Four experimental regions were selected to implement the
combined adjustment of the
ZY3-02
satellite laser altimetry data
and stereo images for mapping without
GCPs
. They are located
in North China, West China, and Central China, as illustrated
in Figure 2.
In the first and second regions, the
ZY3-02
satellite stereo
images and laser altimetry data recorded
at the same time. In the third, the
ZY3-02
satellite images and laser altimetry data
captured in the same region but not syn-
chronized were used. In the fourth region,
some
GLAS
laser data were used to validate
which data are better as elevation control
for
ZY3-02
satellite images. We collected
high-accuracy checkpoints (
CPs
) in the four
regions to validate the result; the
CPs
were
surveyed by
RTK-GPS
, which ensures that
the geo-location accuracy of the
CPs
is bet-
ter than 0.05 m. Detailed information about
the experimental datasets is described in
Table 3, and the experimental regions are
illustrated in Figure 3.
The
GLAS
dataset is derived from the
National Snow and Ice Data Center (
NSIDC
)
and selected according to the criteria to en-
sure an elevation accuracy better than 1.0
m (Gonzalez
et al
., 2010; Li
et al
., 2017).
In addition, atmosphere aerosol and
cloud scatter can contaminate the quality
of the laser points, so not all of them can
be used as elevation control points (Li
et
al
., 2017). In the work described in this
paper, the obvious contaminated
ZY3-02
satellite laser points were deleted accord-
ing to the elevation deviation from exist-
ing public
DSM
datasets AW3D30 (Takaku
et al
., 2016). When the elevation error
between the
ZY3-02
satellite laser data and
the AW3D30 is larger than 5.0 m, which
is the standard deviation of AW3D30, the
laser point will not be used. When the
terrain in the laser footprint point is flat,
the elevation value will be coherent even
if there is a little offset in the horizontal
direction. Moreover, for 1:50 000 scale
mapping, an accuracy of elevation control
better than 1.0 m is acceptable (GB/T,
2008), so we can select the point located
on a small slope, as in Figure 4.
The elevation error caused by the terrain
slope is expressed by the following equation:
Δ
h
≈
Δ
x
tan·
S
(1)
When
Δ
x
is approximately 75.0 m, and if
Δ
h
is less than 1.0 m, then
S
should be no
more than 0.8°. Thus, the laser footprint
point located on a slope less of than 0.8°
can be selected as the elevation control
point for the
ZY3-02
satellite images to con-
duct 1:50 000 scale mapping. According to
Equation 1, the retained laser points will
then be selected based on the terrain slope
less than 0.8° and the roughness less than
1.0 m (GB/T, 2008). The slope value and
the roughness are calculated according to
the existing public
DSM
datasets AW3D30
by the Equations 2 and 3, respectively.
S
S
S
X
Y
=
+
arctan (tan ) (tan ) )
2
2
(2)
ξ
=
−
[
]
=
∑
1
2
1
N
H
H
DSM i
i
i
N
_
(3)
Table 3. Overview of the four test datasets used in this study.
Location Date
Geographic area
Laser
altimetry data
Whether
synchronization
Number
of CPs
Weinan,
Shaanxi
09/08/2016
14/08/2016
[33.97°N,35.31°N],
[108.74°E,110.07°E]
ZY3-02 SLA
Yes
27
Kelamayi,
Xinjiang
13/08/2016
[44.247°N,47.145°N],
[83.536°E,85.233°E]
ZY3-02 SLA
Yes
16
Wutai,
Shanxi
26/07/2016
31/07/2016
[37.15°N,42.05°N],
[113.57°E,115.55°E]
ZY3-02 SLA
No
140
Suizhou,
Hubei
26/07/2016
31/07/2016
20/08/2016
[29.03°N,33.10°N],
[111.21°E,115.27°E]
ZY3-02 SLA +GLAS
No
48
Figure 3. Distribution of
ZY3-02
satellite laser and images in four experimental
regions: (a) Weinan of Shaanxi; (b) Kelamayi of Xinjiang; (c)Wutai of Shanxi;
and (d) Suizhou of Hubei. White circle denotes the checkpoint, gray circle
denotes the
ZY3-02
laser point, and black circle denotes
GLAS
laser point.
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
September 2018
571
