Z. Hu and K. Uchimura (Japan)
Registration problem, Augmented Reality, On-road navigation, camera pose estimation, road model matching
This paper presents dynamical 3D pose estimation and
tracking method for a camera mounted on the moving
passenger car. Tracking of on-board camera’s position
and orientation plays a key role in solving the
registration problem of Augmented Reality (AR).
Especially in our Vision-based Car Navigation System
(VICNAS), a new concept of on-road navigation,
camera pose estimation becomes the most essential
technique for properly superimposing the synthetic
virtual objects like navigation arrows and road
information labels into the real road scene. Traditional
camera pose estimation algorithms always need to have
a fixed and known-structure model as well as the object
depth information to obtain the 3D-2D correlations,
which is not possible in the case of on-road driving.
With the constraints of road structure and on-road
vehicle motion features, this paper presents a new
efficient pose tracking algorithm, which converts the
problem of 3D-2D points correlation to the 2D-2D road
model matching on projective image. We proposed a
multi-lane road model generation method based on the
2D digital road map and the absolute position data
obtained from GPS and Inertial sensors. Additional road
shape lookup table (RSL) concept is also presented in
this paper to calculate the road model matching score.
The algorithms proposed in this paper are validated
with the experimental results from real road test under
different conditions and types of road.