ResumenEn este articulo se presenta un resumen de la evolucion y los principales resultados obten... more ResumenEn este articulo se presenta un resumen de la evolucion y los principales resultados obtenidos en el marco del Proyecto Europeo ORION, desarrollado entre Julio de 2002 y Junio 2003, y cuyo objetivo principal fue la constitucion de una Red de expertos que permitiese crear una Hoja de Ruta para el desarrollo y empleo de la tecnologia 3D en el marco de la arqueologia y, mas especificamente, en el de los museos arqueologicos. Este resumen se plantea como un breve analisis de los principales resultados obtenidos en las diferentes etapas del proyecto – Revision de las practicas y necesidades de los museos arqueologicos y revision cientifica y tecnologica –, etapas que dieron lugar a la configuracion final de la Hoja de Ruta cuyas recomendaciones y conclusiones principales se resumen en la parte final de este articulo.Para la consecucion de estos objetivos se conto con un equilibrado Consorcio de 12 miembros de 7 paises europeos, que agrupo tanto a museos y centros arqueologicos lid...
9Th European Signal Processing Conference, Sep 1, 1998
Ego-motion requires the determination of the position and trajectory of the moving vehicle just b... more Ego-motion requires the determination of the position and trajectory of the moving vehicle just by analysing the sequence of images acquired along its movement. Here, we present an analytical procedure for the estimation of the three-dimensional kinematic parameters of the vehicle, and the depth map that describes the three-dimensional structure of the world that surrounds the acquisition camera (only monocular gray-level images are considered). Since we do not only consider the possibility of translational and rotational movement in the acquisition camera, but we also allow for variations in the focal distance of the lens (zoom), This proposal clearly extends the one proposed by Waxman et al. [IO]. Our proposed solution also provides the way to obtain an estimation of the variation of the focal distance that the camera has applied. The procedure is simple, and relies on the assumption of second order surface patches on the visualised objects.
An analytical procedure for the estimation of the focus of expansion (FOE) location is introduced... more An analytical procedure for the estimation of the focus of expansion (FOE) location is introduced. The method is applied to a vector field which has been obtained from the analysis of the relative translational movement of a rigid body with respect to the acquiring camera (monocular system). A polynomial model is fitted to the vector field; the order and shape
A new segmentation strategy is proposed to precisely extract moving objects in video sequences. I... more A new segmentation strategy is proposed to precisely extract moving objects in video sequences. It is based on the automatic detection of the static elements, and its classification as background and foreground using static differences and contextual information. ...
Video analytics play a critical role in most recent traffic monitoring and driver assistance syst... more Video analytics play a critical role in most recent traffic monitoring and driver assistance systems. In this context, the correct detection and classification of surrounding vehicles through image analysis has been the focus of extensive research in the last years. Most of the pieces of work reported for image-based vehicle verification make use of supervised classification approaches and resort to techniques, such as histograms of oriented gradients (HOG), principal component analysis (PCA), and Gabor filters, among others. Unfortunately, existing approaches are lacking in two respects: first, comparison between methods using a common body of work has not been addressed; second, no study of the combination potentiality of popular features for vehicle classification has been reported. In this study the performance of the different techniques is first reviewed and compared using a common public database. Then, the combination capabilities of these techniques are explored and a metho...
This work addresses the use of image analysis and computer vision in the context of Advance Drive... more This work addresses the use of image analysis and computer vision in the context of Advance Driver Assistance Systems (ADAS). Video-based systems stand as a powerful complement to classical active-sensor based approaches, as their being non-intrusive avoids interferences between sensors. They also allow a deeper understanding of the scene and offer better perspectives in terms of cost and flexibility. The main challenges in this field are the huge within class variability of vehicles, the complexity of the background induced by camera movement, and the effect of illumination and weather conditions on their appearance. Particularly, automatic vehicle detection and tracking from an on-board forward looking camera is considered in this work. A unified approach using statistical methods is proposed as the main contribution, so that the vehicles can be identified from the background irrespective of the scene conditions (e.g., weather, time of the day, own vehicle velocity) and vehicle appearance. The approach is divided into three primary tasks: vehicle hypothesis generation, hypothesis verification and vehicle tracking. Hypothesis generation is based on the use of a rectified domain in which the perspective is removed from the original image. A supervised classification strategy is adopted for the verification of the hypothesized vehicle locations, evaluating the performance of different methods for feature extraction. Finally, a Bayesian tracking framework using particle filters is proposed, in which a constant velocity dynamic model is used together with a multi-cue observation model based on appearance analysis on the rectified and original domains. The evaluation on real sequences demonstrates the robustness of the proposed framework. The exchange of information between processing blocks is fostered so that the maximum degree of adaptation to changes in the environment can be achieved and the computational cost is alleviated, rendering high detection rates in the verification phase, while reducing the number of tracking failures.ResumenEl presente trabajo se centra en el uso de las técnicas análisis de imagen y visión artificial en el ámbito de los sistemas avanzados de ayuda a la conducción, comúnmente conocido por sus siglas en inglés, ADAS (Advanced Driver Assistance Systems). Los sistemas basados en visión artificial suponen un complemento muy potente a los sistemas clásicos que usan sensores activos, ya que su carácter no-intrusivo evita posibles interferencias entre vehículos, ofreciendo una mayor riqueza de información con un coste más asequible. Los principales retos para la visión por ordenador en este campo son la gran heterogeneidad de los vehículos y su entorno, el efecto de la iluminación y las condiciones meteorológicas en su apariencia. Particularmente, se aborda la detección y el seguimiento automático de vehículos con una cámara monocular embarcada y apuntando en la dirección de avance del vehículo. Se plantea como contribución principal un marco global con un enfoque estadístico que tenga en cuenta la variabilidad del entorno, de manera que los vehículos se puedan identificar independientemente de su apariencia y de las condiciones de la escena. El marco se compone de tres bloques: generación de hipótesis, verificación de hipótesis, y seguimiento. La generación de hipótesis, detección de regiones donde potencialmente hay vehículos, se basa en una estimación robusta de la perspectiva de la escena que permite generar una imagen rectificada del plano de la imagen. En la verificación de hipótesis se explora un método de aprendizaje supervisado evaluando el rendimiento de diferentes características. Finalmente, el seguimiento propuesto está constituido por un método Bayesiano basado en filtros de partículas, donde se define un modelo dinámico de velocidad constante, y se propone un modelo de observación combinando modelos de apariencia en el dominio original y transformado. Los resultados obtenidos sobre secuencias reales demuestran la robustez del marco propuesto. Se potencia el intercambio de información entre los diferentes bloques con objeto de obtener el mayor grado posible de adaptación a cambios en el entorno y reducir el coste computacional, permitiendo elevadas tasas de acierto en la fase de verificación y reduciendo el número de fallos en el seguimiento.
ABSTRACT A novel approach to real-time lane modeling using a single camera is proposed. The propo... more ABSTRACT A novel approach to real-time lane modeling using a single camera is proposed. The proposed method is based on an efficient design and implementation of a particle filter which applies the concepts of the Rao-Blackwellized particle filter (RBPF) by separating the state into linear and non-linear parts. As a result the dimensionality of the problem is reduced, which allows the system to perform in real-time in embedded systems. The method is used to determine the position of the vehicle inside its own lane and the curvature of the road ahead to enhance the performance of advanced driver assistance systems. The effectiveness of the method has been demonstrated implementing a prototype and testing its performance empirically on road sequences with different illumination conditions (day and nightime), pavement types, traffic density, etc. Results show that our proposal is capable of accurately determining if the vehicle is approaching the lane markings (Lane Departure Warning), and the curvature of the road ahead, achieving processing times below 2 ms per frame for laptop CPUs, and 12 ms for embedded CPUs.
ResumenEn este articulo se presenta un resumen de la evolucion y los principales resultados obten... more ResumenEn este articulo se presenta un resumen de la evolucion y los principales resultados obtenidos en el marco del Proyecto Europeo ORION, desarrollado entre Julio de 2002 y Junio 2003, y cuyo objetivo principal fue la constitucion de una Red de expertos que permitiese crear una Hoja de Ruta para el desarrollo y empleo de la tecnologia 3D en el marco de la arqueologia y, mas especificamente, en el de los museos arqueologicos. Este resumen se plantea como un breve analisis de los principales resultados obtenidos en las diferentes etapas del proyecto – Revision de las practicas y necesidades de los museos arqueologicos y revision cientifica y tecnologica –, etapas que dieron lugar a la configuracion final de la Hoja de Ruta cuyas recomendaciones y conclusiones principales se resumen en la parte final de este articulo.Para la consecucion de estos objetivos se conto con un equilibrado Consorcio de 12 miembros de 7 paises europeos, que agrupo tanto a museos y centros arqueologicos lid...
9Th European Signal Processing Conference, Sep 1, 1998
Ego-motion requires the determination of the position and trajectory of the moving vehicle just b... more Ego-motion requires the determination of the position and trajectory of the moving vehicle just by analysing the sequence of images acquired along its movement. Here, we present an analytical procedure for the estimation of the three-dimensional kinematic parameters of the vehicle, and the depth map that describes the three-dimensional structure of the world that surrounds the acquisition camera (only monocular gray-level images are considered). Since we do not only consider the possibility of translational and rotational movement in the acquisition camera, but we also allow for variations in the focal distance of the lens (zoom), This proposal clearly extends the one proposed by Waxman et al. [IO]. Our proposed solution also provides the way to obtain an estimation of the variation of the focal distance that the camera has applied. The procedure is simple, and relies on the assumption of second order surface patches on the visualised objects.
An analytical procedure for the estimation of the focus of expansion (FOE) location is introduced... more An analytical procedure for the estimation of the focus of expansion (FOE) location is introduced. The method is applied to a vector field which has been obtained from the analysis of the relative translational movement of a rigid body with respect to the acquiring camera (monocular system). A polynomial model is fitted to the vector field; the order and shape
A new segmentation strategy is proposed to precisely extract moving objects in video sequences. I... more A new segmentation strategy is proposed to precisely extract moving objects in video sequences. It is based on the automatic detection of the static elements, and its classification as background and foreground using static differences and contextual information. ...
Video analytics play a critical role in most recent traffic monitoring and driver assistance syst... more Video analytics play a critical role in most recent traffic monitoring and driver assistance systems. In this context, the correct detection and classification of surrounding vehicles through image analysis has been the focus of extensive research in the last years. Most of the pieces of work reported for image-based vehicle verification make use of supervised classification approaches and resort to techniques, such as histograms of oriented gradients (HOG), principal component analysis (PCA), and Gabor filters, among others. Unfortunately, existing approaches are lacking in two respects: first, comparison between methods using a common body of work has not been addressed; second, no study of the combination potentiality of popular features for vehicle classification has been reported. In this study the performance of the different techniques is first reviewed and compared using a common public database. Then, the combination capabilities of these techniques are explored and a metho...
