Sven Fleck | University of Tübingen (original) (raw)
Papers by Sven Fleck
2005 13th European Signal Processing Conference, Sep 1, 2005
Tracking applications using distributed sensor networks are emerging today, both in the field of ... more Tracking applications using distributed sensor networks are emerging today, both in the field of surveillance (airports, train stations, museums, public spots) and industrial vision (visual servoing, factory automation). Traditional centralized approaches offer several drawbacks, due to limited communication bandwidth, computational requirements and thus also limited spatial camera resolution and framerate. In this paper we present a network-enabled Smart Camera for probabilistic tracking. It is capable of tracking objects in real-time and offers a very bandwidth-conservative approach, as it only transmits the tracking results which are on a higher level of abstraction.
13th European Signal …, 2005
A SMART CAMERA APPROACH TO REAL-TIME TRACKING Sven Fleck, Sven Lanwer, Wolfgang Straßer WSI/GRIS,... more A SMART CAMERA APPROACH TO REAL-TIME TRACKING Sven Fleck, Sven Lanwer, Wolfgang Straßer WSI/GRIS, University of Tübingen Sand 14, 72076 Tübingen, Germany phone: +(49) 7071 2970435, fax: + (49) 7071 295466, email: {fleck,lanwer,strasser}@gris.uni ...
Proceedings of the third …, 2007
Google, Inc. (search). ...
Proceedings of The IEEE, 2008
... of Graphical-Interactive Systems, University of Tübingen, Tübingen, Germany (e-mail: fleck@gr... more ... of Graphical-Interactive Systems, University of Tübingen, Tübingen, Germany (e-mail: fleck@gris.uni-tuebingen.de; strasser@gris.uni-tuebingen ... A distributed smart camera approach is proposed that instead qualifies for surveillance in privacy sensitive areas due to the fact that ...
Western societies are aging rapidly. According to [1] the US population over age 85 will triple f... more Western societies are aging rapidly. According to [1] the US population over age 85 will triple from 4.5 million in 2003 to 14.2 million by 2040. Automated 24/7 surveillance to ensure safety of the elderly while respecting privacy becomes a major challenge. As a first step, we present a fully distributed system and architecture based on a network of smart cameras to perform geo-referenced tracking and activity recognition. Additionally an intuitive visualization is presented, where all information is integrated in one world model that is available ubiquitously. At a real world installation in an assisted living home, we demonstrate first results regarding falling person detection as one major activity for assisted living systems.
Autonomous mobile robots will play a major role in future security and surveillance tasks for lar... more Autonomous mobile robots will play a major role in future security and surveillance tasks for large scale environments such as shopping malls, airports, hospitals and museums. Robotic security guards will autonomously survey such environments, unless a remote human operator takes over control. In this context a 3D model can convey much more useful information than the typical 2D maps used in many robotic applications today, both for visualization of information and as human machine interface for remote control.
We present a probabilistic framework for matching of point clouds. Variants of the ICP algorithm ... more We present a probabilistic framework for matching of point clouds. Variants of the ICP algorithm typically pair points to points or points to lines. Instead, we pair data points to probability functions that are thought of having generated the data points. Then an energy function is derived from a maximum likelihood formulation. Each such distribution is a mixture of a bivariate Normal Distribution to capture the local structure of points and an explicit outlier term to achieve robustness. We apply our approach to the SLAM problem in robotics using a 2D laser range scanner.
One major challenge of 3DTV is content acquisition. Here, we present a method to acquire a realis... more One major challenge of 3DTV is content acquisition. Here, we present a method to acquire a realistic, visually convincing 3D model of indoor environments based on a mobile platform that is equipped with a laser range scanner and a panoramic camera. The data of the 2D laser scans are used to solve the simultaneous localization and mapping problem and to extract walls. Textures for walls and floor are built from the images of a calibrated panoramic camera. Multiresolution blending is used to hide seams in the generated textures. The scene is further enriched by 3D-geometry calculated from a graph cut stereo technique. We present experimental results from a moderately large real environment. 1
Model based tracking is one key component of many systems today, e.g. within video surveillance o... more Model based tracking is one key component of many systems today, e.g. within video surveillance or human computer interfaces (HCI). Our approach consists of a combination of particle filters (PFs) and active appearance models (AAMs): the PFAAM. It combines the robustness of PFs with the precision of AAMs. Experimental results are given. PFAAM shows superior perfomance compared to both standard AAMs and PFs using AAMs as cues only, i.e. without using a local optimization loop.
