• Publikationen
Show publication details

Borrmann, Daniel; Kuijper, Arjan [1. Gutachter]; Franek, Andreas [2. Gutachter]

Incorporating the Unscented Transform in Rao-Blackwellised Visual-Inertial SLAM

2016

Darmstadt, TU, Master Thesis, 2016

Monocular visual-inertial simultaneous localisation and mapping (SLAM) systems use a single camera and inertial sensors to build a sparse map of an environment and to simultaneously estimate the camera's position. Applications can be found, for example, in the robotics domain or in virtual and augmented reality applications. Nowadays, monocular visual SLAM systems are capable of operating in real-time on different platforms. The application of Rao-Blackwellised filtering techniques in visual SLAM has further improved the real-time capability and maximum map size of these systems. In many implementations, extended Kalman filters (EKFs) are used to estimate the camera's pose and landmark positions. Since the required dynamic and measurement functions in a visual SLAM system are usually non-linear, EKF-based implementations need to linearise these functions, which leads to linearisation errors in the state estimate. In these cases, the unscented Kalman filter (UKF) proved to provide better results in many scenarios. In this thesis, the unscented transform (UT) is incorporated into a Rao-Blackwellised visual-inertial SLAM system. It shows how the camera state and landmark positions can be estimated in unscented Rao-Blackwellised filtering. To this end, an appropriate motion model, as well as an inverse depth parametrisation for the landmark estimation, are incorporated into the filter. The proposed filter is tested on a visual-inertial dataset in different setups. Results are given for the UKF-based Rao- Blackwellised filter and an equivalent EKF-based implementation, to enable comparison between the two approaches.

Show publication details

Franek, Andreas; Ackermann, Jens [Betreuer]; Goesele, Michael [Gutachter]

Enhancing GPS Precision using Structure from Motion

2013

Darmstadt, TU, Master Thesis, 2013

GPS positional data has been used to enhance Structure from Motion (SfM) reconstructions of large scenes [Lhu11, Ell04]. Here we evaluate the opposite approach; utilizing SfM reconstructions of scenes between 50 and 500m to enhance the GPS meta data associated with the images used in SfM. These images are here taken from online community photo collections. The statistical properties of SfM reconstructed camera positions and GPS measurements are considered and found to be very distinct. Further, the distribution of geotags created manually by users and that of GPS receiver created geotags in online community photo collections is examined. The result shows that images in online community photo collections scarcely contain reliable GPS meta data. Because of this it is hard to estimate a transformation between GPS and SfM space. With such a transformation it is possible to improve the accuracy of inaccurate geotags and create geotags for images not containing one previously. An algorithm is designed that suits the statistical properties of SfM reconstructions and GPS measurements. Using this algorithm it is possible to enhance the geotags of SfM image bundles even when only few reliable are present initially.

Show publication details

Franek, Andreas; Bein, Matthias [Betreuer]

Evolutionäre B-Spline Approximation mit markierten und uniformen Knoten

2010

Darmstadt, TU, Bachelor Thesis, 2010

B-Splines eignen sich zur Approximation von Hand eingegebener Kurven. In diesem speziellen Fall soll eine Approximation gefunden werden die möglichst nah an der eingegebenen Kurve ist, jedoch durch wenige Kontrollpunkte definiert ist. Die Approximation solch einer Kurve ist jedoch ein schwieriges Optimierungsproblem. Evolutionäre Algorithmen haben sich bei dieser Art von Optimierungsproblemen als nützlich erwiesen. Auch in anderen Arbeiten wurden schon evolutionäre Algorithmen zur Approximation mit B-Splines eingesetzt. Da die hier verwendete Repräsentation der B-Splines sehr speziell ist sind die dort verwendeten Techniken allerdings nur bedingt anwendbar. Die Approximation von Kurven durch B-Splines ebenso wie Evolutionäre Algorithmen haben viele austauschbare Komponenten, wie beispielsweise Fitnessfunktion, das Maß mit dem die Kontrollpunkte den Fehler beeinflussen, Seeding- und Evolutionsfunktionen. Um den Algorithmus erweiterbar für diese verschiedenen Möglichkeiten und auch neue Entwicklungen zu machen wird der Algorithmus in ein Framework eingebettet.