September 2019 Full - page 656

geometric deformations mainly affect the camera’s height
positioning accuracy. The GNSS lever arms and IMU boresight
misalignment mainly affect the camera’s height accuracy.
AT Based on GNSS Lever Arms and IMU Boresight Misalignment Calibra-
tion Values and CAM Files
Based on the updated GNSS/IMU observations and calibrated
CAM files, AT is performed on Data B again in various GCP over-
lay schemes. The GCP overlay schemes are listed as follows.
i. Only one GCP is laid in the center of each strip, and there
are four GCPs involved in AT.
ii. One GCP is laid at both ends of each strip, and there are
eight GCPs involved in AT.
iii. One GCP is laid at the beginning, center and end of each
strip, and there are 12 GCPs involved in AT.
iv. Two GCPs are laid at both ends of each strip, and there are
16 GCPs involved in AT. At each end, one GCP is near the
top of the strip, and the other G CP is close to the bottom.
v. Two GCPs are laid at both ends, one GCP is laid in the cen-
ter of the strip, and 20 GCPs are theoretically involved in
AT. Due to image overlay, 18 GCPs actually take part in AT.
vi. Two GCPs are separately laid at the beginning, center and
end of each strip, and 24 GCPs are theoretically involved in
AT. Due to image overlay, 20 GCPs actually take part in AT.
vii. Three GCPs are laid at both ends of each strip, and 24 GCPs
are theoretically involved in AT. At each end, one GCP is
near the top of the strip, one GCP is near the middle, and
another GCP is close to the bottom. Due to image overlay,
20 GCPs actually take part in AT.
viii. Three GCPs are separately laid at the beginning, center
and end of each strip, and 36 GCPs are theoretically in-
volved in AT. Due to image overlay, 30 GCPs actually take
part in AT.
The experimental results using different GCP overlay schemes
are listed in Table 9. For Data B, there are 108 GCPs in the cov-
erage area. All remaining GCPs are kept as checkpoints except
those taking part in AT. The statistical accuracy in Table 9 is
the positioning accuracy of the checkpoints.
Table 9 shows that, based on the updated GNSS/IMU
observations and calibrated CAM files, only one GCP can sig-
nificantly enhance planar DG accuracy. However, the height
accuracy deteriorates, and the DG result is unstable. With
increasing GCPs, both the planar accuracy and the height ac-
curacy improve and tend to be stable. For schemes iv to vii,
the planar accuracy and the height accuracy remain basically
constant, which proves that scheme iv (four corners layout
scheme) can ensure the AT accuracy and that more dense GCP
layout schemes have littleeffect on accuracy refinement.
Using scheme iv and performing AT on Data A, B, C and D,
the experimental results are listed in Table 10.
The experimental results in Table 10 demonstrate that the
updated GNSS/IMU observations and the calibrated CAM files
provide good initial values for AT. With several well-distributed
GCPs (such as scheme iv), DG accuracy can be ensured. The
ground coverage of Data A, B, and C is mostly hills, which occu-
py approximately 60%–70% of the area, and the rest is moun-
tains. The main landform of Data D is the plain urban area.
According to the “Specifications for aerotriangulation of digital
aerophotogrammetry” distributed by the China Surveying and
Mapping Bureau (GB/T.23236-2009, 2009) (Table 11), the direct
Table 9. AT results for Data B using various GCP overlay schemes.
Scheme
Accuracy in X direction (meters) Accuracy in Y direction (meters) Accuracy in Z direction (meters)
Max Min Mean Std Max Min Mean Std Max Min Mean Std
i. 0.877 −0.915 −0.093 0.217 0.987 −0.903 0.139 0.395 5.914 −10.325 −1.638 5.171
ii. 0.831 −0.868 0.067 0.207 0.694 −0.661 −0.067 0.156 2.167 −4.949 −0.648 2.037
iii. 0.829 −0.678 0.064 0.199 0.637 −0.604 −0.028 0.128 1.306 −3.422 −0.351 1.126
iv. 0.485 −0.219 0.035 0.110 0.425 −0.274 −0.004 0.091 0.612 −0.503 0.005 0.196
v. 0.471 −0.205 0.033 0.102 0.416 −0.263 −0.003 0.090 0.609 −0.494 0.004 0.195
vi. 0.462 −0.203 0.032 0.102 0.415 −0.261 −0.003 0.089 0.608 −0.492 0.004 0.194
vii. 0.469 −0.201 0.031 0.101 0.396 −0.257 −0.003 0.089 0.610 −0.481 0.004 0.193
viii. 0.455 −0.197 0.032 0.100 0.405 −0.240 −0.002 0.088 0.603 −0.473 0.003 0.192
Table 10. Re−AT results.
Test
Data
Accuracy in X direction (meters) Accuracy in Y direction (meters) Accuracy in Z direction (meters)
Max Min Mean Std Max Min Mean Std Max Min Mean Std
Data A 0.436 −0.506 −0.002 0.216 0.441 −0.365 0.002 0.217 0.723 −0.778 −0.017 0.227
Data B 0.485 −0.219 0.035 0.110 0.425 −0.274 −0.004 0.091 0.612 −0.503 0.005 0.196
Data C 0.548 −0.284 −0.015 0.102 0.297 −0.242 0.008 0.119 0.459 −0.892 0.014 0.229
Data D 0.607 −0.795 −0.020 0.298 0.713 −0.603 0.003 0.252 0.913 −0.807 0.154 0.271
Table 11. Specifications for aerotriangulation of digital aerophotogrammetry (unit: meters).
Topographic
Mapping Scale
Planar RMSE Height RMSE
Flat Hills Mountains High Mountains Flat Hills Mountains High Mountains
1:1000 0.5 0.5 0.7 0.7 0.28
(0.15) 0.4 0.6 1.2
1:2000 1.0 1.0 1.4 1.4 0.28
(0.15) 0.4 1.0 1.5
656 September 2019 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING