——M
Using the
1al trajectory
Hz machine,
| pixel. This
ssment. The
>d resolution
Variations in
1ft was set
iust be taken
15 scans per
scan lines in
y the PMIIM
-image cross
Yr orientation
the measured
erned by the
ar orientation
> the noise in
ct a 3.6 GSD
dependently
ultiplying by
method must
> flight line.
ixel locations
red trajectory
file was then
3.669 meters.
ırtosis
)84
103
l.
the measured
:anner modi
ns of Table?
1 rate of 1.919
.] locations d
] trajectory 4
.
Paul Pope
inputs. This took approximately 23 minutes to run on the 200 MHz machine. The output point file was then accuracy
assessed by comparing (differencing) it with the actual location point file. The RMSE was 20.961 meters. The
planimetric error statistics are provided in Table 5. A plot of the actual and corrected trajectories is provided in Figure
2
to
Component | Minimum Maximum Median | Mean | Std. Dev. Skewness Kurtosis
AX -116.457 53.987 -3.786 -3.064 | 19.150 -0.1975 0.1365
AY -90.654 96.765 2.654 2.4647 13.001 -0.2758 0.7543
Table 5. Planimetric error statistics, in units of meters, from PMIIM correction of a noisy input trajectory.
- æ ® " s ®
Figure 2. A plot of the actual (blue) and corrected (red) trajectories. The top row of plots (from left to right) are x, y.
and z versus time. The bottom row of plots (from left to right) are roll, pitch, and yaw versus time.
34.3 Neglecting roll and pitch corrections: The six exterior orientation parameters are not all independent.
Planimetric errors induced through roll action can be mimicked through x and/or y motions. Planimetric errors due to
pitch can be also be achieved through x and/or y and z motions. Therefore, perturbations in roll and pitch were set to
zero and the PMIIM method was used to correct the trajectorys of the remaining four exterior orientation parameters.
Note that the number of perturbations per scan line skip value was reduced from 15,625 to 5% = 625. The
"correlate airborne image.pro" code was used to correct the measured trajectory by using the reference image, DEM,
reduced resolution airborne image, reduced resolution scanner model, scan line skip equal to 18, minimum cross
correlation equal to 0.7, and the exterior orientation perturbations of Table 3 (modified as explained previously) as
inputs. This took approximately 30 minutes to run on the 200 MHz machine. This implies a computation rate of 875
trajectory perturbations per minute. The "correct airborne image.pro" code was used to correct the pixel locations of
the full resolution airborne scanner image by using the full resolution scanner model, DEM, and corrected trajectory as
inputs. This took approximately 23 minutes to run on the 200 MHz machine. The output point file was then accuracy
assessed by comparing (differencing) it with the actual location point file. The RMSE was 32.442 meters. The
planimetric error statistics are provided in Table 6.
Component Minimum Maximum Median Mean Std. Dev. Skewness — Kurtosis
AX -126.262 78.222 -1.576 -2.693 23.130 -0.2157 0.1151
AY -90.262 97.403 -3.109 -6.524 28.567 -0.1272 -0.02276
Table 6. Planimetric error statistics, in units of meters, for the case of neglecting roll and pitch corrections.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 737