Moon finding radar horizon tracking method based on dynamic search center

Abstract

The invention discloses a lunar radar horizon tracking method based on a dynamic search center, and solves the problems that an existing method is high in dependence on artificial experience and low in processing efficiency, and the false horizon misjudgment rate is high due to lunar soil non-uniformity interference. Comprising the following steps: performing Hilbert transform on a radar signal to extract an envelope; based on historical horizon data, predicting a dynamic search center of a current channel through Gaussian weighted average and determining a search window; screening local maximum value points in the window as candidate points, and calculating signal intensity, smoothness and edge direction scores of the candidate points; selecting an optimal candidate point as a horizon position through multi-feature linear fusion; and according to the lunar soil interface form and the signal-to-noise ratio dynamic adjustment parameters, tracking and outputting a final horizon path one by one. According to the method, historical path constraints and local signal features are fused, the identification precision and robustness of the lunar soil layered interface in the complex electromagnetic environment are remarkably improved, and the method is suitable for radar data processing of lunar exploration tasks such as Chang'e 4 and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of deep space exploration data processing, and in particular to a lunar radar layer tracking method based on a dynamic search center. Background Art

[0002] Traditional lunar radar layer tracking methods rely primarily on manual experience-based picking or fixed threshold detection mechanisms, which present significant limitations. Manual methods are inefficient when processing large-scale lunar surface exploration data and struggle to meet rapid processing requirements. Unsupervised algorithms are susceptible to random scattering interference caused by lunar soil medium inhomogeneities (such as rock distribution and pore structure), leading to increased layer misjudgment rates. Furthermore, existing algorithms lack the ability to dynamically integrate historical layer trends, are prone to path deviations in areas with sudden interface changes or low signal-to-noise ratios, and struggle to accurately characterize the spatial morphological characteristics of lunar soil stratification interfaces.

[0003] Therefore, a lunar radar layer tracking method is urgently needed to solve the above problems. Summary of the Invention

[0004] The purpose of the present invention is to provide a lunar radar layer tracking method based on dynamic search center to solve the technical problems of existing layer tracking methods, such as strong dependence on manual experience, low processing efficiency, and high false layer misjudgment rate caused by interference from lunar soil heterogeneity.

[0005] To achieve the above objectives, the present invention provides the following technical solutions:

[0006] The present invention provides a lunar radar layer tracking method based on dynamic search center, comprising the following steps:

[0007] Step 1: Perform Hilbert transform on the corrected radar signal to construct an analytical signal, calculate its modulus to extract the signal envelope, and obtain historical layer data;

[0008] Step 2: Based on the historical horizon data extracted by envelope, the search center of the current trace is predicted by Gaussian weighted average. The search window is determined based on the search center and the search radius.

[0009] Step 3: Extract the signal envelope value of the current channel within the search window from the radar data, select the local maximum point set as the candidate point, and calculate the signal strength score, smoothness score and edge direction score for each candidate point;

[0010] Step 4: By linearly combining the signal strength score, smoothness score and edge direction score, the candidate point with the highest comprehensive score is selected as the layer position of the current trace;

[0011] Step 5: According to the lunar soil stratification interface morphology and data signal-to-noise ratio, repeat steps 2 to 4, adjust the search radius, history window length, smoothing factor and edge direction weight, track each path until the end point, and output the final layer path.

[0012] Furthermore, in step 1, the original radar signal contains a high-frequency oscillation component, and its envelope is extracted to highlight the reflected energy distribution. The signal envelope is calculated by the analytical signal modulus value, and the calculation formula is:

[0013]

[0014] Where E(t) represents the signal envelope, s a (t) represents the analytical signal, s(t) represents the radar signal, represents the Hilbert transform operator.

[0015] Furthermore, the specific steps of step 2 are:

[0016] Assume that the current processing track number is j, the history window length n controls the path smoothness, and extracts the previous n tracked positions along the tracking direction. Use Gaussian weight wk ＝ exp(-(3k / n) 2 / 2) Calculate the weighted average position using the formula:

[0017]

[0018] As the search center of the current track j, its physical meaning is to predict the current interface position based on the historical horizon trend; is the search center, l is the search radius, and the search window is determined to be

[0019] Furthermore, the specific steps of step 3 are:

[0020] Detect local maximum point sets within the search window Each candidate point satisfies E(c i )≥E(c i ±1);

[0021] For each c i Calculate three types of features:

[0022] The signal strength score is the ratio of the candidate point envelope value to the maximum envelope value in the window, which quantifies the relative intensity of the reflected energy. The calculation formula is:

[0023]

[0024] Where, E(c i) represents the envelope value of the candidate point, and E(C) is the envelope value within the window;

[0025] The smoothness score is a normalized value of the deviation between the candidate point and the prediction center, reflecting the deviation between the candidate point and the prediction center. The calculation formula is:

[0026]

[0027] The edge direction score is calculated by the difference in the envelope values ​​on both sides of the interface transition direction, and the calculation formula is:

[0028]

[0029] Where, d = ±1 represents the interface transition direction, K = min(l,c i ) Ensure window validity.

