Rock Location and Property Analysis of Lunar Regolith at Chang'E 4 Landing Site Based On Local Correlation and Sem Blance Analysis
Rock Location and Property Analysis of Lunar Regolith at Chang'E 4 Landing Site Based On Local Correlation and Sem Blance Analysis
Technical Note
Rock Location and Property Analysis of Lunar Regolith at
Chang’E‐4 Landing Site Based on Local Correlation and Sem‐
blance Analysis
Hanjie Song 1, Chao Li 2,3, Jinhai Zhang 2,3, Xing Wu 1, Yang Liu 1,* and Yongliao Zou 1
1 State Key Laboratory of Space Weather, National Space Science Center, Chinese Academy of Sciences, Bei‐
jing 100190, China; songhanjie@nssc.ac.cn (H.S.); wuxing@nssc.ac.cn (X.W.); yangliu@nssc.ac.cn (Y.L.);
zouyongliao@nssc.ac.cn (Y.Z.)
2 Key Laboratory of Earth and Planetary Physics, Institute of Geology and Geophysics, Chinese Academy of
Sciences, Beijing 100029, China
3 Innovation Academy for Earth Science, Chinese Academy of Sciences, Beijing 100029, China; super‐
lee@mail.iggcas.ac.cn (C.L.); zjh@mail.iggcas.ac.cn (J.Z.)
* Correspondence: yangliu@nssc.ac.cn
Remote Sens. 2021, 13, 48. https://doi.org/10.3390/rs13010048 www.mdpi.com/journal/remotesensing
Remote Sens. 2021, 13, 48 2 of 15
The LPR aboard on the Yutu‐2 Rover of CE‐4 is a nanosecond imaging radar which
is carrier‐free and operates in the time domain [11], and has two channels (CH‐1 and
CH‐2) with center frequencies at 60 and 500 MHz, respectively [10]. CH‐1 is used to map
the structure of the shallow lunar crust with meter‐level resolution and the CH‐2 is used
to detect the structure of regolith with a depth resolution of 0.3 m, which contains one
transmitting antenna and two receiving antennas of different offsets—namely, CH‐2A
and CH‐2B [12]. The LPR measurements suggest an underestimation of the global lunar
regolith thickness by other methods and reveal a vast volume from the last volcanic
eruption [8]. Yutu‐2 has obtained a 425 m LPR profile along the rover tracks in the first 16
lunar days, displaying clear and complex subsurface structures of the landing area,
which would help us to reveal the fine structure of lunar regolith.
The subsurface velocity of electromagnetic waves is a vital parameter for strati‐
graphic division, rock location estimates, and calculating rock properties including rela‐
tive permittivity, density, and content of FeO and TiO2 during the interpretation of LPR
data. Meanwhile, the location of rocks in regolith can cause diffractions in LPR data, and
the vertex position of the diffractions can help estimate the possible location of rocks.
This is different from the main reflections from the layers, which are useful in strati‐
graphic analysis. Analysis of the rock location and the properties of the lunar regolith can
help to reveal the formation and evolution history of the landing site. Previous studies on
geological stratification and parameter inversion of the lunar regolith were mainly based
on the CH‐2B data only [3,8,12]. For example, Feng et al. proposed a hyperbola fitting
method for radar velocity analysis in the CH‐2B radar‐gram [12]. Dong et al. calculated
the parameters of the regolith by relative reflection amplitudes [13]. Lai et al. acquired
the radar velocity based on the two‐way delay method [14]. Hu et al. proposed an adap‐
tive rock extraction method based on local similarity constraints to achieve the rock lo‐
cation and quantitative analysis for regolith [15]. Dong et al. analyzed the regolith prop‐
erties from the LPR data of Chang’E‐4 based on the 3D velocity spectrum [16]. However,
a complex subsurface structure and interference of noise always result in incomplete,
interlaced, and amplitude‐varying hyperbolas which cause the inaccurate velocity anal‐
ysis of field LPR data [16], and the previous studies cannot reach a satisfied accuracy. In
addition, the determination of the rock location or the vertex position of the diffractions is
the key procedure to obtain a high‐precision diffraction velocity analysis spectrum. Alt‐
hough Zhang et al. used two sets of CH‐2A and CH‐2B data to estimate the electrical
parameters and the iron–titanium content of regolith [17], the subsurface velocity esti‐
mation is not as accurate as expected due to the limitation of their method.
