Improved moon detection radar data processing method

Abstract

The invention discloses an improved moon detection radar data processing method, and solves the technical problems of redundant channel interference, time shift deviation, noise pollution and signal attenuation in the prior art. Comprising the following steps: acquiring high-frequency radar CH-2B channel echo data; removing repeated detection data from the acquired data based on Sobel edge enhancement filtering, sliding window median filtering, adaptive threshold screening and data continuity correction; determining an optimal correction value by calculating a standardized Euclidean distance between two adjacent channels of data, and performing phase correction on the segmented dislocation data; performing signal correction on the data after phase correction, wherein the signal correction comprises band-pass filtering, background removal processing and spherical surface-index composite compensation gain; by counting the evolution law of the signal standard deviation along with the propagation time, the effective signal cut-off time delay is determined, and the time domain boundary of inversion and interpretation is delimited. According to the method, the quality and interpretation reliability of lunar radar data are remarkably improved, and the method is suitable for high-precision detection of the lunar subsurface structure.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and particularly to an improved method for processing lunar radar data. Background Art

[0002] As a core device for detecting the lunar surface and subsurface structures, the lunar radar system usually adopts time-domain pulse radar technology. By transmitting ultra-wideband electromagnetic pulses and receiving reflected signals, it realizes the analysis of the lunar soil thickness distribution and shallow geological structures. For example, the lunar radar system carried by the Chang'e-4 mission adopts a dual-channel (low-frequency CH-1 and high-frequency CH-2) collaborative working mode. Among them, the low-frequency channel (center frequency 60 MHz) is used for detecting the deep lunar crust structure, and the high-frequency channel (center frequency 500 MHz) specializes in the fine stratification analysis of shallow lunar soil. The system consists of a transmitter, a receiver, an antenna array and a control module, and its data acquisition ability directly determines the accuracy and reliability of geological interpretation.

[0003] However, there are significant technical bottlenecks in the existing methods for processing lunar radar data. Specifically, they are as follows:

[0004] (1) Redundant channel interference: Due to the periodic stays during the movement of the lunar rover, multiple sets of repeated detection data may be generated at the same geographical location, resulting in redundant signals in the radar profile and reducing the data validity;

[0005] (2) Time shift deviation: The discontinuous movement and positioning error of the lunar rover are likely to cause misalignment of segmented data, resulting in phase misalignment and affecting the analysis of formation continuity;

[0006] (3) Noise pollution: Residual antenna coupling, shallow multiple reflections and high-frequency random noise will mask the effective signals, especially interfering with the analysis of deep weak reflection layers;

[0007] (4) Energy attenuation: When electromagnetic waves propagate in the lunar soil medium, geometric diffusion and dielectric loss cause the signal energy to attenuate with depth, and the signal-to-noise ratio of deep reflection layers drops significantly.

[0008] In the prior art, although the above problems can be partially alleviated by basic filtering or gain compensation means, there is a lack of a systematic solution. For example, traditional redundant channel elimination methods rely on manual screening, which is inefficient and prone to misjudgment; time shift correction mostly uses a fixed offset, which is difficult to adapt to dynamic positioning errors; noise suppression and energy recovery technologies are usually implemented independently and do not form a hierarchical processing flow. In addition, there is a lack of a quantitative basis for determining the cut-off time delay of effective signals, resulting in blurred inversion time domain boundaries and affecting the accuracy of geological interpretation.

[0009] Therefore, it is urgent to construct a complete set of lunar radar data processing technology systems to systematically improve the data quality and provide reliable support for the high-precision inversion of the lunar subsurface structure. Summary of the Invention

[0010] The object of the present invention is to provide an improved method for processing lunar radar data, so as to solve the technical problems of low data quality and poor interpretation reliability in existing lunar radar data caused by redundant channel interference, time shift deviation, noise pollution, signal attenuation and boundary ambiguity.

[0011] To achieve the above object, the present invention provides the following technical solutions:

[0012] An improved method for processing lunar radar data provided by the present invention includes:

[0013] Obtaining the echo data of the high-frequency radar CH-2B channel;

[0014] Based on Sobel edge enhancement filtering, sliding window median filtering, adaptive threshold screening and data continuity correction for the obtained data, eliminating duplicate detection data;

[0015] By calculating the standardized Euclidean distance between adjacent two-channel data, determining the optimal correction amount, and performing phase correction on the segmented misaligned data;

[0016] Performing signal correction on the phase-corrected data, including band-pass filtering, background removal processing and spherical-exponential composite compensation gain;

[0017] By statistically analyzing the evolution law of the signal standard deviation with the propagation time, determining the effective signal cut-off delay, and delimiting the time domain boundary for inversion and interpretation.

[0018] Further, the step of based on Sobel edge enhancement filtering, sliding window median filtering, adaptive threshold screening and data continuity correction for the obtained data, eliminating duplicate detection data includes:

[0019] Edge enhancement filtering: Performing convolution calculation using a 3×5 Sobel horizontal gradient kernel to enhance the horizontal difference between effective channels;

[0020] Noise suppression: Using [3×5] sliding window median filtering to eliminate residual random noise, retaining the edge sharpness of the effective signal, and smoothing isolated noise points at the same time;

[0021] Adaptive threshold screening: Calculating the difference scores of each channel signal and screening the effective signal channels;

[0022] Data continuity correction: Respectively setting the starting time window and the filling time window to avoid mis-elimination of the effective signal segment.

[0023] Further, the starting time window is marked as the start of the effective segment only when 20 consecutive channels exceed the threshold, and the filling time window allows the existence of a low-score area with a length ≤ 20 channels within the effective segment.

[0024] Furthermore, the signal correction applied to the phase-corrected data includes:

[0025] Band-pass filtering: Design a 30th-order FIR filter using the frequency sampling method, with the passband boundaries being 250 - 750 MHz;

[0026] Background removal processing: Eliminate antenna coupling residues and shallow multiple reflection noises through column-wise mean filtering;

[0027] Spherical-exponential composite compensation gain: Use a composite gain combining spherical compensation and exponential compensation for energy recovery, and construct a spherical diffusion compensation model and an exponential absorption compensation model by setting parameters.

[0028] Furthermore, for the CH-2 high-frequency channel data, the band-pass filtering is as follows: First, remove the invalid data segment of the first 28 ns of each channel signal to eliminate the transceiver antenna delay difference, and then perform the FIR filter processing using the frequency sampling method. The specific algorithm steps are as follows:

[0029] Normalized frequency mapping: Convert the physical frequency to the normalized angular frequency:

[0030]

[0031] where f sample = 3.2 GHz is the sampling frequency, set the passband boundaries f low = 250 MHz and f high = 750 MHz, corresponding to the normalized frequencies:

[0032]

[0033] Frequency response definition: Construct a discrete frequency point vector f and its corresponding amplitude m:

[0034] f = [0, f low，norm ,f low，norm ,f high，norm ,f high，norm ,1](12),

[0035] m = [0, 0, 1, 1, 0, 0](13);

[0036] Transition band optimization: Suppress the Gibbs phenomenon through the ramp factor r = 0.05:

[0037] Δf ramp = r × (f high，norm - f low，norm ) = 0.015625(14),

[0038] The corrected frequency point vector is:

[0039] f′ = [0, 0.140625, 0.171875, 0.453125, 0.484375, 1](15);

[0040] Filter coefficient generation: Design a 30 - order FIR filter using the frequency sampling method:

[0041] Interpolate on a 512 - point uniform grid to obtain the ideal frequency response H ideal (k),

[0042] After complementing symmetry to H ideal Perform the inverse Fourier transform:

[0043]

[0044] Apply a Hamming window to suppress sidelobes:

[0045]

[0046] Time - domain convolution filtering: Convolve the generated filter coefficients h w with the radar signal D:

[0047]

[0048] Furthermore, the spherical diffusion compensation model indicates that the geometric diffusion loss has a linear relationship with the propagation time t, and the compensation function is defined as:

[0049] G sc (t i ) = v·t i (20),

[0050] where t i = i·Δt, Δt = 0.3125 ns is the sampling interval, and v = 0.168 m / ns;

[0051] The exponential absorption compensation model is that when electromagnetic waves propagate in lunar soil medium, the high - frequency attenuation caused by dielectric loss is corrected by exponential gain:

[0052] G exp (t i ) = exp(αt i )(α = πf0tanδ)(21),

[0053] where f0 = 500 MHz is the center frequency and tanδ = 0.005 is the tangent of the loss angle of lunar soil.