[0030] Furthermore, in step 4, the candidate point with the highest comprehensive score is calculated by the following formula:

[0031] S total (c i )=(1-λ-α)S s (c i )+λS g (c i )+αS e (c i ),

[0032] In the formula, λ is the smoothing factor, α is the edge direction weight, and S is selected total The maximum value of c i As the current layer position, the path iterative update is completed.

[0033] Furthermore, in step 5:

[0034] The search radius controls the spatial range of candidate points, increasing when the lunar soil interface is highly undulating and decreasing in flat areas.

[0035] The length of the history window determines the time inertia of the dynamic search center, which increases in continuous sedimentary layers and decreases in abrupt interfaces;

[0036] The smoothing factor adjusts the strength of the path geometry constraint, increasing it in high-noise data and decreasing it in high signal-to-noise ratio areas;

[0037] The edge direction weight controls the directional sensitivity of the interface transition, increasing at the strong reflection edge and decreasing in the fuzzy transition area.

[0038] Furthermore, it also includes preprocessing of radar data. The specific process of the preprocessing is: enhancing the effective signal edge features through Sobel horizontal gradient kernel convolution, combining with sliding window median filtering to suppress random noise; designing an adaptive threshold screening mechanism based on the median of the inter-channel difference score, and combining the start time window and fill time window constraints to achieve accurate removal of redundant channels; using the Euclidean distance minimization criterion to optimize the phase correction amount of adjacent data segments to eliminate the time shift caused by positioning error; finally, effectively recovering the signal energy of the deep reflection layer through signal gain compensation.

[0039] Furthermore, the signal gain compensation adopts a composite gain model combining spherical diffusion compensation and exponential absorption compensation.

[0040] Based on the above technical solution, the embodiments of the present invention can produce at least the following technical effects:

[0041] The dynamic search center-based lunar radar horizon tracking method provided by this invention enables accurate identification of lunar soil horizon interfaces in complex electromagnetic environments. Compared with traditional manual picking methods, this algorithm integrates historical path constraints and local signal characteristics through a Gaussian weighted prediction mechanism, effectively overcoming the problems of strong reliance on manual experience and low processing efficiency. Compared with unsupervised algorithms, its path constraint mechanism can suppress random scattering interference caused by lunar soil inhomogeneities (such as rocks and pores), reducing the rate of false layer misjudgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying any creative work.

[0043] Figure 1 It is a flow chart of the lunar radar horizon tracking method of the present invention;

[0044] Figure 2 is the radar signal and its envelope curve after correction according to the embodiment of the present invention;

[0045] Figure 3 It is the initial parameter layer tracking result of the embodiment of the present invention. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in this field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0047] The technical solution of the present invention is described in detail below with reference to the accompanying drawings.

[0048] The complete process in this embodiment is as follows:

[0049] 1. Preprocessing of lunar radar data

[0050] 1. Eliminate redundant channels and use the following process to eliminate them:

[0051] Edge enhancement filtering: Convolution is performed using a 3×5 Sobel horizontal gradient kernel to enhance the horizontal differences between valid traces.

[0052] Noise suppression: Use [3×5] sliding window median filtering to eliminate residual random noise, retain the edge sharpness of the valid signal, and smooth out isolated noise points;

[0053] Adaptive threshold screening: calculate the difference score of each signal channel and screen the effective signal channel;

[0054] 2. Data continuity correction: Set the start time window and fill time window separately to avoid the mistaken elimination of valid signal segments.

[0055] 2. Implement signal correction on the pre-processed data, including:

[0056] 1. Bandpass filtering: A 30th-order FIR filter is designed using the frequency sampling method, with a passband boundary of 250-750MHz to suppress high-frequency noise and DC offset.

[0057] 2. Background removal: Antenna coupling residue and shallow multiple reflection noise are eliminated based on column-wise mean filtering.

[0058] 3. Spherical-exponential composite compensation gain (SEC): Combines geometric diffusion compensation and dielectric absorption compensation models to restore deep signal energy.

[0059] Specifically, the signal-to-noise ratio (SNR) of deep radar signals is significantly reduced due to geometric diffusion attenuation and dielectric absorption attenuation. Therefore, a composite gain (Spherical Exponential Compensation, SEC) combining spherical compensation (Gsc) and exponential compensation (Gexp) is required for energy recovery. By setting parameters, the following compensation model is constructed:

[0060] Spherical diffusion compensation model: It indicates that the geometric diffusion loss is linearly related to the propagation time t. The compensation function is defined as:

[0061] G sc (t i )=v·t i ,

[0062] Where t i =i·Δt, Δt=0.3125ns is the sampling interval, v=0.168m / ns.

[0063] Exponential absorption compensation model: When electromagnetic waves propagate in the lunar soil medium, the high-frequency attenuation caused by dielectric loss is corrected using exponential gain:

[0064] G exp (t i )=exp(αt i )α=πf0tanδ),

[0065] Where f0=500MHz is the center frequency, and tanδ=0.005 is the tangent value of the lunar soil loss angle.

[0066] 3. Radar signal envelope extraction: Perform Hilbert transform on the corrected radar signal, construct an analytical signal, calculate its modulus value to extract the signal envelope, and obtain historical layer data.

[0067] Specifically, the corrected radar signal s(t) contains phase information, which is expressed as alternating positive and negative values. Its envelope E(t) needs to be extracted to highlight the distribution of reflected energy. The mathematical definition of is:

[0068]

[0069] Where pv represents the Cauchy principal value integral. This transform is equivalent to the convolution of the signal with the impulse response h(t) = 1 / (πt) in the time domain. Its physical essence is to apply a 90-degree phase shift to each frequency component of the signal: the positive frequency component is phase-delayed by π / 2, and the negative frequency component is phase-advanced by π / 2. The frequency domain transfer function can be expressed as:

[0070]

[0071] Construct an analytic signal via the Hilbert transform:

[0072]

[0073] The Fourier spectrum of the analytical signal retains only the positive frequency components and eliminates the negative frequency interference. The envelope E(t) is calculated from the modulus of the analytical signal:

[0074]

[0075] like Figure 2 As shown, the blue curve is the corrected radar signal, and the red curve is the extracted signal envelope. This process converts the oscillating signal into a non-negative envelope curve, eliminating the interference of phase oscillation on maximum value detection.

[0076] 4. Dynamic search center update mechanism: Based on the historical layer data extracted by envelope, the search center of the current track is predicted by Gaussian weighted average. The search window is determined based on the search center and the search radius.

[0077] Specifically, let the current processing track number be j, the history window length n controls the path smoothness, and extract the previous n tracked positions along the tracking direction Using Gaussian weight ω k =exp(-(3k / n) 2 / 2) Calculate the weighted average position using the formula:

[0078]

[0079] As the search center of the current track j, its physical meaning is to predict the current interface position based on the historical horizon trend; is the search center, l is the search radius, and the candidate point range is limited to Avoid path deviation caused by local interference.

[0080] 5. Candidate point detection and feature extraction: The signal envelope value of the current channel within the search window is intercepted from the radar data, and the local maximum point set is selected as the candidate point. The signal strength score, smoothness score and edge direction score are calculated for each candidate point.

[0081] Specifically, detect the local maximum point set within the search window Each candidate point satisfies E(c i )≥E(c i ±1); for each c i Calculate three types of features:

[0082] Signal Strength Score: Quantify the relative intensity of reflected energy, E(c i ) represents the envelope value of the candidate point, and E(C) is the envelope value within the window;

[0083] Smoothness score: The deviation degree between the candidate point and the predicted center;

[0084] Edge direction score: Where d = ±1 represents the interface transition direction (strong → weak or weak → strong), K = min (l, c i ) Ensure window validity.

[0085] 6. Multi-feature fusion decision: By linearly combining the signal strength score, smoothness score and edge direction score, the candidate point with the highest comprehensive score is selected as the layer position of the current track.

[0086] Specifically, the candidate point with the highest comprehensive score is calculated by the following formula:

[0087] S total (c i )=(1-λ-α)S s (c i )+λS g (c i )+αS e (c i ),

[0088] In the formula, λ is the smoothing factor, α is the edge direction weight, and S is selected total The maximum value of c i As the current layer position, the path iterative update is completed.

[0089] 7. According to the lunar soil stratification interface morphology and data signal-to-noise ratio, repeat steps 2-4, adjust the search radius, historical window length, smoothing factor and edge direction weight, track each path until the end point, and output the final layer path.

[0090] In this embodiment, the synergistic effect of the parameter set (l,n,λ,α,d) directly affects the tracking accuracy and robustness. The parameter setting needs to comprehensively consider the target formation morphological characteristics and data signal-to-noise ratio:

[0091] Search radius l: Controls the spatial range of candidate points. When the lunar surface is highly undulating (e.g., at the rim of an impact crater), l should be increased to account for potential offsets. In flat areas (e.g., the mare basalt plains), l can be reduced to prevent tracking of discontinuous surfaces.

[0092] History window n: This determines the time inertia of the dynamic search center. A larger n is suitable for continuous sedimentary layers (such as lunar regolith) to suppress local interference through long-range historical constraints. Abrupt interfaces (such as the lunar regolith-bedrock contact zone) require a smaller n to improve response speed.

[0093] Smoothing factor λ: Adjusts the strength of the path's geometric constraints. Increase λ for high-noise data to enhance path smoothness; decrease λ for high-SNR areas to preserve detailed features.

[0094] Edge weight α and direction d: Controls the sensitivity of the interface transition direction. For strongly reflective edges, α should be increased to locate the edge boundary, and the direction of the edge should be locked using the transition direction d of the boundary signal strength. For fuzzy transition areas, α should be reduced to avoid misjudgment.

[0095] In specific implementation, Figure 3 As shown in Figure 2, the initial parameters of the dynamic search center algorithm are set to l = 20, n = 20, λ = 0, and α = 0. The second layer tracking delay error is 1.1%, but the correlation coefficient is only 0.56. This phenomenon is due to the sensitivity of the Pearson correlation coefficient to low-variance data: when the true layer approximates a straight line, its standard deviation σ x →0 leads to calculation distortion. The delay error of the third layer is 0.4%, and the correlation coefficient is 0.93, which shows that the algorithm can effectively identify the medium-depth layer. The delay error of the fourth layer increases to 2.1%, and the correlation coefficient drops to 0.84, reflecting the influence of deep signal attenuation on tracking accuracy ( Figure 3 ).

[0096] Parameter sensitivity analysis (as shown in Table 1 below) shows that when the search radius l is reduced to 10, the algorithm misses a horizon at a horizontal position of 5 m due to an undersized window, and the error increases to 3.3%. When l is increased to 30, noise interference increases, and the error rises to 2.3%. Adjusting the historical window n within the range of 10-30 has no significant impact on the results, with error fluctuations less than 0.2%. When the smoothing factor λ is increased to 0.3, the correlation coefficient increases to 0.88, improving path continuity. When the edge weight α is set to 0.3 and the direction d = -1 is specified, the delay error decreases to 0.9%, which is the optimal parameter combination.

[0097] Table 1 Parameter sensitivity analysis results

[0098]

[0099] Parameter optimization reduced the fourth-layer tracking delay error from 2.1% to 0.9%, a 57% reduction in relative error. This demonstrates that the introduction of edge directional features effectively suppresses noise interference. Simulation experiments provided a basis for parameter selection for the measured data processing used in this application: λ = 0.3 was used in flat areas to enhance path continuity, while α = 0.3 was used in complex terrain to improve edge recognition.

[0100] The dynamic search center layer tracking algorithm proposed in this paper effectively solves two key problems in traditional methods for lunar soil layer identification by fusing historical path constraints with local signal characteristics. The algorithm uses a Gaussian weighted prediction mechanism to balance historical trends and current signal characteristics, and achieves robust tracking in noisy environments through multi-feature fusion decision-making. Simulation experiments show that when the search radius l = 20 and the historical window n = 20, the algorithm's shallow layer identification error is less than 1%; to address the signal attenuation problem in deep layers (> 170ns), the introduction of edge direction weights (α = 0.3) can reduce the tracking error by 56%. Parameter sensitivity analysis confirms that when the smoothing factor λ > 0.3, the path continuity is improved by 12%, but the ability to respond to interface mutations is weakened. This method provides an adaptive solution for the layer interpretation of Chang'e-4 measured data, and its parameter adjustment mechanism can adapt to the electrical structure differences of different lunar soil geomorphic units.

