September 2019 Full - page 651

multiple strips. There are 108 control points within the cover-
age area of Data B, with 50 control points in the overlapping
area. There are 85 control points within the coverage area of
Data C, with 13 control points in the overlapping area.
Hegang, Heilongjiang Province, China Tests
In the second project, three groups of blocks were flown in He-
gang, Heilongjiang Province, China on October 14, 17, and 21,
2017. In Figure 9, the area enveloped by the black solid line is
the Hegang area, and the three groups of blocks are all within
that area. Blocks flown on October 17, 2017, were selected for
the experiment. The red translucent area shows the four cross
strips flown at 2600 m height, and the green solid line marked
area shows the four cross strips flown at 1700 m height. At
2600 m height, two strips in the north-south direction were
flown bidirectional, and the overlapping ratio is nearly 70%.
Two strips in the east-west direction are also flown bidirec-
tional, and the overlapping ratio is nearly 70%. The same
flight design was applied to blocks flown at 1700 m height.
The main landform of the area isthe plain urban area. 200
GCPs are collected by means of GPS field measurements. The
GCPs are designed to evenly distribute in the blocks flown at
a height of 2600 m. Due to buildings occlusions and other
ground conditions, GCP spacing is minorly adjusted, but the
overall uniformity is maintained. The blocks flown at 2600
m height (referred to as Data D) are chosen for AT and calibra-
tion. After image matching, 9211 points are reserved for AT
and calibration.
Experimental Results and Analysis
The experimental tests include six parts as follows:
1. Initial uncontrolled DG : Using the initially designed
technical parameters of the camera, uncontrolled DG was
performed to evaluate the GFXJ camera’s accuracy.
2. AT and geometric calibration: The two-step iterative
calibration scheme was applied on four blocks of flight
data to assess the GFXJ ’s accuracy after AT and calibrate the
camera lens and CCD line distortion, the GNSS lever arms
and the IMU boresight misalignment. The GNSS lever arms
and IMU boresight misalignment calibration values and
CAM files were obtained after the experiments.
3. GNSS lever arms and IMU boresight misalignment calibra-
tion: Analysis was performed on the GNSS lever arms and
IMU boresight misalignment calibration values, and the
calibration values were evaluated by DG using updated
GNSS/IMU observations.
4. CAM files calibration: From calibrated CAM files, we can
calculate the viewing angles (Ψx, Ψ y) of each element in
the forward/nadir/backward view CCD line arrays. For
each CCD line, two viewing angle curves will be formed
independently along the CCD line direction and along the
flight direction. A comparison and analysis will be made
on these viewing angle curves.
To evaluate the reliability and accuracy of CAM files, DG
will be implemented again using CAM files and original
GNSS/IMU observations.
5. Redirect geopositioning based on GNSS lever arms and
IMU boresight misalignment calibration values and CAM
files: From the above GNSS lever arms and IMU boresight
misalignment calibration values and CAM files, average
values can be calculated. Using these average values,
GNSS/IMU observations are updated. Using these updated
GNSS/IMU observations and CAM files, DG was performed
repeatedly to evaluate the calibration results.
6. AT based on GNSS lever arms and IMU boresight misalign-
ment calibration valuesand CAM files: The design goal of
the GFXJ camera is to accomplish a 1:1000 scale mapping
task with the support of several GCPs. Based on the GNSS
lever arms and IMU boresight misalignment calibration
values and CAM files, AT is carried out again with several
GCPs to assess the GFXJ ’s camera mapping capabilities.
Initial Uncontrolled DG
Using the GFXJ’s initial technical parameters, including the
nominal viewing angle, focus length, pixel size, and GNSS/
IMU observations, uncontrolled DG was performed to evaluate
the accuracy of the GFXJ camera. The experimental results are
shown in Table 2. The DG accuracy of the GFXJ camera is poor,
and the accuracy for the four datasets is equivalent. There
are great systematic errors in the DG experiment using GNSS/
IMU observations and initial technical camera parameters. The
planar accuracy is approximately 3 to 4 m, the height accu-
racy of Data A, Data B, and Data C is approximately 6 meters,
and that of Data D is nearly 9 m. The values in Table 2 repre-
sent the geometric accuracy of all GCPs, 111 GCPs for Data A,
108 GCPs for Data B, 85 GCPs for Data C, and 200 GCPs for Data
D. Here, Std means the standard deviation. Max means the
maximum value, Min means the minimum value, and Mean
represents the mean of the residuals.
AT and Geometric Calibration
For Data A, 61 GCPs take part in AT & Calibration and 50
reserved GCPs are used as checkpoints for accuracy assess-
ment. The values in Table 3 represent the geometric ac-
curacy of the checkpoints. For Data B, 58 GCPs are used in
AT, and 50 reserved GCPs are used as checkpoints. For Data
C, 50 GCPs are involved in AT & Calibration, and 35 GCPs are
reserved as checkpoints. For Data D, 150 GCPs take part in AT
Table 2. Initial uncontrolled DG results (unit: meters).
Test
Data
Accuracy in X (east-west) direction Accuracy in Y (north-south) direction Accuracy in Z (height) direction
Max Min Mean Std Max Min Mean Std Max Min Mean Std
Data A 6.340 −4.739 1.402 3.897 5.073 −4.980 1.438 3.654 −2.803 −8.932 −6.930 6.487
Data B 6.293 −5.193 0.687 4.033 5.481 −5.956 0.379 3.938 −3.840 −8.255 −6.873 6.573
Data C 6.315 −5.150 0.516 4.623 3.590 −3.729 −0.159 2.456 −3.500 −7.824 −5.755 5.886
Data D 8.312 −6.113 2.611 4.855 7.724 −8.943 1.880 3.719 −6.571 −12.671 −8.928 9.037
Table 3. AT and geometric calibration results (unit: meters).
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 0.443 −0.403 −0.002 0.194 0.308 −0.386 −0.001 0.214 0.692 −0.767 0.016 0.219
Data B 0.455 −0.197 0.032 0.100 0.405 −0.240 −0.002 0.088 0.603 −0.473 0.003 0.192
Data C 0.514 −0.274 −0.004 0.082 0.271 −0.355 0.006 0.080 0.525 −0.796 0.001 0.216
Data D 0.589 −0.724 −0.015 0.252 0.597 −0.542 0.001 0.239 0.859 −0.792 0.147 0.269
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING September 2019 651