[0054] Furthermore, the determination method of the effective signal cut - off delay is:

[0055] Statistically analyze the variation law of the standard deviation of the pre - processed data with the propagation time;

[0056] When the standard deviation enters the steady state after 520 ns, it is determined as the system noise-dominated region.

[0057] Based on the above technical solutions, the embodiments of the present invention can at least produce the following technical effects:

[0058] The improved lunar radar data processing method provided by the present invention constructs a complete full-process processing technology system for high-frequency radar data, and obtains the processed data of the CH-2 channel of the Chang'e-4 lunar radar at 1617.28 m in the lunar subsurface. A hierarchical solution is proposed for the problems existing in the original data, such as redundant channel interference, time shift misalignment, noise pollution, and energy attenuation: First, based on the Sobel edge enhancement and adaptive threshold screening method, the automatic and efficient removal of redundant channels is realized; Secondly, the time shift correction amount is optimized through phase correlation analysis to solve the problem of misalignment of segmented data; Further, the data quality is improved by combining band-pass filtering, background noise suppression, and SEC model gain; Through the statistical analysis of the evolution law of the standard deviation of the radar signal, 520 ns is determined as the cut-off time delay of the effective signal, which delimits a reliable time domain boundary for subsequent inversion and interpretation. Description of the Drawings

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on the structures shown in these drawings.

[0060] Figure 1 is the process framework diagram of the embodiments of the present invention;

[0061] Figure 2 is the original data diagram of the present invention;

[0062] Figure 3 is the data feature diagram after Sobel kernel filtering of the present invention;

[0063] Figure 4 is the median filtering processing effect diagram of the present invention;

[0064] Figure 5 is the inter-channel difference score and threshold determination diagram of the present invention;

[0065] Figure 6 is the schematic diagram of time shift misalignment between data segments of the present invention;

[0066] Figure 7 is the Euclidean distance distribution diagram of adjacent two-channel data under different relative offsets of the present invention;

[0067] Figure 8 is the result diagram after the pre - processing of the present invention;

[0068] Figure 9 is the high - frequency data correction diagram of the lunar - sounding radar of the present invention at 600 - 800m; among them, (a) is the data after pre - processing; (b) is the data after band - pass filtering; (c) is the data after background removal; (d) is the data after SEC gain;

[0069] Figure 10 is the characteristic diagram of the standard deviation of the radar signal evolving with time after 400ns of the present invention, and the gray area represents the noise - stable area. Specific embodiments

[0070] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. In addition, the technical solutions between various embodiments can be combined with each other, but it must be based on the fact that those of ordinary skill in the art can implement them. When the combination of technical solutions appears to be contradictory or unable to be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0071] The object of the present invention is achieved through the following technical solutions:

[0072] As Figure 1 shown, an improved method for processing lunar - sounding radar data includes the following steps:

[0073] Step 1: Obtain the echo data of the high - frequency radar CH - 2B channel;

[0074] It should be noted that the data used is to obtain the original lunar - sounding radar data through the lunar - sounding radar system. The lunar - sounding radar system adopts a dual - channel collaborative working mode, which respectively includes a low - frequency CH - 1 channel and a high - frequency CH - 2 channel.

[0075] Specifically, the basic parameters of the lunar - sounding radar are shown in Table 1 below. The low - frequency CH - 1 channel (center frequency 60MHz) has a detection depth of ≥100m, and the longitudinal resolution is at the meter level, which is suitable for detecting the shallow structure of the lunar crust; the high - frequency CH - 2 channel (center frequency 500MHz) has a longitudinal resolution better than 30cm and a maximum detection depth of 30m, and is dedicated to finely depicting the lunar soil sequence. The high - frequency channel contains two groups of receiving antennas, and usually the B - antenna data is used in the research. The antennas of the two channels adopt a differential layout scheme. CH - 1 uses a top - dipole antenna, and CH - 2 uses a bottom - butterfly antenna, and the installation height is about 0.3m from the lunar surface.

[0076] Table 1 Basic Parameters of the Lunar Radar

[0077]

[0078] The time-resolution characteristics of the pulse signal are achieved through a nanosecond-level sampling interval. The sampling intervals of CH-1 and CH-2 channels are 2.5 ns and 0.3125 ns respectively. This design ensures that the system can not only capture the macroscopic characteristics of deep structures but also resolve the microscopic inhomogeneities of shallow media. After processing such as time-frequency analysis, noise suppression, and migration correction of radar data, the dielectric parameters of lunar regolith can be extracted and the spatial distribution of underground media can be reconstructed.

[0079] The lunar radar data used in this invention are from the echo data of the B antenna of the high-frequency radar of the "Yutu-2" lunar rover of Chang'e-4 released by the Lunar and Planetary Data Release System of the National Astronomical Observatories, Chinese Academy of Sciences. The data acquisition period covers from January 4, 2019, to March 16, 2024, for a total of 65 lunar days.

[0080] In addition, the lunar radar data are divided into a three-layer processing system, namely Level 0 data, Level 1 data, and Level 2 data.

[0081] Specifically, the classification criteria and processing methods are as follows:

[0082] Level 0 data: It includes two sub-levels, Level 0A and Level 0B. Level 0A data are the source packet data formed after basic processing such as frame synchronization, descrambling, and RS decoding of the original signal by the ground station; Level 0B data are further reorganized into data blocks according to the sampling period (20480 ns / channel for CH-1 channel, 640 ns / channel for CH-2B channel) through operations such as multi-station data merging, sorting and duplicate removal, and packet header stripping.

[0083] Level 1 data: Physical quantity conversion processing is completed on the basis of Level 0B. Convert the 16-bit unsigned integer data to a signed integer according to Equation (1):

[0084] X(m) = S(m) - 128×(A + 1) (1),

[0085] where m is the sampling point index, S(m) is the original unsigned data, A represents the number of accumulation times, and the converted X(m) contains the electromagnetic wave phase information.

[0086] Level 2 data: It includes three-level enhancement processing:

[0087] Level 2A: Normalize the Level 1 data (in the range of -128 to 128), and eliminate the influence of variable gain and DC drift through Equation (2):

[0088]

[0089] Where N is the sliding window size (201 points for the low-frequency channel and 101 points for the high-frequency channel), and y(n) is the normalized signal amplitude.

[0090] Level 2B: Fuse the lunar rover positioning information in the Level 2A data, including the attitude parameters provided by the inertial navigation system.

[0091] Level 2C: Apply band-pass filtering to the Level 2B data.

[0092] Among them, the Level 2B product has achieved an accurate association between electromagnetic characteristics and geographical coordinates, and can be directly used for lunar soil tomography and geological interpretation. The high-frequency data of the lunar radar at Level 2B is selected in the present invention, and its vertical resolution reaches 30 cm. Compared with the 10 m resolution of the low-frequency channel (CH-1), it is more suitable for the fine analysis of the millimeter-level structure of the sub-surface of lunar soil. The data resampling interval is 0.3125 ns, which meets the requirements for identifying shallow geological features.

[0093] Step 2: Preprocess the acquired data, including:

[0094] Step 21: Redundant trace elimination: Based on Sobel edge enhancement filtering, sliding window median filtering, adaptive threshold screening, and data continuity correction, eliminate the repeated detection data;

[0095] Since the lunar radar adopts a time-triggered acquisition mode (fixed sampling interval of 0.3125 ns), and the Yutu-2 lunar rover has periodic dwells during its movement on the lunar surface (such as scientific instrument operation or path planning), resulting in multiple repeated detection data at the same geographical location. Such redundant data appears as a time series with horizontal continuity and highly similar waveform features in the radar profile, as Figure 2 shown, and needs to be eliminated through the following processing flow:

[0096] Step 211: Edge enhancement filtering: Perform convolution operation using a 3×5 Sobel horizontal gradient kernel, and its operator matrix is defined as:

[0097]

[0098] This operation enhances the horizontal difference between effective traces through Equation (3):

[0099]

[0100] Where D(i,j) represents the signal amplitude of the jth sampling point in the ith trace. As Figure 3 shown, after the filtering process, the low-gradient features of the redundant traces are effectively suppressed.

[0101] Step 212: Noise suppression: Use [3×5] sliding window median filtering to eliminate the remaining random noise:

[0102]

[0103] As Figure 4 shown, this operation can preserve the edge sharpness of the valid signal while smoothing the isolated noise points.

[0104] Step 213, Adaptive threshold screening: Calculate the difference scores of each channel signal

[0105]

[0106] where N is the number of sampling points per channel. According to the manual screening experience, 2.5 times the median of the difference scores of the full data set is selected as the dynamic threshold. At this time, the results of automatic screening are basically the same as those of manual screening:

[0107] T = 2.5×medianS(i)(7),

[0108] When S(i) < T, it is determined as a redundant channel, as Figure 5 shown.

[0109] Step 214, Data continuity correction: To avoid the misremoval of valid signal segments, set double constraint conditions: Set the start time window and the filling time window respectively to avoid the misremoval of valid signal segments.

[0110] Start time window: Continuously 20 channels exceeding the threshold T are marked as the start of the valid segment;

[0111] Filling time window: Low-score regions with a length ≤ 20 channels are allowed within the valid segment.

[0112] Among the 713,264 channels of data in the original data set, 40,696 channels of valid signals are finally screened out. According to the lunar rover positioning information, the total driving mileage corresponding to the valid data is calculated to be 1,617.28 m, and the average channel spacing is 4.0 cm, meeting the spatial resolution requirements of lunar soil tomography.

[0113] Step 22: Time shift deviation elimination: Determine the optimal correction amount by calculating the standardized Euclidean distance between adjacent two-channel data, and perform phase correction on the segmented misaligned data.

[0114] Specifically, after the redundant channel removal, the data has segmented misalignment due to the discontinuous movement of the lunar rover and positioning errors, as Figure 6 shown. Such time shift deviations can be eliminated by phase correction. For adjacent two-channel data D i and D i+1 , calculate the standardized Euclidean distance within the relative offset range δ∈[-30, 30]:

[0115]

[0116] As shown Figure 7 in the figure, the optimal correction amount is determined by traversing the δ value and a cyclic shift is performed on D i+1 As shown Figure 8 in the figure, the processed radar profile shows that the time shift deviation in the original data has been effectively corrected, and thus the preprocessing process of the 2B-level data is completed.

[0117] Step 3: Perform signal correction on the preprocessed data, including:

[0118] Step 31: Band-pass filtering: Design a 30th-order FIR filter using the frequency sampling method, with the passband boundary being 250 - 750 MHz, to suppress high-frequency noise and DC offset.

[0119] Specifically, the lunar radar signal is subject to antenna coupling delay during transmission and needs to be corrected through zero-time correction and frequency-domain filtering. For the CH-2 high-frequency channel data, first remove the invalid data segment of the first 28 ns of each channel signal to eliminate the difference in transceiver antenna delays. Subsequently, perform finite impulse response (FIR) filter processing using the frequency sampling method. The specific algorithm steps are as follows:

[0120] Step 311: Normalized frequency mapping: Convert the physical frequency to the normalized angular frequency:

[0121]

[0122] where f sample = 3.2 GHz is the sampling frequency. Set the passband boundaries f low = 250 MHz and f high = 750 MHz, corresponding to the normalized frequencies:

[0123]

[0124] Step 312: Frequency response definition: Construct a discrete frequency point vector f and its corresponding amplitude m:

[0125] f = [0, f low，norm， f low，norm f high，norm f high，norm , 1](12),

[0126] m = [0, 0, 1, 1, 0, 0](13).

[0127] Step 313: Transition band optimization: Suppress the Gibbs phenomenon through the ramp factor r = 0.05:

[0128] Δf ramp = r × (f high，norm - f​low，norm ) = 0.015625(14),

[0129] The corrected frequency point vector is:

[0130] f′ = [0, 0.140625, 0.171875, 0.453125, 0.484375, 1](15).

[0131] Step 314, filter coefficient generation: Design a 30 - order FIR filter using the frequency sampling method:

[0132] Interpolate on a 512 - point uniform grid to obtain the ideal frequency response H ideal (k),

[0133] For H ideal After complementing symmetry, perform the inverse Fourier transform:

[0134]

[0135] Apply a Hamming window to suppress sidelobes:

[0136]

[0137] Step 315, time - domain convolution filtering: Convolve the generated filter coefficients h w with the radar signal D:

[0138]

[0139] As Figure 9 a - Figure 9 b shows, after the above - mentioned processing, the DC offset and high - frequency noise in the radar signal are effectively suppressed.

[0140] Step 32, background removal processing: Eliminate antenna coupling residues and shallow multiple - reflection noises based on column - wise mean filtering.

[0141] Specifically, the background noise in the lunar radar signal is mainly composed of antenna coupling residues and shallow multiple reflections, which need to be eliminated through mean filtering. Based on the transmit pulse width (about 2 ns), set the sliding window length to 7 sampling points (corresponding to 2.1875 ns). Perform column - wise mean filtering on the radar profile D (i,j) Implement column - wise mean filtering:

[0142]

[0143] As Figure 9 b - Figure 9 c shows, after processing, the background noise is effectively suppressed.

[0144] Step 33, Spherical-Exponential Composite Compensation Gain (SEC): Combining the geometric diffusion compensation and the medium absorption compensation models to recover the deep signal energy.

[0145] Specifically, since the deep radar signal has a significant reduction in signal-to-noise ratio due to geometric diffusion attenuation and medium absorption attenuation, a composite gain (Spherical Exponential Compensation, SEC) combining spherical compensation (Spherical Compensation, Gsc) and exponential compensation (Exponential Compensation, Gexp) is required for energy recovery. By setting parameters, the following compensation model is constructed:

[0146] Step 331, Spherical diffusion compensation model: It indicates that the geometric diffusion loss has a linear relationship with the propagation time t, and the compensation function is defined as:

[0147] G sc (t i ) = v·t i (20),

[0148] where t i = i·Δt, Δt = 0.3125 ns is the sampling interval, and v = 0.168 m / ns.

[0149] Step 332, Exponential absorption compensation model: When electromagnetic waves propagate in the lunar soil medium, the high-frequency attenuation caused by dielectric loss is corrected by exponential gain:

[0150] G exp (t i ) = exp(αt i )(α = πf0tanδ)(21),

[0151] where f0 = 500 MHz is the center frequency, and tanδ = 0.005 is the tangent of the loss angle of the lunar soil.

[0152] As Figure 9 c- Figure 9 d shows, after applying the SEC gain to the radar signal, the energy of the deep reflection layer is significantly enhanced.

[0153] Step 4: By statistically analyzing the evolution law of the signal standard deviation with the propagation time, determine that 520 ns is the cut-off delay of the effective signal, and delimit the time domain boundary for inversion and interpretation.

[0154] Specifically, the radar signal shows a noise-dominated characteristic in the deep time delay region due to the attenuation of the effective reflection energy. It is necessary to determine the cut-off delay of the effective signal through statistical analysis methods. Based on the variation law of the signal standard deviation with the propagation time, as Figure 10As shown, the standard deviation of the preprocessed full dataset is calculated, and the time delay intervals of 400 - 500 ns and 400 - 520 ns are linearly fitted. The analysis results show that in the 400 - 520 ns interval, the standard deviation decreases linearly with time, reflecting the gradual attenuation of the effective signal components; after 520 ns, the standard deviation enters a stable state, indicating that the radar signal after this time delay is completely dominated by system noise. Based on this, a noise floor threshold of 520 ns is established for the identification of effective signals in subsequent data processing.

[0155] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. An improved lunar radar data processing method, characterized in that: include: Acquire high frequency radar CH-2B channel echo data; The acquired data is filtered based on Sobel edge enhancement, sliding window median filtering, adaptive threshold screening and data continuity correction to eliminate repeated detection data; By calculating the standardized Euclidean distance between two adjacent data, the optimal correction amount is determined and the phase correction is performed on the segmented misaligned data. Implementing signal correction on the phase-corrected data, including bandpass filtering, background removal processing, and spherical-exponential composite compensation gain; By analyzing the evolution law of the signal standard deviation with propagation time, the effective signal cut-off delay is determined and the time domain boundary of inversion and interpretation is defined.

2. The improved lunar radar data processing method according to claim 1, characterized in that: The data acquired is based on Sobel edge enhancement filtering, sliding window median filtering, adaptive threshold screening and data continuity correction, and the repeated detection data is eliminated, including: Edge enhancement filtering: Use 3×5 Sobel horizontal gradient kernel for convolution to enhance the horizontal difference between effective channels; Noise suppression: Use [3×5] sliding window median filtering to eliminate residual random noise, retain the edge sharpness of the effective signal, and smooth out isolated noise points; Adaptive threshold screening: calculate the difference score of each signal channel and screen the effective signal channel; Data continuity correction: Set the start time window and fill time window separately to avoid mistakenly eliminating valid signal segments.

3. The improved lunar radar data processing method according to claim 2, characterized in that: The start time window is when 20 consecutive tracks exceed the threshold value and are marked as the start of a valid segment. The filling time window is when a low-scoring area with a length of ≤20 tracks is allowed to exist in the valid segment.

4. The improved lunar radar data processing method according to claim 1, characterized in that: The performing of signal correction on the phase-corrected data comprises: Bandpass filtering: A 30th-order FIR filter is designed using the frequency sampling method, with a passband boundary of 250-750MHz; Background removal: Antenna coupling residue and shallow multiple reflection noise are eliminated through column-wise mean filtering; Spherical-exponential composite compensation gain: A composite gain combining spherical compensation and exponential compensation is used for energy recovery, and a spherical diffusion compensation model and an exponential absorption compensation model are constructed by setting parameters.

5. The improved lunar radar data processing method according to claim 4, characterized in that: The bandpass filtering is for the CH-2 high frequency channel data. First, the invalid data segment of 28ns before each signal is removed to eliminate the delay difference between the transmitting and receiving antennas. Then, the FIR filter processing of the frequency sampling method is implemented. The specific algorithm steps are as follows: Normalized frequency mapping: Convert physical frequency to normalized angular frequency: where f sample =3.2GHz is the sampling frequency, set the passband boundary f low =250MHz and f high =750MHz, corresponding to the normalized frequency: Frequency response definition: Construct a discrete frequency point vector f and its corresponding amplitude m: f=[0,f low,norm ,f low,norm ,f high，norm ,f high,norm ,1] (12), m＝[0,0,1,1,0,0] (13)； Transition zone optimization: Suppression of the Gibbs phenomenon by means of a ramp factor r = 0.05: △f ramp =r×(f high,norm -f low，norm )=0.015625 (14) The corrected frequency point vector is: f′＝[0,0.140625,0.171875,0.453125,0.484375,1] (15)； Filter coefficient generation: Design a 30th order FIR filter using frequency sampling method: The ideal frequency response H is obtained by interpolating on a 512-point uniform grid. ideal (k) For H ideal Perform an inverse Fourier transform after complement symmetry: Add a Hamming window to suppress side lobes: Time domain convolution filtering: The generated filter coefficient h w Convolution operation with radar signal D:

6. The improved lunar radar data processing method according to claim 4, characterized in that: The spherical diffusion compensation model shows that the geometric diffusion loss is linearly related to the propagation time t, and the compensation function is defined as: G sc (t i )=υ·t i (20), Where t i =i·Δt, Δt=0.3125ns is the sampling interval, v=0.168m / ns; The exponential absorption compensation model uses exponential gain correction to correct the high-frequency attenuation caused by dielectric loss when electromagnetic waves propagate in the lunar soil medium: G exp (t i )=exp(αt i (α=πf0tanδ) (21) In the formula, f0=500MHz is the center frequency, and tanδ=0.005 is the tangent value of the lunar soil loss angle.

7. The improved lunar radar data processing method according to claim 1, characterized in that: The method for determining the effective signal cut-off delay is: The variation of standard deviation of data after statistical preprocessing with propagation time; When the standard deviation enters a stable state after 520ns, it is determined to be the system noise-dominated area.

Images