In this paper, two sets of CH‐2A and CH‐2B data are used to estimate velocity and
the relative permittivity; thus, the other properties of lunar regolith including density
and content of FeO and TiO2 can be calculated based on the relative permittivity. Firstly,
we consider the measurement similarity [18,19] between the two sets of CH‐2A and
CH‐2B data based on the local correlation, and develop an extraction method to deter‐
mine the vertex position of the diffractions and estimate the rock location. Secondly, we
apply the normalized velocity spectrum based on the semblance analysis to estimate the
subsurface velocity of the lunar regolith [20]. Finally, we utilize the interpolation method
to obtain the velocity of the subsurface structure and derive the distribution of the rela‐
tive permittivity.
2. Materials and Methods
2.1. LPR Data Processing
The CE‐4 LPR is a dual‐frequency ground penetrating radar (GPR) system, operat‐
ing at 60 MHz (low‐frequency) with a frequency band of 40~80 MHz and 500 MHz
(high‐frequency) with frequency band of 250~750 MHz [11,21]. The CH‐2 radar data were
collected during the first 16 lunar days along the Yutu‐2′s traverse of about 425 m, as
shown in Figure 1, which is consistent with the results acquired by Lin et al. [22].
Remote Sens. 2021, 13, 48 3 of 15
Figure 1. The location of Chang’E‐4 landing site and the traverse of Yutu‐2 rover for the first 16
lunar days. The lunar background was obtained using the imagery by the Lunar Reconnaissance
Orbiter Camera.
To reveal the near‐surface structure of the regolith, we processed and interpreted the
high‐frequency LPR data (CH‐2A and CH‐2B) by excluding the effects of the electro‐
magnetic coupling with the rover’s metallic body. The CH‐2 antenna is mounted at the
bottom of the lunar rover about 0.3 m away from the ground, and the space between two
adjacent antenna elements (one transmitting antenna and two receiving antennae) is
about 0.16 m [23]. The receiving antenna for CH‐2A data is closer to the transmitting an‐
tenna than that of CH‐2B, and thus has a lower signal‐to‐noise ratio. Therefore, we de‐
veloped a series of procedures to process the CH‐2A and CH‐2B data as follows: (1) data
deleting; (2) time delay removal; (3) band‐pass filter application; (4) removal of DC bias;
(5) background removal; (6) mean spatial filter application. Then, we could obtain the
high‐resolution CH‐2A and CH‐2B data with 1958 samples with 0.3125 ns time intervals
and 11,661 traces of 0.0365 m spatial intervals, as shown in Figure 2.
Figure 2. High‐frequency LPR profile along the track of rover Yutu‐2 after the series of data processing. (a) The CH‐2A
data; (b) the CH‐2B data.
2.2. Local Correlation.
Remote Sens. 2021, 13, 48 4 of 15
The processed LPR CH‐2A and CH‐2B data are always affected by noise signifi‐
cantly, and the strong coherence of the noise may cause local maximums [15]. Here, we
introduce local correlation [18] to determine the rock position in order to take advantage
of both LPR CH‐2A and CH‐2B data simultaneously.
The global correlation coefficient between two discrete signals, ai and bi , can be
defined as [18]:
N
a b i i
i 1
, (1)
N N
a b
i 1
2
i
i 1
2
i
which can be represented as two least‐squares inverse based on a linear algebra notation:
= 1 2 , (2)
a b ,
2 1
1 arg min b 1 a = aT a T
(3)
1
b a ,
2 1
2 arg min a 2 b = bT b T
(4)
2
1
c1 12 I S A T A 12 I SA T b, (5)
1
c 2 22 I S BT B 22 I SBT a , (6)
whose elements are given by c i c 1,* i c 2 ,i 1 i N . The local correlation is a
measure of the similarity between two signals [15,19].
