Fine-grained Indoor Tracking by Fusing Inertial Sensor and Physical Layer Information in WLANs

Indoor positioning has become an emerging research area because of huge commercial demands for location-based services in indoor environments. Channel State Information (CSI) as a fine-grained physical layer information has been recently proposed to achieve high positioning accuracy by using range-based methods, e.g., trilateration. In this work, we propose to fuse the CSI-based ranges and velocity estimated from inertial sensors by an enhanced particle filter to achieve highly accurate tracking. The algorithm relies on some enhanced ranging methods and further mitigates the remaining ranging errors by a weighting technique. Additionally, we provide an efficient method to estimate the velocity based on inertial sensors. The algorithms are designed in a network-based system, which uses rather cheap commercial devices as anchor nodes. We evaluate our system in a complex environment along three different moving paths. Our proposed tracking method can achieve 1:3m for mean accuracy and 2:2m for 90% accuracy, which is more accurate and stable than pedestrian dead reckoning and range-based positioning.

Fine-grained indoor positioning and tracking systems

Indoor positioning has attracted considerable attention for decades due to the increasing demands for location based services. In the past years, although numerous methods have been proposed for indoor positioning, it is still challenging to find a convincing solution that combines high positioning accuracy and ease of deployment. Radio-based indoor positioning has emerged as a dominant method due to its ubiquitousness, especially for WiFi. RSSI (Received Signal Strength Indicator) has been investigated in the area of indoor positioning for decades. However, it is prone to multipath propagation and hence fingerprinting has become the most commonly used method for indoor positioning using RSSI. The drawback of fingerprinting is that it requires intensive labour efforts to calibrate the radio map prior to experiments, which makes the deployment of the positioning system very time consuming. Using time information as another way for radio-based indoor positioning is challenged by time synchronization among anchor nodes and timestamp accuracy. Besides radio-based positioning methods, intensive research has been conducted to make use of inertial sensors for indoor tracking due to the fast developments of smartphones. However, these methods are normally prone to accumulative errors and might not be available for some applications, such as passive positioning. This thesis focuses on network-based indoor positioning and tracking systems, mainly for passive positioning, which does not require the participation of targets in the positioning process. To achieve high positioning accuracy, we work on some information of radio signals from physical-layer processing, such as timestamps and channel information. The contributions in this thesis can be divided into two parts: time-based positioning and channel information based positioning. First, for time-based indoor positioning (especially for narrow-band signals), we address challenges for compensating synchronization offsets among anchor nodes, designing timestamps with high resolution, and developing accurate positioning methods. Second, we work on range-based positioning methods with channel information to passively locate and track WiFi targets. Targeting less efforts for deployment, we work on range-based methods, which require much less calibration efforts than fingerprinting. By designing some novel enhanced methods for both ranging and positioning (including trilateration for stationary targets and particle filter for mobile targets), we are able to locate WiFi targets with high accuracy solely relying on radio signals and our proposed enhanced particle filter significantly outperforms the other commonly used range-based positioning algorithms, e.g., a traditional particle filter, extended Kalman filter and trilateration algorithms. In addition to using radio signals for passive positioning, we propose a second enhanced particle filter for active positioning to fuse inertial sensor and channel information to track indoor targets, which achieves higher tracking accuracy than tracking methods solely relying on either radio signals or inertial sensors.

Fine-grained Indoor Tracking by Fusing Inertial Sensor and Physical Layer Information in WLANs

Indoor positioning has become an emerging research area because of huge commercial demands for location-based services in indoor environments. Channel State Information (CSI) as fine-grained physical layer information has been recently proposed to achieve high positioning accuracy by using range based methods, e.g., trilateration. In this work, we propose to fuse the CSI-based ranging and velocity estimated from inertial sensors by an enhanced particle filter to achieve highly accurate tracking. The algorithm relies on some enhanced ranging methods and further mitigates the remaining ranging errors by a weighting technique. Additionally, we provide an efficient method to estimate the velocity based on inertial sensors. The algorithms are designed in a network-based system, which uses rather cheap commercial devices as anchor nodes. We evaluate our system in a complex environment along three different moving paths. Our proposed tracking method can achieve 1.3m for mean accuracy and 2.2m for 90% accuracy, which is more accurate and stable than pedestrian dead reckoning and range-based positioning.

Video analysis based vehicle detection and tracking using an MCMC sampling framework

This article presents a probabilistic method for vehicle detection and tracking through the analysis of monocular images obtained from a vehicle-mounted camera. The method is designed to address the main shortcomings of traditional particle filtering approaches, namely Bayesian methods based on importance sampling, for use in traffic environments. These methods do not scale well when the dimensionality of the feature space grows, which creates significant limitations when tracking multiple objects. Alternatively, the proposed method is based on a Markov chain Monte Carlo (MCMC) approach, which allows efficient sampling of the feature space. The method involves important contributions in both the motion and the observation models of the tracker. Indeed, as opposed to particle filter-based tracking methods in the literature, which typically resort to observation models based on appearance or template matching, in this study a likelihood model that combines appearance analysis with information from motion parallax is introduced. Regarding the motion model, a new interaction treatment is defined based on Markov random fields (MRF) that allows for the handling of possible inter-dependencies in vehicle trajectories. As for vehicle detection, the method relies on a supervised classification stage using support vector machines (SVM). The contribution in this field is twofold. First, a new descriptor based on the analysis of gradient orientations in concentric rectangles is dened. This descriptor involves a much smaller feature space compared to traditional descriptors, which are too costly for real-time applications. Second, a new vehicle image database is generated to train the SVM and made public. The proposed vehicle detection and tracking method is proven to outperform existing methods and to successfully handle challenging situations in the test sequences.

Segmentación de Secuencias de Vídeo: análisis y desarrollo de estrategias para la detección de cambios de toma y para la detección de objetos móviles

The important technological advances experienced along the last years have resulted in an important demand for new and efficient computer vision applications. On the one hand, the increasing use of video editing software has given rise to a necessity for faster and more efficient editing tools that, in a first step, perform a temporal segmentation in shots. On the other hand, the number of electronic devices with integrated cameras has grown enormously. These devices require new, fast, and efficient computer vision applications that include moving object detection strategies. In this dissertation, we propose a temporal segmentation strategy and several moving object detection strategies, which are suitable for the last generation of computer vision applications requiring both low computational cost and high quality results. First, a novel real-time high-quality shot detection strategy is proposed. While abrupt transitions are detected through a very fast pixel-based analysis, gradual transitions are obtained from an efficient edge-based analysis. Both analyses are reinforced with a motion analysis that allows to detect and discard false detections. This analysis is carried out exclusively over a reduced amount of candidate transitions, thus maintaining the computational requirements. On the other hand, a moving object detection strategy, which is based on the popular Mixture of Gaussians method, is proposed. This strategy, taking into account the recent history of each image pixel, adapts dynamically the amount of Gaussians that are required to model its variations. As a result, we improve significantly the computational efficiency with respect to other similar methods and, additionally, we reduce the influence of the used parameters in the results. Alternatively, in order to improve the quality of the results in complex scenarios containing dynamic backgrounds, we propose different non-parametric based moving object detection strategies that model both background and foreground. To obtain high quality results regardless of the characteristics of the analyzed sequence we dynamically estimate the most adequate bandwidth matrices for the kernels that are used in the background and foreground modeling. Moreover, the application of a particle filter allows to update the spatial information and provides a priori knowledge about the areas to analyze in the following images, enabling an important reduction in the computational requirements and improving the segmentation results. Additionally, we propose the use of an innovative combination of chromaticity and gradients that allows to reduce the influence of shadows and reflects in the detections.

Moving Object Detection for Real-Time High-Quality Lightweight Applications on Smart Cameras

Here, a novel and efficient strategy for moving object detection by non-parametric modeling on smart cameras is presented. Whereas the background is modeled using only color information, the foreground model combines color and spatial information. The application of a particle filter allows the update of the spatial information and provides a priori information about the areas to analyze in the following images, enabling an important reduction in the computational requirements and improving the segmentation results

Applications for wireless sensor networks : tracking with binary proximity sensors

El interés cada vez mayor por las redes de sensores inalámbricos pueden ser entendido simplemente pensando en lo que esencialmente son: un gran número de pequeños nodos sensores autoalimentados que recogen información o detectan eventos especiales y se comunican de manera inalámbrica, con el objetivo final de entregar sus datos procesados a una estación base. Los nodos sensores están densamente desplegados dentro del área de interés, se pueden desplegar al azar y tienen capacidad de cooperación. Por lo general, estos dispositivos son pequeños y de bajo costo, de modo que pueden ser producidos y desplegados en gran numero aunque sus recursos en términos de energía, memoria, velocidad de cálculo y ancho de banda están enormemente limitados. Detección, tratamiento y comunicación son tres elementos clave cuya combinación en un pequeño dispositivo permite lograr un gran número de aplicaciones. Las redes de sensores proporcionan oportunidades sin fin, pero al mismo tiempo plantean retos formidables, tales como lograr el máximo rendimiento de una energía que es escasa y por lo general un recurso no renovable. Sin embargo, los recientes avances en la integración a gran escala, integrado de hardware de computación, comunicaciones, y en general, la convergencia de la informática y las comunicaciones, están haciendo de esta tecnología emergente una realidad. Del mismo modo, los avances en la nanotecnología están empezando a hacer que todo gire entorno a las redes de pequeños sensores y actuadores distribuidos. Hay diferentes tipos de sensores tales como sensores de presión, acelerómetros, cámaras, sensores térmicos o un simple micrófono. Supervisan las condiciones presentes en diferentes lugares tales como la temperatura, humedad, el movimiento, la luminosidad, presión, composición del suelo, los niveles de ruido, la presencia o ausencia de ciertos tipos de objetos, los niveles de tensión mecánica sobre objetos adheridos y las características momentáneas tales como la velocidad , la dirección y el tamaño de un objeto, etc. Se comprobara el estado de las Redes Inalámbricas de Sensores y se revisaran los protocolos más famosos. Así mismo, se examinara la identificación por radiofrecuencia (RFID) ya que se está convirtiendo en algo actual y su presencia importante. La RFID tiene un papel crucial que desempeñar en el futuro en el mundo de los negocios y los individuos por igual. El impacto mundial que ha tenido la identificación sin cables está ejerciendo fuertes presiones en la tecnología RFID, los servicios de investigación y desarrollo, desarrollo de normas, el cumplimiento de la seguridad y la privacidad y muchos más. Su potencial económico se ha demostrado en algunos países mientras que otros están simplemente en etapas de planificación o en etapas piloto, pero aun tiene que afianzarse o desarrollarse a través de la modernización de los modelos de negocio y aplicaciones para poder tener un mayor impacto en la sociedad. Las posibles aplicaciones de redes de sensores son de interés para la mayoría de campos. La monitorización ambiental, la guerra, la educación infantil, la vigilancia, la micro-cirugía y la agricultura son solo unos pocos ejemplos de los muchísimos campos en los que tienen cabida las redes mencionadas anteriormente. Estados Unidos de América es probablemente el país que más ha investigado en esta área por lo que veremos muchas soluciones propuestas provenientes de ese país. Universidades como Berkeley, UCLA (Universidad de California, Los Ángeles) Harvard y empresas como Intel lideran dichas investigaciones. Pero no solo EE.UU. usa e investiga las redes de sensores inalámbricos. La Universidad de Southampton, por ejemplo, está desarrollando una tecnología para monitorear el comportamiento de los glaciares mediante redes de sensores que contribuyen a la investigación fundamental en glaciología y de las redes de sensores inalámbricos. Así mismo, Coalesenses GmbH (Alemania) y Zurich ETH están trabajando en diversas aplicaciones para redes de sensores inalámbricos en numerosas áreas. Una solución española será la elegida para ser examinada más a fondo por ser innovadora, adaptable y polivalente. Este estudio del sensor se ha centrado principalmente en aplicaciones de tráfico, pero no se puede olvidar la lista de más de 50 aplicaciones diferentes que ha sido publicada por la firma creadora de este sensor específico. En la actualidad hay muchas tecnologías de vigilancia de vehículos, incluidos los sensores de bucle, cámaras de video, sensores de imagen, sensores infrarrojos, radares de microondas, GPS, etc. El rendimiento es aceptable, pero no suficiente, debido a su limitada cobertura y caros costos de implementación y mantenimiento, especialmente este ultimo. Tienen defectos tales como: línea de visión, baja exactitud, dependen mucho del ambiente y del clima, no se puede realizar trabajos de mantenimiento sin interrumpir las mediciones, la noche puede condicionar muchos de ellos, tienen altos costos de instalación y mantenimiento, etc. Por consiguiente, en las aplicaciones reales de circulación, los datos recibidos son insuficientes o malos en términos de tiempo real debido al escaso número de detectores y su costo. Con el aumento de vehículos en las redes viales urbanas las tecnologías de detección de vehículos se enfrentan a nuevas exigencias. Las redes de sensores inalámbricos son actualmente una de las tecnologías más avanzadas y una revolución en la detección de información remota y en las aplicaciones de recogida. Las perspectivas de aplicación en el sistema inteligente de transporte son muy amplias. Con este fin se ha desarrollado un programa de localización de objetivos y recuento utilizando una red de sensores binarios. Esto permite que el sensor necesite mucha menos energía durante la transmisión de información y que los dispositivos sean más independientes con el fin de tener un mejor control de tráfico. La aplicación se centra en la eficacia de la colaboración de los sensores en el seguimiento más que en los protocolos de comunicación utilizados por los nodos sensores. Las operaciones de salida y retorno en las vacaciones son un buen ejemplo de por qué es necesario llevar la cuenta de los coches en las carreteras. Para ello se ha desarrollado una simulación en Matlab con el objetivo localizar objetivos y contarlos con una red de sensores binarios. Dicho programa se podría implementar en el sensor que Libelium, la empresa creadora del sensor que se examinara concienzudamente, ha desarrollado. Esto permitiría que el aparato necesitase mucha menos energía durante la transmisión de información y los dispositivos sean más independientes. Los prometedores resultados obtenidos indican que los sensores de proximidad binarios pueden formar la base de una arquitectura robusta para la vigilancia de áreas amplias y para el seguimiento de objetivos. Cuando el movimiento de dichos objetivos es suficientemente suave, no tiene cambios bruscos de trayectoria, el algoritmo ClusterTrack proporciona un rendimiento excelente en términos de identificación y seguimiento de trayectorias los objetos designados como blancos. Este algoritmo podría, por supuesto, ser utilizado para numerosas aplicaciones y se podría seguir esta línea de trabajo para futuras investigaciones. No es sorprendente que las redes de sensores de binarios de proximidad hayan atraído mucha atención últimamente ya que, a pesar de la información mínima de un sensor de proximidad binario proporciona, las redes de este tipo pueden realizar un seguimiento de todo tipo de objetivos con la precisión suficiente. Abstract The increasing interest in wireless sensor networks can be promptly understood simply by thinking about what they essentially are: a large number of small sensing self-powered nodes which gather information or detect special events and communicate in a wireless fashion, with the end goal of handing their processed data to a base station. The sensor nodes are densely deployed inside the phenomenon, they deploy random and have cooperative capabilities. Usually these devices are small and inexpensive, so that they can be produced and deployed in large numbers, and so their resources in terms of energy, memory, computational speed and bandwidth are severely constrained. Sensing, processing and communication are three key elements whose combination in one tiny device gives rise to a vast number of applications. Sensor networks provide endless opportunities, but at the same time pose formidable challenges, such as the fact that energy is a scarce and usually non-renewable resource. However, recent advances in low power Very Large Scale Integration, embedded computing, communication hardware, and in general, the convergence of computing and communications, are making this emerging technology a reality. Likewise, advances in nanotechnology and Micro Electro-Mechanical Systems are pushing toward networks of tiny distributed sensors and actuators. There are different sensors such as pressure, accelerometer, camera, thermal, and microphone. They monitor conditions at different locations, such as temperature, humidity, vehicular movement, lightning condition, pressure, soil makeup, noise levels, the presence or absence of certain kinds of objects, mechanical stress levels on attached objects, the current characteristics such as speed, direction and size of an object, etc. The state of Wireless Sensor Networks will be checked and the most famous protocols reviewed. As Radio Frequency Identification (RFID) is becoming extremely present and important nowadays, it will be examined as well. RFID has a crucial role to play in business and for individuals alike going forward. The impact of ‘wireless’ identification is exerting strong pressures in RFID technology and services research and development, standards development, security compliance and privacy, and many more. The economic value is proven in some countries while others are just on the verge of planning or in pilot stages, but the wider spread of usage has yet to take hold or unfold through the modernisation of business models and applications. Possible applications of sensor networks are of interest to the most diverse fields. Environmental monitoring, warfare, child education, surveillance, micro-surgery, and agriculture are only a few examples. Some real hardware applications in the United States of America will be checked as it is probably the country that has investigated most in this area. Universities like Berkeley, UCLA (University of California, Los Angeles) Harvard and enterprises such as Intel are leading those investigations. But not just USA has been using and investigating wireless sensor networks. University of Southampton e.g. is to develop technology to monitor glacier behaviour using sensor networks contributing to fundamental research in glaciology and wireless sensor networks. Coalesenses GmbH (Germany) and ETH Zurich are working in applying wireless sensor networks in many different areas too. A Spanish solution will be the one examined more thoroughly for being innovative, adaptable and multipurpose. This study of the sensor has been focused mainly to traffic applications but it cannot be forgotten the more than 50 different application compilation that has been published by this specific sensor’s firm. Currently there are many vehicle surveillance technologies including loop sensors, video cameras, image sensors, infrared sensors, microwave radar, GPS, etc. The performance is acceptable but not sufficient because of their limited coverage and expensive costs of implementation and maintenance, specially the last one. They have defects such as: line-ofsight, low exactness, depending on environment and weather, cannot perform no-stop work whether daytime or night, high costs for installation and maintenance, etc. Consequently, in actual traffic applications the received data is insufficient or bad in terms of real-time owed to detector quantity and cost. With the increase of vehicle in urban road networks, the vehicle detection technologies are confronted with new requirements. Wireless sensor network is the state of the art technology and a revolution in remote information sensing and collection applications. It has broad prospect of application in intelligent transportation system. An application for target tracking and counting using a network of binary sensors has been developed. This would allow the appliance to spend much less energy when transmitting information and to make more independent devices in order to have a better traffic control. The application is focused on the efficacy of collaborative tracking rather than on the communication protocols used by the sensor nodes. Holiday crowds are a good case in which it is necessary to keep count of the cars on the roads. To this end a Matlab simulation has been produced for target tracking and counting using a network of binary sensors that e.g. could be implemented in Libelium’s solution. Libelium is the enterprise that has developed the sensor that will be deeply examined. This would allow the appliance to spend much less energy when transmitting information and to make more independent devices. The promising results obtained indicate that binary proximity sensors can form the basis for a robust architecture for wide area surveillance and tracking. When the target paths are smooth enough ClusterTrack particle filter algorithm gives excellent performance in terms of identifying and tracking different target trajectories. This algorithm could, of course, be used for different applications and that could be done in future researches. It is not surprising that binary proximity sensor networks have attracted a lot of attention lately. Despite the minimal information a binary proximity sensor provides, networks of these sensing modalities can track all kinds of different targets classes accurate enough.

Versatile Bayesian classifier for moving object detection by non-parametric background-foreground modeling

Along the recent years, several moving object detection strategies by non-parametric background-foreground modeling have been proposed. To combine both models and to obtain the probability of a pixel to belong to the foreground, these strategies make use of Bayesian classifiers. However, these classifiers do not allow to take advantage of additional prior information at different pixels. So, we propose a novel and efficient alternative Bayesian classifier that is suitable for this kind of strategies and that allows the use of whatever prior information. Additionally, we present an effective method to dynamically estimate prior probability from the result of a particle filter-based tracking strategy.

Fast human pose estimation using 3D Zernike descriptors

Markerless video-based human pose estimation algorithms face a high-dimensional problem that is frequently broken down into several lower-dimensional ones by estimating the pose of each limb separately. However, in order to do so they need to reliably locate the torso, for which they typically rely on time coherence and tracking algorithms. Their losing track usually results in catastrophic failure of the process, requiring human intervention and thus precluding their usage in real-time applications. We propose a very fast rough pose estimation scheme based on global shape descriptors built on 3D Zernike moments. Using an articulated model that we configure in many poses, a large database of descriptor/pose pairs can be computed off-line. Thus, the only steps that must be done on-line are the extraction of the descriptors for each input volume and a search against the database to get the most likely poses. While the result of such process is not a fine pose estimation, it can be useful to help more sophisticated algorithms to regain track or make more educated guesses when creating new particles in particle-filter-based tracking schemes. We have achieved a performance of about ten fps on a single computer using a database of about one million entries.

Bayesian scene analysis for multi-camera 3D tracking and camera positioning

Los sistemas de seguimiento mono-cámara han demostrado su notable capacidad para el análisis de trajectorias de objectos móviles y para monitorización de escenas de interés; sin embargo, tanto su robustez como sus posibilidades en cuanto a comprensión semántica de la escena están fuertemente limitadas por su naturaleza local y monocular, lo que los hace insuficientes para aplicaciones realistas de videovigilancia. El objetivo de esta tesis es la extensión de las posibilidades de los sistemas de seguimiento de objetos móviles para lograr un mayor grado de robustez y comprensión de la escena. La extensión propuesta se divide en dos direcciones separadas. La primera puede considerarse local, ya que está orientada a la mejora y enriquecimiento de las posiciones estimadas para los objetos móviles observados directamente por las cámaras del sistema; dicha extensión se logra mediante el desarrollo de un sistema multi-cámara de seguimiento 3D, capaz de proporcionar consistentemente las posiciones 3D de múltiples objetos a partir de las observaciones capturadas por un conjunto de sensores calibrados y con campos de visión solapados. La segunda extensión puede considerarse global, dado que su objetivo consiste en proporcionar un contexto global para relacionar las observaciones locales realizadas por una cámara con una escena de mucho mayor tamaño; para ello se propone un sistema automático de localización de cámaras basado en las trayectorias observadas de varios objetos móviles y en un mapa esquemático de la escena global monitorizada. Ambas líneas de investigación se tratan utilizando, como marco común, técnicas de estimación bayesiana: esta elección está justificada por la versatilidad y flexibilidad proporcionada por dicho marco estadístico, que permite la combinación natural de múltiples fuentes de información sobre los parámetros a estimar, así como un tratamiento riguroso de la incertidumbre asociada a las mismas mediante la inclusión de modelos de observación específicamente diseñados. Además, el marco seleccionado abre grandes posibilidades operacionales, puesto que permite la creación de diferentes métodos numéricos adaptados a las necesidades y características específicas de distintos problemas tratados. El sistema de seguimiento 3D con múltiples cámaras propuesto está específicamente diseñado para permitir descripciones esquemáticas de las medidas realizadas individualmente por cada una de las cámaras del sistema: esta elección de diseño, por tanto, no asume ningún algoritmo específico de detección o seguimiento 2D en ninguno de los sensores de la red, y hace que el sistema propuesto sea aplicable a redes reales de vigilancia con capacidades limitadas tanto en términos de procesamiento como de transmision. La combinación robusta de las observaciones capturadas individualmente por las cámaras, ruidosas, incompletas y probablemente contaminadas por falsas detecciones, se basa en un metodo de asociación bayesiana basado en geometría y color: los resultados de dicha asociación permiten el seguimiento 3D de los objetos de la escena mediante el uso de un filtro de partículas. El sistema de fusión de observaciones propuesto tiene, como principales características, una gran precisión en términos de localización 3D de objetos, y una destacable capacidad de recuperación tras eventuales errores debidos a un número insuficiente de datos de entrada. El sistema automático de localización de cámaras se basa en la observación de múltiples objetos móviles y un mapa esquemático de las áreas transitables del entorno monitorizado para inferir la posición absoluta de dicho sensor. Para este propósito, se propone un novedoso marco bayesiano que combina modelos dinámicos inducidos por el mapa en los objetos móviles presentes en la escena con las trayectorias observadas por la cámara, lo que representa un enfoque nunca utilizado en la literatura existente. El sistema de localización se divide en dos sub-tareas diferenciadas, debido a que cada una de estas tareas requiere del diseño de algoritmos específicos de muestreo para explotar en profundidad las características del marco desarrollado: por un lado, análisis de la ambigüedad del caso específicamente tratado y estimación aproximada de la localización de la cámara, y por otro, refinado de la localización de la cámara. El sistema completo, diseñado y probado para el caso específico de localización de cámaras en entornos de tráfico urbano, podría tener aplicación también en otros entornos y sensores de diferentes modalidades tras ciertas adaptaciones. ABSTRACT Mono-camera tracking systems have proved their capabilities for moving object trajectory analysis and scene monitoring, but their robustness and semantic possibilities are strongly limited by their local and monocular nature and are often insufficient for realistic surveillance applications. This thesis is aimed at extending the possibilities of moving object tracking systems to a higher level of scene understanding. The proposed extension comprises two separate directions. The first one is local, since is aimed at enriching the inferred positions of the moving objects within the area of the monitored scene directly covered by the cameras of the system; this task is achieved through the development of a multi-camera system for robust 3D tracking, able to provide 3D tracking information of multiple simultaneous moving objects from the observations reported by a set of calibrated cameras with semi-overlapping fields of view. The second extension is global, as is aimed at providing local observations performed within the field of view of one camera with a global context relating them to a much larger scene; to this end, an automatic camera positioning system relying only on observed object trajectories and a scene map is designed. The two lines of research in this thesis are addressed using Bayesian estimation as a general unifying framework. Its suitability for these two applications is justified by the flexibility and versatility of that stochastic framework, which allows the combination of multiple sources of information about the parameters to estimate in a natural and elegant way, addressing at the same time the uncertainty associated to those sources through the inclusion of models designed to this end. In addition, it opens multiple possibilities for the creation of different numerical methods for achieving satisfactory and efficient practical solutions to each addressed application. The proposed multi-camera 3D tracking method is specifically designed to work on schematic descriptions of the observations performed by each camera of the system: this choice allows the use of unspecific off-the-shelf 2D detection and/or tracking subsystems running independently at each sensor, and makes the proposal suitable for real surveillance networks with moderate computational and transmission capabilities. The robust combination of such noisy, incomplete and possibly unreliable schematic descriptors relies on a Bayesian association method, based on geometry and color, whose results allow the tracking of the targets in the scene with a particle filter. The main features exhibited by the proposal are, first, a remarkable accuracy in terms of target 3D positioning, and second, a great recovery ability after tracking losses due to insufficient input data. The proposed system for visual-based camera self-positioning uses the observations of moving objects and a schematic map of the passable areas of the environment to infer the absolute sensor position. To this end, a new Bayesian framework combining trajectory observations and map-induced dynamic models for moving objects is designed, which represents an approach to camera positioning never addressed before in the literature. This task is divided into two different sub-tasks, setting ambiguity analysis and approximate position estimation, on the one hand, and position refining, on the other, since they require the design of specific sampling algorithms to correctly exploit the discriminative features of the developed framework. This system, designed for camera positioning and demonstrated in urban traffic environments, can also be applied to different environments and sensors of other modalities after certain required adaptations.

Bayesian scene analysis for multi-camera 3D tracking and camera positioning

Los sistemas de seguimiento mono-cámara han demostrado su notable capacidad para el análisis de trajectorias de objectos móviles y para monitorización de escenas de interés; sin embargo, tanto su robustez como sus posibilidades en cuanto a comprensión semántica de la escena están fuertemente limitadas por su naturaleza local y monocular, lo que los hace insuficientes para aplicaciones realistas de videovigilancia. El objetivo de esta tesis es la extensión de las posibilidades de los sistemas de seguimiento de objetos móviles para lograr un mayor grado de robustez y comprensión de la escena. La extensión propuesta se divide en dos direcciones separadas. La primera puede considerarse local, ya que está orientada a la mejora y enriquecimiento de las posiciones estimadas para los objetos móviles observados directamente por las cámaras del sistema; dicha extensión se logra mediante el desarrollo de un sistema multi-cámara de seguimiento 3D, capaz de proporcionar consistentemente las posiciones 3D de múltiples objetos a partir de las observaciones capturadas por un conjunto de sensores calibrados y con campos de visión solapados. La segunda extensión puede considerarse global, dado que su objetivo consiste en proporcionar un contexto global para relacionar las observaciones locales realizadas por una cámara con una escena de mucho mayor tamaño; para ello se propone un sistema automático de localización de cámaras basado en las trayectorias observadas de varios objetos móviles y en un mapa esquemático de la escena global monitorizada. Ambas líneas de investigación se tratan utilizando, como marco común, técnicas de estimación bayesiana: esta elección está justificada por la versatilidad y flexibilidad proporcionada por dicho marco estadístico, que permite la combinación natural de múltiples fuentes de información sobre los parámetros a estimar, así como un tratamiento riguroso de la incertidumbre asociada a las mismas mediante la inclusión de modelos de observación específicamente diseñados. Además, el marco seleccionado abre grandes posibilidades operacionales, puesto que permite la creación de diferentes métodos numéricos adaptados a las necesidades y características específicas de distintos problemas tratados. El sistema de seguimiento 3D con múltiples cámaras propuesto está específicamente diseñado para permitir descripciones esquemáticas de las medidas realizadas individualmente por cada una de las cámaras del sistema: esta elección de diseño, por tanto, no asume ningún algoritmo específico de detección o seguimiento 2D en ninguno de los sensores de la red, y hace que el sistema propuesto sea aplicable a redes reales de vigilancia con capacidades limitadas tanto en términos de procesamiento como de transmision. La combinación robusta de las observaciones capturadas individualmente por las cámaras, ruidosas, incompletas y probablemente contaminadas por falsas detecciones, se basa en un metodo de asociación bayesiana basado en geometría y color: los resultados de dicha asociación permiten el seguimiento 3D de los objetos de la escena mediante el uso de un filtro de partículas. El sistema de fusión de observaciones propuesto tiene, como principales características, una gran precisión en términos de localización 3D de objetos, y una destacable capacidad de recuperación tras eventuales errores debidos a un número insuficiente de datos de entrada. El sistema automático de localización de cámaras se basa en la observación de múltiples objetos móviles y un mapa esquemático de las áreas transitables del entorno monitorizado para inferir la posición absoluta de dicho sensor. Para este propósito, se propone un novedoso marco bayesiano que combina modelos dinámicos inducidos por el mapa en los objetos móviles presentes en la escena con las trayectorias observadas por la cámara, lo que representa un enfoque nunca utilizado en la literatura existente. El sistema de localización se divide en dos sub-tareas diferenciadas, debido a que cada una de estas tareas requiere del diseño de algoritmos específicos de muestreo para explotar en profundidad las características del marco desarrollado: por un lado, análisis de la ambigüedad del caso específicamente tratado y estimación aproximada de la localización de la cámara, y por otro, refinado de la localización de la cámara. El sistema completo, diseñado y probado para el caso específico de localización de cámaras en entornos de tráfico urbano, podría tener aplicación también en otros entornos y sensores de diferentes modalidades tras ciertas adaptaciones. ABSTRACT Mono-camera tracking systems have proved their capabilities for moving object trajectory analysis and scene monitoring, but their robustness and semantic possibilities are strongly limited by their local and monocular nature and are often insufficient for realistic surveillance applications. This thesis is aimed at extending the possibilities of moving object tracking systems to a higher level of scene understanding. The proposed extension comprises two separate directions. The first one is local, since is aimed at enriching the inferred positions of the moving objects within the area of the monitored scene directly covered by the cameras of the system; this task is achieved through the development of a multi-camera system for robust 3D tracking, able to provide 3D tracking information of multiple simultaneous moving objects from the observations reported by a set of calibrated cameras with semi-overlapping fields of view. The second extension is global, as is aimed at providing local observations performed within the field of view of one camera with a global context relating them to a much larger scene; to this end, an automatic camera positioning system relying only on observed object trajectories and a scene map is designed. The two lines of research in this thesis are addressed using Bayesian estimation as a general unifying framework. Its suitability for these two applications is justified by the flexibility and versatility of that stochastic framework, which allows the combination of multiple sources of information about the parameters to estimate in a natural and elegant way, addressing at the same time the uncertainty associated to those sources through the inclusion of models designed to this end. In addition, it opens multiple possibilities for the creation of different numerical methods for achieving satisfactory and efficient practical solutions to each addressed application. The proposed multi-camera 3D tracking method is specifically designed to work on schematic descriptions of the observations performed by each camera of the system: this choice allows the use of unspecific off-the-shelf 2D detection and/or tracking subsystems running independently at each sensor, and makes the proposal suitable for real surveillance networks with moderate computational and transmission capabilities. The robust combination of such noisy, incomplete and possibly unreliable schematic descriptors relies on a Bayesian association method, based on geometry and color, whose results allow the tracking of the targets in the scene with a particle filter. The main features exhibited by the proposal are, first, a remarkable accuracy in terms of target 3D positioning, and second, a great recovery ability after tracking losses due to insufficient input data. The proposed system for visual-based camera self-positioning uses the observations of moving objects and a schematic map of the passable areas of the environment to infer the absolute sensor position. To this end, a new Bayesian framework combining trajectory observations and map-induced dynamic models for moving objects is designed, which represents an approach to camera positioning never addressed before in the literature. This task is divided into two different sub-tasks, setting ambiguity analysis and approximate position estimation, on the one hand, and position refining, on the other, since they require the design of specific sampling algorithms to correctly exploit the discriminative features of the developed framework. This system, designed for camera positioning and demonstrated in urban traffic environments, can also be applied to different environments and sensors of other modalities after certain required adaptations.

Parallel Computer Vision Algorithms for Graphics Processing Units

La evolución de los teléfonos móviles inteligentes, dotados de cámaras digitales, está provocando una creciente demanda de aplicaciones cada vez más complejas que necesitan algoritmos de visión artificial en tiempo real; puesto que el tamaño de las señales de vídeo no hace sino aumentar y en cambio el rendimiento de los procesadores de un solo núcleo se ha estancado, los nuevos algoritmos que se diseñen para visión artificial han de ser paralelos para poder ejecutarse en múltiples procesadores y ser computacionalmente escalables. Una de las clases de procesadores más interesantes en la actualidad se encuentra en las tarjetas gráficas (GPU), que son dispositivos que ofrecen un alto grado de paralelismo, un excelente rendimiento numérico y una creciente versatilidad, lo que los hace interesantes para llevar a cabo computación científica. En esta tesis se exploran dos aplicaciones de visión artificial que revisten una gran complejidad computacional y no pueden ser ejecutadas en tiempo real empleando procesadores tradicionales. En cambio, como se demuestra en esta tesis, la paralelización de las distintas subtareas y su implementación sobre una GPU arrojan los resultados deseados de ejecución con tasas de refresco interactivas. Asimismo, se propone una técnica para la evaluación rápida de funciones de complejidad arbitraria especialmente indicada para su uso en una GPU. En primer lugar se estudia la aplicación de técnicas de síntesis de imágenes virtuales a partir de únicamente dos cámaras lejanas y no paralelas—en contraste con la configuración habitual en TV 3D de cámaras cercanas y paralelas—con información de color y profundidad. Empleando filtros de mediana modificados para la elaboración de un mapa de profundidad virtual y proyecciones inversas, se comprueba que estas técnicas son adecuadas para una libre elección del punto de vista. Además, se demuestra que la codificación de la información de profundidad con respecto a un sistema de referencia global es sumamente perjudicial y debería ser evitada. Por otro lado se propone un sistema de detección de objetos móviles basado en técnicas de estimación de densidad con funciones locales. Este tipo de técnicas es muy adecuada para el modelado de escenas complejas con fondos multimodales, pero ha recibido poco uso debido a su gran complejidad computacional. El sistema propuesto, implementado en tiempo real sobre una GPU, incluye propuestas para la estimación dinámica de los anchos de banda de las funciones locales, actualización selectiva del modelo de fondo, actualización de la posición de las muestras de referencia del modelo de primer plano empleando un filtro de partículas multirregión y selección automática de regiones de interés para reducir el coste computacional. Los resultados, evaluados sobre diversas bases de datos y comparados con otros algoritmos del estado del arte, demuestran la gran versatilidad y calidad de la propuesta. Finalmente se propone un método para la aproximación de funciones arbitrarias empleando funciones continuas lineales a tramos, especialmente indicada para su implementación en una GPU mediante el uso de las unidades de filtraje de texturas, normalmente no utilizadas para cómputo numérico. La propuesta incluye un riguroso análisis matemático del error cometido en la aproximación en función del número de muestras empleadas, así como un método para la obtención de una partición cuasióptima del dominio de la función para minimizar el error. ABSTRACT The evolution of smartphones, all equipped with digital cameras, is driving a growing demand for ever more complex applications that need to rely on real-time computer vision algorithms. However, video signals are only increasing in size, whereas the performance of single-core processors has somewhat stagnated in the past few years. Consequently, new computer vision algorithms will need to be parallel to run on multiple processors and be computationally scalable. One of the most promising classes of processors nowadays can be found in graphics processing units (GPU). These are devices offering a high parallelism degree, excellent numerical performance and increasing versatility, which makes them interesting to run scientific computations. In this thesis, we explore two computer vision applications with a high computational complexity that precludes them from running in real time on traditional uniprocessors. However, we show that by parallelizing subtasks and implementing them on a GPU, both applications attain their goals of running at interactive frame rates. In addition, we propose a technique for fast evaluation of arbitrarily complex functions, specially designed for GPU implementation. First, we explore the application of depth-image–based rendering techniques to the unusual configuration of two convergent, wide baseline cameras, in contrast to the usual configuration used in 3D TV, which are narrow baseline, parallel cameras. By using a backward mapping approach with a depth inpainting scheme based on median filters, we show that these techniques are adequate for free viewpoint video applications. In addition, we show that referring depth information to a global reference system is ill-advised and should be avoided. Then, we propose a background subtraction system based on kernel density estimation techniques. These techniques are very adequate for modelling complex scenes featuring multimodal backgrounds, but have not been so popular due to their huge computational and memory complexity. The proposed system, implemented in real time on a GPU, features novel proposals for dynamic kernel bandwidth estimation for the background model, selective update of the background model, update of the position of reference samples of the foreground model using a multi-region particle filter, and automatic selection of regions of interest to reduce computational cost. The results, evaluated on several databases and compared to other state-of-the-art algorithms, demonstrate the high quality and versatility of our proposal. Finally, we propose a general method for the approximation of arbitrarily complex functions using continuous piecewise linear functions, specially formulated for GPU implementation by leveraging their texture filtering units, normally unused for numerical computation. Our proposal features a rigorous mathematical analysis of the approximation error in function of the number of samples, as well as a method to obtain a suboptimal partition of the domain of the function to minimize approximation error.

Contribution to Real-Time Estimation of Crop Phenological States in a Dynamical Framework Based on NDVI Time Series: Data Fusion With SAR and Temperature

In this study, a methodology based in a dynamical framework is proposed to incorporate additional sources of information to normalized difference vegetation index (NDVI) time series of agricultural observations for a phenological state estimation application. The proposed implementation is based on the particle filter (PF) scheme that is able to integrate multiple sources of data. Moreover, the dynamics-led design is able to conduct real-time (online) estimations, i.e., without requiring to wait until the end of the campaign. The evaluation of the algorithm is performed by estimating the phenological states over a set of rice fields in Seville (SW, Spain). A Landsat-5/7 NDVI series of images is complemented with two distinct sources of information: SAR images from the TerraSAR-X satellite and air temperature information from a ground-based station. An improvement in the overall estimation accuracy is obtained, especially when the time series of NDVI data is incomplete. Evaluations on the sensitivity to different development intervals and on the mitigation of discontinuities of the time series are also addressed in this work, demonstrating the benefits of this data fusion approach based on the dynamic systems.

Quantifying simulator discrepancy in discrete-time dynamical simulators

When making predictions with complex simulators it can be important to quantify the various sources of uncertainty. Errors in the structural specification of the simulator, for example due to missing processes or incorrect mathematical specification, can be a major source of uncertainty, but are often ignored. We introduce a methodology for inferring the discrepancy between the simulator and the system in discrete-time dynamical simulators. We assume a structural form for the discrepancy function, and show how to infer the maximum-likelihood parameter estimates using a particle filter embedded within a Monte Carlo expectation maximization (MCEM) algorithm. We illustrate the method on a conceptual rainfall-runoff simulator (logSPM) used to model the Abercrombie catchment in Australia. We assess the simulator and discrepancy model on the basis of their predictive performance using proper scoring rules. This article has supplementary material online. © 2011 International Biometric Society.

Short-term electricity demand and gas price forecasts using wavelet transforms and adaptive models

This paper presents some forecasting techniques for energy demand and price prediction, one day ahead. These techniques combine wavelet transform (WT) with fixed and adaptive machine learning/time series models (multi-layer perceptron (MLP), radial basis functions, linear regression, or GARCH). To create an adaptive model, we use an extended Kalman filter or particle filter to update the parameters continuously on the test set. The adaptive GARCH model is a new contribution, broadening the applicability of GARCH methods. We empirically compared two approaches of combining the WT with prediction models: multicomponent forecasts and direct forecasts. These techniques are applied to large sets of real data (both stationary and non-stationary) from the UK energy markets, so as to provide comparative results that are statistically stronger than those previously reported. The results showed that the forecasting accuracy is significantly improved by using the WT and adaptive models. The best models on the electricity demand/gas price forecast are the adaptive MLP/GARCH with the multicomponent forecast; their MSEs are 0.02314 and 0.15384 respectively.