Heterogene Sensordatenfusion zur robusten Objektverfolgung im automobilen Straßenverkehr

Future driver assistance systems will require not only information about the current state of a vehicle but also a perception and modelling of the environment. Depending on the complexity of the assistance function it will have a variable scale of details. Safety systems - like automatic breaking - have to meet the highest requirements. The geometric as well as the dynamic parameters of potential collision partners have to be calculated as precisely as possible. Because no state-of-the-art sensor device is able to meet the whole variety of requirements, a sensor data fusion of heterogenous sensors is therefore essential. Here the advantages of radar sensors measuring the dynamic properties of objects can be combined with the advantages of optical sensors measuring their geometric dimension. In order to achieve an optimal fusion, however, different aspects have to be considered. Three of those are covered by this thesis and will be presented after a detailed bibliography: The processing of asynchronous sensor data, an improved sensor model for optical sensors in particular, and finally a fusion architecture for a varying observability of the state vector. This thesis will show different possibilities to modify a Kalman filter architecture in order to stabilize object tracking especially when the perspective is changing - which is typical for overtaking manoeuvres. The advantages of the proposed algorithms will be demonstrated on the basis of real sensor data.

Level thesis:degree_level

thesis.doctoral

Grantor dc:publisher

Universität Oldenburg

Year

2004

Author dc:creator

• Stüker, Dirk

• [Keine Schlagwörter von Autor/in vergeben.]

Repository record source_url

http://oops.uni-oldenburg.de/201

OAI identifier oai:identifier

oai:oops.uni-oldenburg.de:201