2.3. Semblance Analysis
The high‐frequency electromagnetic wave will be reflected, diffracted, and refracted
simultaneously when propagating the interface with discontinuous dielectric properties,
such as boulders, voids, and soil inhomogeneities [24]. The LPR CH‐2A and CH‐2B data
represent the common offset profiles, and the responses analyzed in semblance analysis
are diffraction hyperbola instead of reflection hyperbola in the common midpoint data.
The normal moveout equation was reconfigured for diffraction trajectories in the com‐
mon offset profile, thus the traveltime was approximated as:
Remote Sens. 2021, 13, 48 5 of 15
4 x x0
2
t x t x0
2
, (7)
st2
where x is the position along the common offset profile, x0 is surface position verti‐
cally above the diffracting target, t x0 is the two‐way traveltime to the target when
rms
2
t rms
2
t
interval ,n ,n n , n 1 n 1
, (8)
t n t n 1
where interval ,n is the interval velocity of n‐th layer and rms is the root‐mean‐square
velocity.
Semblance analysis provides a measure of the coherency of energy along trial tra‐
jectories defined by the substitution of trial pairs of st and t x0 , which was con‐
ducted over an aperture of L traces and 2M+1 samples as:
2
iM
L 1
1 j i M k 0
Q j ,k
S , (9)
L iM
L 1
2
j ,k Q
j i M k 0
where L is 11,661 in LPR data, i and j are time sample indices, k is a trace number and
Q j , k is the amplitude of the j‐th sample of the k‐th trace [27]. The selection of M would
not influence the peak location of contours for semblance analysis. A proper selection of
M would gain a better tradeoff between resolution (traveltime and velocity) and noise
compression. Our sensitivity analysis indicates that a value of M = 3 would give a rea‐
sonable resolution and less noise level, so here we set M = 3.
2.4. Property Calculation
The relative permittivity without considering the electric conductivity and magnetic
permeability can be approximated by the propagating velocity as [8]:
2
c
, (10)
where c is the speed of light in vacuum, which is 0.3 m/ns. The relation between the
relative permittivity and density of lunar regolith [28] is:
Remote Sens. 2021, 13, 48 6 of 15
3. Results
3.1. Simulation Data Results
We verify the effectiveness of the proposed method in a synthetic model using the
finite difference time domain (FDTD) method [29], as shown in Figure 3a. The synthetic
model consists of a heterogeneous background and several anomalous rocks with ran‐
dom relative permittivities less than 5.0 [24]. The dominant frequency of a Ricker wave‐
let is 500 MHz, the time windows and the sampling interval are 120 and 0.02 ns, respec‐
tively, and the trace interval is 0.005 m.
The forward simulation results of CH‐2A with the offset of 0.16 m and CH‐2B with
the offset of 0.32 m are illustrated in Figure 3b,c, respectively. The double diffraction
hyperbolas were generated due to the upper and bottom surfaces of the rocks [16]. We
focused on the upper diffraction hyperbola to reduce the error of velocity estimation.
Furthermore, we demonstrate the results by applying the local correlation based on the
frequency wavenumber (FK) fan filtered data in Figure 3d, where the red‐cross indicates
the vertex position of the diffractions. Dong et al. determined the vertex positions of the
diffractions by directly applying a 3D velocity spectrum for the whole data [16], which
would cause a relatively large error because the maximum value position of semblance
analysis would be significantly influenced by the noise. The vertex position of the dif‐
fraction determinations based on local correlation would have higher accuracy than the
direct 3D velocity spectrum. Accurate determination of vertex positions of the diffrac‐
tions is the basis of high‐precision semblance analysis for the velocity.
Remote Sens. 2021, 13, 48 7 of 15
Figure 3. (a) The synthetic regolith model; (b) forward result with offset of 0.16 m; (c) forward re‐
sult with offset of 0.32 m; (d) the local correlation defined in Section 2.2 for the two filtered data.
The red‐cross indicates the vertex position of the diffractions and the white color indicates the
maximum value of local correlation.