[0101] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A lunar radar layer tracking method based on dynamic search center, characterized in that: The following steps are involved: Step 1: Perform Hilbert transform on the corrected radar signal to construct an analytical signal, calculate its modulus to extract the signal envelope, and obtain historical layer data; Step 2: Based on the historical horizon data extracted by envelope, the search center of the current trace is predicted by Gaussian weighted average. The search window is determined based on the search center and the search radius. Step 3: Extract the signal envelope value of the current channel within the search window from the radar data, select the local maximum point set as the candidate point, and calculate the signal strength score, smoothness score and edge direction score for each candidate point; Step 4: By linearly combining the signal strength score, smoothness score and edge direction score, the candidate point with the highest comprehensive score is selected as the layer position of the current trace; Step 5: According to the lunar soil stratification interface morphology and data signal-to-noise ratio, repeat steps 2 to 4, adjust the search radius, history window length, smoothing factor and edge direction weight, track each path until the end point, and output the final layer path.

2. The lunar radar layer tracking method based on dynamic search center according to claim 1 is characterized in that: In step 1, the original radar signal contains a high-frequency oscillation component, and its envelope is extracted to highlight the reflected energy distribution. The signal envelope is calculated by the analytical signal modulus value, and the calculation formula is: Where E(t) represents the signal envelope, s a (t) represents the analytical signal, s(t) represents the radar signal, represents the Hilbert transform operator.

3. The lunar radar layer tracking method based on dynamic search center according to claim 1 is characterized in that: The specific steps of step 2 are: Assume that the current processing track number is j, the history window length n controls the path smoothness, and extracts the previous n tracked positions along the tracking direction. Use Gaussian weight w k =exp(-(3k / n) 2 / 2) Calculate the weighted average position using the formula: As the search center of the current track j, its physical meaning is to predict the current interface position based on the historical horizon trend; is the search center, l is the search radius, and the search window is determined to be 4. The lunar radar layer tracking method based on dynamic search center according to claim 1 is characterized in that: The specific steps of step 3 are: Detect local maximum point sets within the search window Each candidate point satisfies E(c i )≥E(c i ±1); For each c i Calculate three types of features: The signal strength score is the ratio of the candidate point envelope value to the maximum envelope value in the window, which quantifies the relative intensity of the reflected energy. The calculation formula is: Where, E(c i ) represents the envelope value of the candidate point, and E(C) is the envelope value within the window; The smoothness score is a normalized value of the deviation between the candidate point and the prediction center, reflecting the deviation between the candidate point and the prediction center. The calculation formula is: The edge direction score is calculated by the difference in the envelope values ​​on both sides of the interface transition direction, and the calculation formula is: Where, d = ±1 represents the interface transition direction, K = min(l,c i ) Ensure window validity.

5. The lunar radar layer tracking method based on dynamic search center according to claim 1 is characterized in that: In step 4, the candidate point with the highest comprehensive score is calculated by the following formula: S total (c i )=(1-λ-α)S s (c i )+λS g (c i )+αS e (c i ), In the formula, λ is the smoothing factor, α is the edge direction weight, and S is selected total The maximum value of c i As the current layer position, the path iterative update is completed.

6. The lunar radar layer tracking method based on dynamic search center according to claim 1 is characterized in that: In step 5: The search radius controls the spatial range of candidate points, increasing when the lunar soil interface is highly undulating and decreasing in flat areas. The length of the history window determines the time inertia of the dynamic search center, which increases in continuous sedimentary layers and decreases in abrupt interfaces; The smoothing factor adjusts the strength of the path geometry constraint, increasing it in high-noise data and decreasing it in high signal-to-noise ratio areas; The edge direction weight controls the directional sensitivity of the interface transition, increasing at the strong reflection edge and decreasing in the fuzzy transition area.

7. The lunar radar layer tracking method based on dynamic search center according to claim 1 is characterized in that: The method also includes preprocessing of radar data. The specific process of the preprocessing is as follows: enhancing the effective signal edge features through Sobel horizontal gradient kernel convolution and suppressing random noise by combining sliding window median filtering; designing an adaptive threshold screening mechanism based on the median of the inter-channel difference score, and combining the start time window and fill time window constraints to achieve accurate removal of redundant channels; using the Euclidean distance minimization criterion to optimize the phase correction amount of adjacent data segments to eliminate the time shift caused by positioning error; and finally, effectively recovering the signal energy of the deep reflection layer through signal gain compensation.

8. The lunar radar layer tracking method based on dynamic search center according to claim 7 is characterized in that: The signal gain compensation adopts a composite gain model combining spherical diffusion compensation and exponential absorption compensation.

Images