Mechanisms for automatically acquiring high quality 3D content of both indoor and outdoor scenes ... more Mechanisms for automatically acquiring high quality 3D content of both indoor and outdoor scenes are essential research topics within 3D Television (3DTV). The envisioned goal is that a photo-realistic 3D real-time rendering from the actual and potentially arbitrary viewpoint of the beholder who is watching 3DTV becomes possible. Such scenes include movie sets in studios, e.g., for talk shows, TV series and blockbuster movies, but also outdoor scenes, e.g., buildings in a neighborhood for a car chase or cultural heritage sites for a documentary. The goal of 3D model acquisition is to provide the 3D background models where potential 3D actors can be embedded. Although model acquisition platforms have become available, the dynamic range of the used cameras is too limited to fully cover real world environments. We present two multi-sensor systems for 3D model acquisition, demonstrate the need for high dynamic range (HDR) capable acquisition platforms and propose a system that performs HDR computation to produce 3D models with tonemapped HDR information. Combining depth and HDR color information, a Markov Random Field based technique is used to get best out of both worlds -the high spatial resolution of the tonemapped HDR image with the depth field of a time-of-flight camera.
Applying the machine-learning technique of inference in Markov Random Fields we build improved 3D... more Applying the machine-learning technique of inference in Markov Random Fields we build improved 3D models by integrating two different modalities. Visual input from a standard color camera delivers high-resolution texture data but also enables us to enhance the 3D data calculated from the range output of a 3D time-of-flight camera in terms of noise and spatial resolution. The proposed method to increase the visual quality of the 3D data makes this kind of camera a promising device for various upcoming 3DTV applications. With our two-camera setup we believe that the design of lowcost, fast and highly portable 3D scene acquisition systems will be possible in the near future.
Multi-touch capable displays are one of the central emerging technologies in Human Computer Inter... more Multi-touch capable displays are one of the central emerging technologies in Human Computer Interfaces and many commercial applications like the Apple iPhone or the Microsoft Surface already show the benefit of this interaction technique. But most of the applications are limited to 2D interaction and only little effort has been spent on intuitive 3D interaction techniques. Since 3D scene navigation is a central aspect of 3DTV applications we present easy and intuitive multi-touch gestures to control a virtual 3D camera. These gestures were implemented on The DabR - a complete and robust multi-touch system designed at our institute. The underlying computer vision algorithms to cover the tracking of the users' fingers are GPU accelerated to ensure very low latency for the interaction.
Tracking applications based on distributed and embedded sensor networks are emerging today, both ... more Tracking applications based on distributed and embedded sensor networks are emerging today, both in the field of surveillance (airports, lab facilities, train stations, museums, public spots) and industrial vision (visual servoing, factory automation). Traditional centralized approaches offer several drawbacks, due to limited communication bandwidth, computational requirements and thus also limited spatial camera resolution and framerate. In this paper, we present a network-enabled Smart Camera for probabilistic tracking. It is capable of tracking objects adaptively in real-time and offers a very bandwidthconservative approach, as the whole computation is performed embedded in the Smart Camera, and only the tracking results are transmitted which are on a higher level of abstraction.
The demand for surveillance systems has increased extremely over recent times. We present a syste... more The demand for surveillance systems has increased extremely over recent times. We present a system consisting of a distributed network of cameras that allows for tracking and handover of multiple persons in real time. The intercamera tracking results are embedded as live textures in an integrated 3D world model which is available ubiquitously and can be viewed from arbitrary perspectives independent of the persons' movements. We mainly concentrate on our implementation of embedded camera nodes in the form of smart cameras and discuss the benefits of such a distributed surveillance network compared to a host centralized approach. We also briefly describe our way of hassle free 3D model acquisition to cover the complete system from setup to operation and finally show some results of both an indoor and an outdoor system in operation.
Autonomous mobile robots will play a major role in future security and surveillance tasks for lar... more Autonomous mobile robots will play a major role in future security and surveillance tasks for large scale environments such as shopping malls, airports, hospitals and museums. Robotic security guards will autonomously survey such environments, unless a remote human operator takes over control. In this context a 3D model can convey much more useful information than the typical 2D maps used in many robotic applications today, both for visualization of information and as human machine interface for remote control. This paper addresses the challenge of building such a model of a large environment (50x60m2) using data from the robot's own sensors: a 2D laser scanner and a panoramic camera. The data are processed in a pipeline that comprises automatic, semiautomatic and manual stages. The user can interact with the reconstruction process where necessary to ensure robustness and completeness of the model. A hybrid representation, tailored to the application, has been chosen: floors and walls are represented efficiently by textured planes. Non-planar structures like stairs and tables, which are represented by point clouds, can be added if desired. Our methods to extract these structures include: simultaneous localization and mapping in 2D and wall extraction based on laser scanner range data, building textures from multiple omnidirectional images using multiresolution blending, and calculation of 3D geometry by a graph cut stereo technique. Various renderings illustrate the usability of the model for visualizing the security guard's position and environment.