Based on the picking positions in Figure 3d, we could scan the stacking velocity
within the local scope of the simulation data. We display the contours from the sem‐
blance analysis for the velocity ranges from 0.1 to 0.3 m/ns within a 4 ns time window in
Figure 4. The vertex positions of the diffractions are shown in the upper right corner of
Remote Sens. 2021, 13, 48 8 of 15
the graph. The stacking velocity and the calculated depth of each rock are listed in Table
1.
Figure 4. Semblance analysis contours of selected diffraction hyperbolas for the simulation model, where the normalized
stacked amplitude threshold of 0.6 was used. The red‐cross indicates the maximum value 1 of the semblance analysis.
Table 1. The estimated results of velocity and depth for the simulation model.
The scanning positions are well consistent with the actual positions of rocks except
the two rocks in the shallow layer (Figure 5). The inconsistency may be due to the high
velocity in the shallow layers which results in flat diffraction hyperbola, thus the accu‐
racy of semblance analysis for velocity is insufficient. Thus, the simulation shows that
our proposed procedures based on local correlation and semblance analysis is effective
to obtain the subsurface parameters.
Figure 5. The scanning location of the rocks in the synthetic model. The red‐cross indicates the
scanning location of the diffractions.
Remote Sens. 2021, 13, 48 9 of 15
3.2. LPR Data Results
The amplitude of echoes in the first ~150 ns (length N=501 with 0.3125 ns time in‐
terval), with a depth of about 10 m for the LPR CH‐2A and CH‐2B data, is displayed in
Figure 6a,b, respectively. The local correlation for the filtered CH‐2A and CH‐2B data is
shown in Figure 6c, where the red‐cross indicates the vertex position of the diffractions
after selection.
Figure 6. The profile of lunar penetrating radar (LPR) data within the region of 0~150 ns. (a) The
CH‐2A data; (b) the CH‐2B data; (c) the local correlation defined in Section 2.2 for the filtered
CH‐2A and CH‐2B data. The red‐cross indicates the vertex position of the diffractions and the
white color indicates the maximum value of local correlation.
The semblance analysis contour of diffraction hyperbolas for the LPR CH‐2B data
was based on the rock location extraction by applying the local correlation, and the re‐
sults are shown in Figure 7. The red‐cross in the contour maps represents the maximum
value of the semblance analysis. The stacking velocity for the rocks was obtained by
picking the maximum value of the semblance analysis, and the calculated depths are
listed in Table 2.
Remote Sens. 2021, 13, 48 10 of 15
Figure 7. Semblance analysis contour of diffraction hyperbolas after screening the LPR CH‐2B data,
where the normalized stacked amplitude threshold of 0.6 was used. The red‐cross indicates the
maximum value of 1 for the semblance analysis.
Table 2. The estimated results of velocity and depth for the LPR data.
Remote Sens. 2021, 13, 48 11 of 15
Meanwhile, we compared the diffraction hyperbola for No.30 and No.34 rocks with
the CH‐2B data. As shown in Figure 8, the white dotted lines represent the diffraction
hyperbola at 302.2930 m and 106.5625 ns for the stacking velocity of 0.142 m/ns (Figure
8a) and at 341.3845 m and 101.5625 ns for the stacking velocity of 0.165 m/ns (Figure 8b),
respectively. The trend of the hyperbola shows great consistency with the CH‐2B data,
which indicates the semblance analysis procedure can obtain the stacking velocity with
high precision.
Remote Sens. 2021, 13, 48 12 of 15
Figure 8. The comparison between the diffraction hyperbola and the CH‐2B data within the local
region. (a) The white dotted lines represent the diffraction hyperbola at 302.2930 m and 106.5625
ns for the stacking velocity of 0.142 m/ns; (b) the white dotted lines represent the diffraction hy‐
perbola at 341.3845 m and 101.5625 ns for the stacking velocity of 0.165 m/ns.
We applied the spline image interpolation method based on the 40 selected rocks to
obtain the stacking velocity of the substructures. The stacking velocity and the
root‐mean‐square velocity approach to the same when the source‐receiver offset ap‐
proaches zero for the isotropic layered model. The interval velocity can be calculated
using the Dix formula as shown in Figure 9a. Subsequently, the estimated relative per‐
mittivity of the subsurface structure along the Yutu‐2 rover’s traverse, as shown in Fig‐
ure 9b, could be obtained based on the interval velocity distribution. The subsurface ve‐
locity and the relative permittivity are important parameters for calculating other phys‐
ical properties. Based on the relative permittivity of the subsurface structure without
considering the echo power attenuation, we could acquire the subsurface properties of
lunar regolith, which are important for geological interpretation.
Remote Sens. 2021, 13, 48 13 of 15
Figure 9. The estimated subsurface velocity and relative permittivity distribution along the Yutu‐2
rover route. (a) The subsurface velocity distribution; (b) the subsurface relative permittivity dis‐
tribution.
4. Discussion
The noise interference and incomplete hyperbola will produce errors in semblance
analysis. Additionally, the region selection of the data will influence the determination
of the vertex position of the diffractions, which will affect the accuracy of semblance
analysis. The determination of rock locations (i.e., the vertex position of the diffractions)
is the most important procedure to obtain reliable semblance analysis results. Our anal‐
ysis indicates that the semblance analysis in the shallow layers of the simulation model
always has a larger bias than that in the deep layers. One possible reason is that the flat
diffraction hyperbola would influence the accuracy of semblance analysis. Thus, a
high‐precision semblance analysis method such as the weighted AB semblance [30] or
new migration methods [31] should be developed to improve the accuracy in the future
studies. Additionally, LPR data processing should consider the local attributes to sepa‐
rate the data with higher accuracies.
We display the forward result with small Gaussian noise (mean is 0 and variance is
0.005) added and the corresponding local correlation results in Figure 10 to evaluate the
validity of the combined method with noise. The results show that small noise has little
influence on the local correlation results, and the combined method is valid to derive
accurate velocity, permittivity, and bulk density distribution.
Figure 10. (a) Forward result; (b) the local correlation results; (c) forward result with Gaussian
noise added; (d) the local correlation results with Gaussian noise added. The red‐cross indicates
the vertex position of the diffractions and the white color indicates the maximum value of local
correlation.
The local correlation and semblance analysis were combined to estimate the veloci‐
ty structure, which can provide a good estimation of abnormal rock location and relative
permittivity. The future multiple‐input multiple‐output radar onboard the lander from
China’s Chang’E‐5 (CE‐5) mission can help the detection of the regolith structure and the
position of the underlying abnormal rocks in the landing site. The combined method by
using local correlation and semblance analysis could be helpful for further processing on
the CE‐5 radar data. Furthermore, the method can be used in the site selection of the In‐
ternational Lunar Research Station (ILRS) [32] and similar missions in the future.
Remote Sens. 2021, 13, 48 14 of 15
5. Conclusions
We derived the similarity between filtered LPR CH‐2A and CH‐2B data based on
the local correlation, which would help to reveal the rock location distribution in the lu‐
nar regolith. Subsequently, we applied the semblance analysis method for the diffraction
hyperbola to obtain the subsurface velocity structure and the relative permittivity dis‐
tribution. The other properties of lunar regolith including bulk density and FeO and TiO2
abundance can be calculated based on the relative permittivity. A procedure that com‐
bined the local correlation and semblance analysis has an advantage over traditional
methods [12–16] in terms of accuracy and robustness in the interpretation of the
high‐frequency LPR data.
The velocity structure and the properties of the subsurface were derived from dif‐
fraction hyperbolas in the high‐frequency LPR data, which could provide a good struc‐
tural constraint for understanding the formation and evolution of the lunar regolith at
the landing site. The combined procedure is an excellent choice for estimating the sub‐
surface velocity and the other properties of the lunar regolith. Compared with the
methods by analyzing Apollo’s samples [1] and by the microwave remote sensing tech‐
niques [3] to derive the physical parameters of lunar regolith, the application of LPR has
great advantages both in terms of the range and the accuracy. Therefore, it is very im‐
portant to develop parameter estimation methods based on the LPR data, especially its
high‐frequency data. The procedure combined with the local correlation and semblance
analysis, proposed in this paper, is of great significance in the interpretation of LPR
high‐frequency data.
Author Contributions: Conceptualization, H.S. and J.Z.; methodology, H.S.; software, H.S. and
C.L.; validation, H.S. and X.W.; formal analysis, Y.Z.; investigation, H.S., X.W. and C.L.; resources,
Y.L. and Y.Z.; data curation, H.S., J.Z. and C.L.; writing—original draft preparation, H.S.; writ‐
ing—review and editing, J.Z. and Y.L.; visualization, H.S.; supervision, J.Z. and Y.L.; project ad‐
ministration, H.S. and Y.L.; funding acquisition, H.S., J.Z. and Y.L. All authors have read and
agreed to the published version of the manuscript.
Funding: This research was funded by the National Key R&D Program of China (grant number:
2020YFE0202100), the National Natural Science Foundation of China (grant number: 11941001), the
pre‐research project on Civil Aerospace Technologies funded by Chinese National Space Admin‐
istration (CNSA) (grant number: D020201 and D020203) and the Beijing Municipal Science and
Technology Commission (grant number: Z191100004319001 and Z181100002918003). J.Z. was
supported by the National Natural Science Foundation of China (grant number: 41941002). Y.L.
was supported by the pre‐research project on Civil Aerospace Technologies funded by Chinese
National Space Administration (CNSA) (grant number: D020101 and D020102) and the Strategic
Priority Research Program of Chinese Academy of Sciences (grant number: XDB 41000000).
Acknowledgments: We thank Lin H.L. for his assistance with the manuscript. We would like to
thank the editor and reviewers for their reviews that improved the content of this paper. The data
reported in this work are archived at http://moon.bao.ac.cn/searchOrder_pdsData.search.
Conflicts of Interest: The funders had no role in the design of the study; in the collection, analyses,
or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
References
1. McKay, D.S.; Heiken, G.; Basu, A.; Blanford, G.; Simon, S.; Reedy, R.; French, B.M.; Papike, J. The lunar regolith. In Lunar
Source‐Book: A User’s Guide to the Moon; Heiken, G., Ed.; Cambridge University Press: New York, NY, USA, 1991; Volume 7,
pp. 285–357.
2. Nakamura, Y.; Latham, G.V.; Dorman, H.J. Apollo Lunar Seismic Experiment—Final summary. J. Geophys. Res. Space Phys.
1982, 87, A117–A123, doi:10.1029/jb087is01p0a117.
3. Fa, W.; Jin, Y. A primary analysis of microwave brightness temperature of lunar surface from Chang‐E 1 multi‐channel
radiometer observation and inversion of regolith layer thickness. Icarus 2010, 207, 605–615, doi:10.1016/j.icarus.2009.11.034.
4. Neish, C.D.; Barnes, J.W.; Sotin, C.; MacKenzie, S.; Soderblom, J.M.; Le Mouélic, S.; Kirk, R.L.; Stiles, B.W.; Malaska, M.J.;
Le Gall, A.; et al. Spectral properties of titan’s impact craters imply chemical weathering of its surface. Geophys. Res. Lett.
2015, 42, 3746–3754, doi:10.1002/2015GL063824.
Remote Sens. 2021, 13, 48 15 of 15
5. Porcello, L.; Jordan, R.; Zelenka, J.; Adams, G.; Phillips, R.; Brown, W.; Ward, S.; Jackson, P. The Apollo lunar sounder
radar system. Proc. IEEE 1974, 62, 769–783, doi:10.1109/PROC.1974.9517.
6. Ono, T.; Kumamoto, A.; Kasahara, Y.; Yamaguchi, Y.; Yamaji, A.; Kobayashi, T.; Oshigami, S.; Nakagawa, H.; Goto, Y.;
Hashimoto, K.; et al. The Lunar Radar Sounder (LRS) Onboard the KAGUYA (SELENE) Spacecraft. Space Sci. Rev. 2010,
154, 145–192, doi:10.1007/s11214‐010‐9673‐8.
7. Xiao, L.; Zhu, P.; Fang, G.; Xiao, Z.; Zou, Y.; Zhao, J.; Zhao, N.; Yuan, Y.; Qiao, L.; Zhang, X.; et al. A young multilayered
terrane of the northern Mare Imbrium revealed by Chang’E‐3 mission. Science 2015, 347, 1226–1229,
doi:10.1126/science.1259866.
8. Zhang, J.; Yang, W.; Hu, S.; Lin, Y.; Fang, G.; Li, C.; Peng, W.; Zhu, S.; He, Z.; Zhou, B.; et al. Volcanic history of the Im‐
brium basin: A close‐up view from the lunar rover Yutu. Proc. Natl. Acad. Sci. USA 2015, 112, 5342–5347,
doi:10.1073/pnas.1503082112.
9. Li, C.; Su, Y.; Pettinelli, E.; Xing, S.; Ding, C.; Liu, J.; Ren, X.; Lauro, S.E.; Soldovieri, F.; Zeng, X.; et al. The Moon’s farside
shallow subsurface structure unveiled by Chang’E‐4 Lunar Penetrating Radar. Sci. Adv. 2020, 6, eaay6898,
doi:10.1126/sciadv.aay6898.
10. Zhang, J.; Zhou, B.; Lin, Y.; Zhu, M.; Song, H.; Dong, Z.; Gao, Y.; Di, K.; Yang, W.; Lin, H.; et al. Lunar regolith and sub‐
structure at Chang’E‐4 landing site in South Pole‐Aitken basin. Nat. Astron. 2020, 1–6, doi:10.1038/s41550‐020‐1197‐x.
11. Fang, G.; Zhou, B.; Ji, Y.; Zhang, Q.; Shen, S.; Li, Y.; Guan, H.; Tang, C.; Gao, Y.; Lu, W.; et al. Lunar Penetrating Radar
onboard the Chang’e‐3 mission. Res. Astron. Astrophys. 2014, 14, 1607–1622, doi:10.1088/1674‐4527/14/12/009.
12. Feng, J.; Su, Y.; Ding, C.; Xing, S.; Dai, S.; Zou, Y. Dielectric properties estimation of the lunar regolith at CE‐3 landing site
using lunar penetrating radar data. Icarus 2017, 284, 424–430, doi:10.1016/j.icarus.2016.12.005.
13. Dong, Z.; Fang, G.; Ji, Y.; Gao, Y.; Wu, C.; Zhang, X. Parameters and structure of lunar regolith in Chang’E‐3 landing area
from lunar penetrating radar (LPR) data. Icarus 2017, 282, 40–46, doi:10.1016/j.icarus.2016.09.010.
14. Lai, J.; Xu, Y.; Zhang, X.; Tang, Z. Structural analysis of lunar subsurface with Chang׳E‐3 lunar penetrating radar. Planet.
Space Sci. 2016, 120, 96–102, doi:10.1016/j.pss.2015.10.014.
15. Hu, B.; Wang, D.; Zhang, L.; Zeng, Z. Rock Location and Quantitative Analysis of Regolith at the Chang’e 3 Landing Site
Based on Local Similarity Constraint. Remote Sens. 2019, 11, 530, doi:10.3390/rs11050530.
16. Dong, Z.; Feng, X.; Zhou, H.; Liu, C.; Zeng, Z.; Li, J.; Liang, W. Properties Analysis of Lunar Regolith at Chang’E‐4 Landing
Site Based on 3D Velocity Spectrum of Lunar Penetrating Radar. Remote Sens. 2020, 12, 629, doi:10.3390/rs12040629.
17. Zhang, L.; Zeng, Z.; Li, J.; Huo, Z.; Wang, K.; Zhang, J.; Huang, L. Parameter Estimation of Lunar Regolith from Lunar
Penetrating Radar Data. Sensors 2018, 18, 2907, doi:10.3390/s18092907.
18. Fomel, S. Local seismic attributes. Geophysics 2007, 72, 29–33, doi:10.1190/1.2437573.
19. Liu, G.; Fomel, S.; Jin, L.; Chen, X. Stacking seismic data using local correlation. Geophysics 2009, 74, V43–V48,
doi:10.1190/1.3085643.
20. Fisher, E.; McMechan, G.A.; Annan, A.P. Acquisition and processing of wide‐aperture ground‐penetrating radar data.
Geophysics 1992, 57, 495–504, doi:10.1190/1.1443265.
21. Su, Y.; Fang, G.; Feng, J.; Xing, S.; Ji, Y.; Zhou, B.; Gao, Y.; Li, H.; Dai, S.; Xiao, Y.; et al. Data processing and initial results of
Chang’e‐3 lunar penetrating radar. Res. Astron. Astrophys. 2014, 14, 1623–1632, doi:10.1088/1674‐4527/14/12/010.
22. Lin, H.; Lin, Y.; Yang, W.; He, Z.; Hu, S.; Wei, Y.; Xu, R.; Zhang, J.; Liu, X.; Yang, J.; et al. New Insight Into Lunar Regolith‐
Forming Processes by the Lunar Rover Yutu‐2. Geophys. Res. Lett. 2020, 47, e2020GL087949, doi:10.1029/2020gl087949.
23. Angelopoulos, M.; Redman, D.; Pollard, W.H.; Haltigin, T.W.; Dietrich, P. Lunar ground penetrating radar: Minimizing
po‐tential data artifacts caused by signal interaction with a rover body. Adv. Space Res. 2014, 54, 2059–2072,
doi:10.1016/j.asr.2013.09.014.
24. Lv, W.; Li, C.; Song, H.; Zhang, J.; Lin, Y. Comparative analysis of reflection characteristics of lunar penetrating radar data
using nu‐merical simulations. Icarus 2020, 350, 113896, doi:10.1016/j.icarus.2020.113896.
25. Ristic, A.V.; Petrovački, D.; Govedarica, M. A new method to simultaneously estimate the radius of a cylindrical object and
the wave propagation velocity from GPR data. Comput. Geosci. 2009, 35, 1620–1630, doi:10.1016/j.cageo.2009.01.003.
26. Dix, C.H. Seismic velocities from surface measurements. Geophysics 1955, 20, 68–86, doi:10.1190/1.1438126.
27. Neidell, N.S.; Taner, M.T. Sembalnce and other coherency measures for multichannel data. Geophysics 1971, 34, 859–891,
doi:10.1190/1.1440186.
28. Carrier, W.D.; Olhoeft, G.R.; Mendell, W. Physical properties of the lunar surface. In Lunar Source‐Book: A User’s Guide to the
Moon; Heiken, G., Ed.; Cambridge University Press: New York, NY, USA, 1991; Volume 9, pp. 475–594.
29. Irving, J.; Knight, R. Numerical modeling of ground‐penetrating radar in 2‐D using MATLAB. Comput. Geosci. 2006, 32,
1247–1258, doi:10.1016/j.cageo.2005.11.006.
30. Ebrahimi, S.; Kahoo, A.R.; Chen, Y.; Porsani, M. A high‐resolution weighted AB semblance for dealing with am‐
pli‐tude‐variation‐with‐offset phenomenon. Geophysics 2017, 82, V85–V93, doi:10.1190/geo2016‐0047.1.
31. Sun, H.; Gao, C.; Zhang, Z.; Liao, X.; Wang, X.; Yang, J. High‐Resolution Anisotropic Prestack Kirchhoff Dynamic Focused
Beam Migration. IEEE Sens. J. 2019, 20, 11753–11760, doi:10.1109/jsen.2019.2933200.
32. Pei, Z.; Liu, J.; Wang, Q.; Kang, Y.; Zou, Y.; Zhang, H.; Zhang, Y.; He, H.; Wang, Q.; Yang, R. Overview of lunar exploration
and International Lunar Research Station. Chin. Sci. Bull. 2020, 65, 2577–2586, doi:10.1360/TB‐2020‐0582. (In Chinese)