157
The whole georeferencing process can be summarized
by following relationship:
r, m (0 = r” (?) + R? (0 ■ k r (0 + R c- - r' (?) J (3)
Fig. 5b: Another view of the relationship btw. rover
and stereoscopic camera frame
While the first step is accomplished by the stereophoto-
grammetric module using only digital image proces
sing and photogrammetric techniques, in the second
step we require not only the knowledge about the
istantaneous position of the vehicle as derived by GPS
receiver, but also the information regarding its direc
tion, i.e. the bearing, respect to the mapping frame.
This is needed to calculate the rotation matrix
R 1 " (t) between rover and mapping frame.
Given the absence of the INS, we have developed an
algorithm to recover the bearing by GPS coordinates
only. Certainly, the final accuracy will be lower than
that achievable by MMS adopting the INS, but anyway
we tried in our research to limit the error position to
enough acceptable level for our mobile mapping
purposes.
Ideally the bearing can be obtained from vehicle trajec
tory, computing the curve that represents it, as result of
GPS coordinates interpolation, and the tangent to the
curve at the point of interest (Fig. 7).
Although the GPS receiver provide us with 3 spatial
coordinates, we are interested in positioning objects on
2D map, therefore we used in calculations only hori
zontal coordinates and only the angle tp between rover
and mapping frame (Fig. 6) is computed through the
algorithm described in next section. The estimates of
that angle are performed in post-processing on the
basis of all collected and double differentially corrected
GPS measurements.
3. COMPUTING THE ROVER S BEARING
In order to estimate the direction of the van for each
collected GPS point we employed the prediction-upda
te strategy of the well known Kalman filter [4].
The adopted kinematic model describes the vehicle
trajectory in terms of state vectors represented by time
dependent positional vectors, as derived from discrete
time measurements:
