to optimization of the individual subsystem. To illustrate the efficacy of this approach, we demonstrate multiple versions of an approximate smart camera system that executes various DNN-based unimodal computer vision applications, showcasing how the sensor, memory, compute, and communication subsystems can all be synergistically approximated for energy-efficient edge inference.</p> <p><br></p> <p>Building on this foundation, the second contribution extends AxIS to multimodal AI, harnessing data from multiple sensor modalities to impart human-like cognitive and perceptual abilities to edge devices. By exploring optimization techniques for multiple sensor modalities and subsystems, this research reveals the impact of synergistic modality-aware optimizations on system-level accuracy-efficiency (AE) trade-offs, culminating in the introduction of SysteMMX, the first AE scalable cognitive system that allows efficient multimodal inference at the edge. To illustrate the practicality and effectiveness of this approach, we present an in-depth case study centered around a multimodal system that leverages RGB and Depth sensor modalities for image segmentation tasks.</p> <p><br></p> <p>The final contribution focuses on optimizing the performance of an edge-cloud collaborative inference system through intelligent DNN partitioning and computation offloading. We delve into the realm of distributed inference across edge devices and cloud servers, unveiling the challenges associated with finding the optimal partitioning point in DNNs for significant inference latency speedup. To address these challenges, we introduce PArtNNer, a platform-agnostic and adaptive DNN partitioning framework capable of dynamically adapting to changes in communication bandwidth and cloud server load. Unlike existing approaches, PArtNNer does not require pre-characterization of underlying edge computing platforms, making it a versatile and efficient solution for real-world edge-cloud scenarios.</p> <p><br></p> <p>Overall, this thesis provides novel insights, innovative techniques, and intelligent solutions to enable energy-efficient AI at the edge. The contributions presented herein serve as a solid foundation for future researchers to build upon, driving innovation and shaping the trajectory of research in edge AI.</p>