This work addresses the use of image analysis and computer vision in the context of Advance Drive... more This work addresses the use of image analysis and computer vision in the context of Advance Driver Assistance Systems (ADAS). Video-based systems stand as a powerful complement to classical active-sensor based approaches, as their being non-intrusive avoids interferences between sensors. They also allow a deeper understanding of the scene and offer better perspectives in terms of cost and flexibility. The main challenges in this field are the huge within class variability of vehicles, the complexity of the background induced by camera movement, and the effect of illumination and weather conditions on their appearance. Particularly, automatic vehicle detection and tracking from an on-board forward looking camera is considered in this work. A unified approach using statistical methods is proposed as the main contribution, so that the vehicles can be identified from the background irrespective of the scene conditions (e.g., weather, time of the day, own vehicle velocity) and vehicle appearance. The approach is divided into three primary tasks: vehicle hypothesis generation, hypothesis verification and vehicle tracking. Hypothesis generation is based on the use of a rectified domain in which the perspective is removed from the original image. A supervised classification strategy is adopted for the verification of the hypothesized vehicle locations, evaluating the performance of different methods for feature extraction. Finally, a Bayesian tracking framework using particle filters is proposed, in which a constant velocity dynamic model is used together with a multi-cue observation model based on appearance analysis on the rectified and original domains. The evaluation on real sequences demonstrates the robustness of the proposed framework. The exchange of information between processing blocks is fostered so that the maximum degree of adaptation to changes in the environment can be achieved and the computational cost is alleviated, rendering high detection rates in the verification phase, while reducing the number of tracking failures.ResumenEl presente trabajo se centra en el uso de las técnicas análisis de imagen y visión artificial en el ámbito de los sistemas avanzados de ayuda a la conducción, comúnmente conocido por sus siglas en inglés, ADAS (Advanced Driver Assistance Systems). Los sistemas basados en visión artificial suponen un complemento muy potente a los sistemas clásicos que usan sensores activos, ya que su carácter no-intrusivo evita posibles interferencias entre vehículos, ofreciendo una mayor riqueza de información con un coste más asequible. Los principales retos para la visión por ordenador en este campo son la gran heterogeneidad de los vehículos y su entorno, el efecto de la iluminación y las condiciones meteorológicas en su apariencia. Particularmente, se aborda la detección y el seguimiento automático de vehículos con una cámara monocular embarcada y apuntando en la dirección de avance del vehículo. Se plantea como contribución principal un marco global con un enfoque estadístico que tenga en cuenta la variabilidad del entorno, de manera que los vehículos se puedan identificar independientemente de su apariencia y de las condiciones de la escena. El marco se compone de tres bloques: generación de hipótesis, verificación de hipótesis, y seguimiento. La generación de hipótesis, detección de regiones donde potencialmente hay vehículos, se basa en una estimación robusta de la perspectiva de la escena que permite generar una imagen rectificada del plano de la imagen. En la verificación de hipótesis se explora un método de aprendizaje supervisado evaluando el rendimiento de diferentes características. Finalmente, el seguimiento propuesto está constituido por un método Bayesiano basado en filtros de partículas, donde se define un modelo dinámico de velocidad constante, y se propone un modelo de observación combinando modelos de apariencia en el dominio original y transformado. Los resultados obtenidos sobre secuencias reales demuestran la robustez del marco propuesto. Se potencia el intercambio de información entre los diferentes bloques con objeto de obtener el mayor grado posible de adaptación a cambios en el entorno y reducir el coste computacional, permitiendo elevadas tasas de acierto en la fase de verificación y reduciendo el número de fallos en el seguimiento.
ABSTRACT A novel approach to real-time lane modeling using a single camera is proposed. The propo... more ABSTRACT A novel approach to real-time lane modeling using a single camera is proposed. The proposed method is based on an efficient design and implementation of a particle filter which applies the concepts of the Rao-Blackwellized particle filter (RBPF) by separating the state into linear and non-linear parts. As a result the dimensionality of the problem is reduced, which allows the system to perform in real-time in embedded systems. The method is used to determine the position of the vehicle inside its own lane and the curvature of the road ahead to enhance the performance of advanced driver assistance systems. The effectiveness of the method has been demonstrated implementing a prototype and testing its performance empirically on road sequences with different illumination conditions (day and nightime), pavement types, traffic density, etc. Results show that our proposal is capable of accurately determining if the vehicle is approaching the lane markings (Lane Departure Warning), and the curvature of the road ahead, achieving processing times below 2 ms per frame for laptop CPUs, and 12 ms for embedded CPUs.
Uploads