2005 13th European Signal Processing Conference, Sep 1, 2005
Tracking applications using distributed sensor networks are emerging today, both in the field of ... more Tracking applications using distributed sensor networks are emerging today, both in the field of surveillance (airports, train stations, museums, public spots) and industrial vision (visual servoing, factory automation). Traditional centralized approaches offer several drawbacks, due to limited communication bandwidth, computational requirements and thus also limited spatial camera resolution and framerate. In this paper we present a network-enabled Smart Camera for probabilistic tracking. It is capable of tracking objects in real-time and offers a very bandwidth-conservative approach, as it only transmits the tracking results which are on a higher level of abstraction.
13th European Signal …, 2005
A SMART CAMERA APPROACH TO REAL-TIME TRACKING Sven Fleck, Sven Lanwer, Wolfgang Straßer WSI/GRIS,... more A SMART CAMERA APPROACH TO REAL-TIME TRACKING Sven Fleck, Sven Lanwer, Wolfgang Straßer WSI/GRIS, University of Tübingen Sand 14, 72076 Tübingen, Germany phone: +(49) 7071 2970435, fax: + (49) 7071 295466, email: {fleck,lanwer,strasser}@gris.uni ...
Proceedings of the third …, 2007
Google, Inc. (search). ...
Proceedings of The IEEE, 2008
... of Graphical-Interactive Systems, University of Tübingen, Tübingen, Germany (e-mail: fleck@gr... more ... of Graphical-Interactive Systems, University of Tübingen, Tübingen, Germany (e-mail: fleck@gris.uni-tuebingen.de; strasser@gris.uni-tuebingen ... A distributed smart camera approach is proposed that instead qualifies for surveillance in privacy sensitive areas due to the fact that ...
Western societies are aging rapidly. According to [1] the US population over age 85 will triple f... more Western societies are aging rapidly. According to [1] the US population over age 85 will triple from 4.5 million in 2003 to 14.2 million by 2040. Automated 24/7 surveillance to ensure safety of the elderly while respecting privacy becomes a major challenge. As a first step, we present a fully distributed system and architecture based on a network of smart cameras to perform geo-referenced tracking and activity recognition. Additionally an intuitive visualization is presented, where all information is integrated in one world model that is available ubiquitously. At a real world installation in an assisted living home, we demonstrate first results regarding falling person detection as one major activity for assisted living systems.
Autonomous mobile robots will play a major role in future security and surveillance tasks for lar... more Autonomous mobile robots will play a major role in future security and surveillance tasks for large scale environments such as shopping malls, airports, hospitals and museums. Robotic security guards will autonomously survey such environments, unless a remote human operator takes over control. In this context a 3D model can convey much more useful information than the typical 2D maps used in many robotic applications today, both for visualization of information and as human machine interface for remote control.
We present a probabilistic framework for matching of point clouds. Variants of the ICP algorithm ... more We present a probabilistic framework for matching of point clouds. Variants of the ICP algorithm typically pair points to points or points to lines. Instead, we pair data points to probability functions that are thought of having generated the data points. Then an energy function is derived from a maximum likelihood formulation. Each such distribution is a mixture of a bivariate Normal Distribution to capture the local structure of points and an explicit outlier term to achieve robustness. We apply our approach to the SLAM problem in robotics using a 2D laser range scanner.
One major challenge of 3DTV is content acquisition. Here, we present a method to acquire a realis... more One major challenge of 3DTV is content acquisition. Here, we present a method to acquire a realistic, visually convincing 3D model of indoor environments based on a mobile platform that is equipped with a laser range scanner and a panoramic camera. The data of the 2D laser scans are used to solve the simultaneous localization and mapping problem and to extract walls. Textures for walls and floor are built from the images of a calibrated panoramic camera. Multiresolution blending is used to hide seams in the generated textures. The scene is further enriched by 3D-geometry calculated from a graph cut stereo technique. We present experimental results from a moderately large real environment. 1
Model based tracking is one key component of many systems today, e.g. within video surveillance o... more Model based tracking is one key component of many systems today, e.g. within video surveillance or human computer interfaces (HCI). Our approach consists of a combination of particle filters (PFs) and active appearance models (AAMs): the PFAAM. It combines the robustness of PFs with the precision of AAMs. Experimental results are given. PFAAM shows superior perfomance compared to both standard AAMs and PFs using AAMs as cues only, i.e. without using a local optimization loop.
Mechanisms for automatically acquiring high quality 3D content of both indoor and outdoor scenes ... more Mechanisms for automatically acquiring high quality 3D content of both indoor and outdoor scenes are essential research topics within 3D Television (3DTV). The envisioned goal is that a photo-realistic 3D real-time rendering from the actual and potentially arbitrary viewpoint of the beholder who is watching 3DTV becomes possible. Such scenes include movie sets in studios, e.g., for talk shows, TV series and blockbuster movies, but also outdoor scenes, e.g., buildings in a neighborhood for a car chase or cultural heritage sites for a documentary. The goal of 3D model acquisition is to provide the 3D background models where potential 3D actors can be embedded. Although model acquisition platforms have become available, the dynamic range of the used cameras is too limited to fully cover real world environments. We present two multi-sensor systems for 3D model acquisition, demonstrate the need for high dynamic range (HDR) capable acquisition platforms and propose a system that performs HDR computation to produce 3D models with tonemapped HDR information. Combining depth and HDR color information, a Markov Random Field based technique is used to get best out of both worlds -the high spatial resolution of the tonemapped HDR image with the depth field of a time-of-flight camera.
Applying the machine-learning technique of inference in Markov Random Fields we build improved 3D... more Applying the machine-learning technique of inference in Markov Random Fields we build improved 3D models by integrating two different modalities. Visual input from a standard color camera delivers high-resolution texture data but also enables us to enhance the 3D data calculated from the range output of a 3D time-of-flight camera in terms of noise and spatial resolution. The proposed method to increase the visual quality of the 3D data makes this kind of camera a promising device for various upcoming 3DTV applications. With our two-camera setup we believe that the design of lowcost, fast and highly portable 3D scene acquisition systems will be possible in the near future.
Multi-touch capable displays are one of the central emerging technologies in Human Computer Inter... more Multi-touch capable displays are one of the central emerging technologies in Human Computer Interfaces and many commercial applications like the Apple iPhone or the Microsoft Surface already show the benefit of this interaction technique. But most of the applications are limited to 2D interaction and only little effort has been spent on intuitive 3D interaction techniques. Since 3D scene navigation is a central aspect of 3DTV applications we present easy and intuitive multi-touch gestures to control a virtual 3D camera. These gestures were implemented on The DabR - a complete and robust multi-touch system designed at our institute. The underlying computer vision algorithms to cover the tracking of the users' fingers are GPU accelerated to ensure very low latency for the interaction.
Tracking applications based on distributed and embedded sensor networks are emerging today, both ... more Tracking applications based on distributed and embedded sensor networks are emerging today, both in the field of surveillance (airports, lab facilities, train stations, museums, public spots) and industrial vision (visual servoing, factory automation). Traditional centralized approaches offer several drawbacks, due to limited communication bandwidth, computational requirements and thus also limited spatial camera resolution and framerate. In this paper, we present a network-enabled Smart Camera for probabilistic tracking. It is capable of tracking objects adaptively in real-time and offers a very bandwidthconservative approach, as the whole computation is performed embedded in the Smart Camera, and only the tracking results are transmitted which are on a higher level of abstraction.
The demand for surveillance systems has increased extremely over recent times. We present a syste... more The demand for surveillance systems has increased extremely over recent times. We present a system consisting of a distributed network of cameras that allows for tracking and handover of multiple persons in real time. The intercamera tracking results are embedded as live textures in an integrated 3D world model which is available ubiquitously and can be viewed from arbitrary perspectives independent of the persons' movements. We mainly concentrate on our implementation of embedded camera nodes in the form of smart cameras and discuss the benefits of such a distributed surveillance network compared to a host centralized approach. We also briefly describe our way of hassle free 3D model acquisition to cover the complete system from setup to operation and finally show some results of both an indoor and an outdoor system in operation.
Autonomous mobile robots will play a major role in future security and surveillance tasks for lar... more Autonomous mobile robots will play a major role in future security and surveillance tasks for large scale environments such as shopping malls, airports, hospitals and museums. Robotic security guards will autonomously survey such environments, unless a remote human operator takes over control. In this context a 3D model can convey much more useful information than the typical 2D maps used in many robotic applications today, both for visualization of information and as human machine interface for remote control. This paper addresses the challenge of building such a model of a large environment (50x60m2) using data from the robot's own sensors: a 2D laser scanner and a panoramic camera. The data are processed in a pipeline that comprises automatic, semiautomatic and manual stages. The user can interact with the reconstruction process where necessary to ensure robustness and completeness of the model. A hybrid representation, tailored to the application, has been chosen: floors and walls are represented efficiently by textured planes. Non-planar structures like stairs and tables, which are represented by point clouds, can be added if desired. Our methods to extract these structures include: simultaneous localization and mapping in 2D and wall extraction based on laser scanner range data, building textures from multiple omnidirectional images using multiresolution blending, and calculation of 3D geometry by a graph cut stereo technique. Various renderings illustrate the usability of the model for visualizing the security guard's position and environment.