Simultaneous estimation of vehicle dynamics and lane features for road safety applications

The objective of the proposed algorithm is to estimate the horizontal ego motion and the current pitch angle of a road vehicle at each point in time during its movement. The knowledge of these dynamics helps to improve the reliability and stability of a lane recognition algorithm. We realize this with a monochrome camera and conventional odometry data from the vehicle's CAN bus. For a vanishing point estimation a purely image based algorithm is applied. To associate the vanishing point of the image plane and the vehicle pitch angle, knowledge about their geometric relationship is necessary. We show a way to create this relation, to determine the current vehicle pitch angle and, to update the relation itself accordingly. Finally, the results of the data fusion with the supported lane detection are shown.