Global Positioning System Reference
In-Depth Information
3.
Calculation of measurements
4.
Calculation of the Kalman gain
5.
Calculation of estimations
6.
Calculation of next covariance
7.
Updating of the covariance.
4.2 Bundle block adjustment (BBA) of digital images
The image exterior orientation is determined by the BBA for the mosaicked digital images
where the BBA is a nonlinear least squares optimization method using the tie points of the
inside block (Takagi & Shimoda, 2004). Bundle block configuration increases both the
reliability and accuracy of object reconstruction. An object point is determined by the
intersection of more than two images, which provides local redundancy for gross error
detection and which consequently forms a better intersection geometry (Chen et al., 2003).
Therefore, in this paper, the digital images are set to overlap by more than 50% in the
forward direction, and by more than 30% on each side. The GPS and IMU data obtained in a
previous step allow the automatic setting of tie points in overlapped images and reduce the
time spent searching for tie points by limiting the search area based on the epipolar line. The
epipolar line is the straight line of intersection between the epipolar plane and the image
plane, and it is estimated by the sensor position and attitude, which is derived from
GPS/IMU integration. It connects the image point in one image through the image point in
the next image. Figure 5 shows an image orientation series with tie points that overlap each
other. The image resolution is extremely high (approximately 1.5 cm), so it is easy to detect
small gaps or cracks.
The accuracy of the image orientation (ba) is estimated by comparison with 20 control points
(cp) as shown in Table 3. The average error of the plane is approximately 3 to 6 cm. The
average error of the height is approximately 10 cm. That is, although the BBA is done
automatically, the result is very accurate compared with the differential GPS or the
GPS/IMU integration data, in which the average error is based on GPS accuracy. Moreover,
the processing time is very short. Thus, the BBA's results aid Kalman filtering by initializing
the position and attitude in the next step to acquire a greater accurate trajectory.
Fig. 5. Image orientation with tie points
Search WWH ::
Custom Search