Thomas Gumpp

Unconstrained Real-time Markerless Hand Tracking for Humanoid Interaction

Markerless hand tracking of humans can be applied to a broad range of applications, in robotics, animation and natural human-computer interaction. Traditional motion capture and tracking methods involve the usage of devices such as a data glove, or marker points that are fixed and calibrated on the object to perform tracking. Markerless tracking is free from such needs, and therefore allows for more freedom in movement and spontaneous interaction. In this paper, we analyze how a hand tracking system, which reliably tracks arbitrary hand movements can be implemented. We explored a model based approach that uses particle filters for tracking. In this study we also determine the degree to which the inherent parallel properties of particle filter can be exploited to achieve the goal of real-time tracking. We present the effectiveness of the tracking system via the realtime control of a 20 degrees of freedom dexterous robotic hand

A Situation context aware Dempster-Shafer fusion of digital maps and a road sign recognition system

Speed limit information systems solely based on one modality can hardly overcome their respective intrinsic disadvantages: Digital maps lack support for short-term changes brought by variable message signs and road works, while camera based systems cannot recognize implicit speed limits and may fail in adverse lighting scenarios. In this work we show a fusion approach that is able to overcome these limitations. It is specifically tailored to our task by adapting sensor reliability based on the perceived situation context. This enables the camera based system to easily outvote the digital map in a construction site or the digital map to veto against uncertain camera recognition results during nighttime. The proposed fusion approach was implemented and evaluated on a qualitative base showing very promising results: The situation context aware fusion is able to deduce the correct effective speed limit even when one of the sensors fails. Moreover, it reduces conflicts encountered between the sources compared to a not situation context aware fusion.

Recognition and tracking of temporary lanes in motorway construction sites

We propose a system for detecting a construction site on motorways, and subsequently track its individual lanes using kalman filters. Beacons and yellow lane markers are extracted from color images and provide indicators for a construction site situation. Yellow lane markers are used in this case for tracking the temporary lanes at road works. When leaving the construction site, again white road markers are extracted to track regular motorway lanes.

Fast and reliable recognition of supplementary traffic signs

Supplementary traffic signs are used to alter the meaning of other traffic signs. Assistance systems that recognize traffic signs therefore must also recognize supplementary signs to evaluate their influence on the meaning of detected traffic signs. We propose an algorithm which is able to detect supplementary signs in the vicinity of other signs using a novel rectangle segmentation algorithm. Support vector machines are used for the classification and rejection of other objects. The combination of both components permits to recognize a supplementary sign in less than 40 ms. First quantitative results for a test set with four different supplementary sign types show a very good classification accuracy of more than 96 %.

Recognition and attribution of variable message signs and lanes

Knowledge of the position of traffic signs, lanes and the own vehicle is needed to decide whether a sign applies to oneself. This paper describes a system capable of tracking lanes and recognizing speed limit signs, both static ones and variable message signs. In addition a probabilistic approach mapping detected signs to lanes is presented which is able to correctly attribute signs to lanes even at great distance. All information is extracted from a monocular video camera.

Relevance estimation of traffic elements using Markov logic networks

Complex traffic situations e.g. at intersections consist of many traffic participants, traffic elements and relations between them. The behavior of participants is constrained by implicit and explicit traffic rules. We are interested in estimating whether a given traffic element — a traffic sign, a traffic light — is relevant in the current driving situation, i.e. affects the set of possible legal actions. A wide variety of properties influences the relevance. The route to take for example affects which traffic lights are relevant and the current weather situation affects whether a speed limit restricted by a supplementary sign is in effect. We use first-order logic to model such relations and apply reasoning to decide upon the relevance of static traffic elements. The need for perfect information is alleviated with the help of Markov logic networks, reconciling hard decision rules on the one hand and uncertainty intrinsic to the environment perception process on the other hand. The evaluation of twelve intersection scenes shows very promising results for the relevance estimation of traffic lights: Markov logic networks are able to judge whether enough information is available and determine the relevant traffic lights reliably in such cases.

Kamera-basierte Erkennung von Geschwindigkeitsbeschränkungen auf deutschen Stra\en

An Fahrerassistenzsysteme im industriellen Einsatz werden hohe Anforderungen bezuglich Zuverlassigkeit und Robustheit gestellt. In dieser Arbeit wird die Kombination robuster Verfahren wie der Hough-Transformation und Support-Vektor-Maschinen zu einem Gesamtsystem zur Erkennung von Geschwindigkeitsbeschrankungen beschrieben. Es setzt eine Farbvideokamera als Sensorik ein. Die Evaluation auf Testdaten bestatigt durch die ermittelte hohe Korrektklassifikationsrate bei gleichzeitig geringer Zahl Fehlalarme die Zuverlassigkeit des Systems.

PMD basierte Fahrspurerkennung und -Verfolgung für Fahrerassistenzsysteme

In diesem Artikel wird ein System zur Verfolgung von Fahrspuren unter Verwendung von PMD-Kameras vorgestellt. Er gibt einen uberblick uber die Auswertung der Intensitats- und Tiefenbilder dieses Sensors in einem System zur fremdlichtunabhangigen Fahrspurverfolgung. Ein Kaiman-Filter wird verwendet, um neben den Fahrspurparametern auch die Position und Orientierung des Fahrzeugs relativ zur Fahrspur zu schatzten.

Probabilistic hierarchical detection, representation and scene interpretation of lanes and roads

The focus of this paper is to propose a concept for integrated detection, representation and interpretation of lanes and roads as well as their possible roles in the vehicles surrounding. This includes a hierarchical probabilistic representation using particle approximation of multiple probability density functions for different levels of abstraction. Low- and high-level information can be integrated, leading to mutual bottom-up and top-down refinement of scene representation. Based on this representation bayesian networks are modeled for probabilistically inferring abstract, not directly observable relations. Based on these relations, a consistent subset of all hypotheses is generated to represent the current situation. The approach is highly flexible, able to integrate different information sources of varying levels of abstraction, while preserving a high level of probabilistic detail.

Energy Efficient Driving and Operation Strategies Based on Situation Awareness and Reasoning

Abstract A vehicle´s basic energy consumption is determined by its design, but the driving behavior is another important factor having a major influence on the overall consumption. In this work we show how information technology can be used to gain an understanding of the whole traffic situation and use it to derive energy optimal behavior. It can be used to optimize operation strategies and in cooperation with the driver helping him to maximize fuel saving by anticipatory driving. We summarize the research challenges that need to be solved to create energy efficiency assistance systems and present an overview of our solutions. Zusammenfassung Der Basis-Energieverbrauch eines Fahrzeugs wird durch dessen Design festgelegt, aber das Fahrverhalten hat ebenfalls einen großen Einfluss auf den Gesamtverbrauch. In dieser Arbeit zeigen wir, wie Informationstechnik eingesetzt werden kann, um ein Gesamtverständnis der Verkehrssituation aufzubauen und davon ausgehend energieoptimales Verhalten abzuleiten. Transparent für den Fahrer wird es zur Optimierung von Betriebsstrategien des Fahrzeugs eingesetzt oder in Kooperation mit dem Fahrer als Hilfestellungen zum vorausschauenden Fahren angeboten. Wir fassen die wissenschaftlichen Herausforderungen zusammen, die bei der Realisierung energieeffizienter Assistenzsysteme gelöst werden müssen und geben einen Überblick über unsere Lösungen.