Random noise suppression method and device for ground penetrating radar data

Abstract

The invention relates to the technical field of signal processing, in particular to a random noise suppression method and device for ground penetrating radar data, and the method comprises the steps: carrying out the self-similar block matching processing of original ground penetrating radar data, so as to obtain a self-similar block matrix; performing multiple synchronous extrusion transformation on the self-similar block matrix to obtain a multiple synchronous extrusion transformation result three-dimensional matrix; applying low-rank constraint to the multi-synchronous extrusion transformation result three-dimensional matrix to establish a low-rank matrix and a sparse matrix; and superposing the low-rank matrix to obtain a denoising result of the time domain, and performing effective signal reconstruction on the original ground penetrating radar data according to the denoising result of the time domain to generate the ground penetrating radar data after random noise suppression. Therefore, the problems that an existing ground penetrating radar data random noise suppression method is insufficient in adaptability, the signal structure is damaged, the non-stationary processing capacity is weak, and the low-energy target extraction effect is poor are solved.

Description

Technical Field

[0001] This invention relates to the field of signal processing technology, and in particular to a method and apparatus for suppressing random noise in ground-penetrating radar data. Background Technology

[0002] Ground penetrating radar (GPR), a non-destructive testing technology based on the principle of electromagnetic wave propagation, has become one of the core technologies in the field of shallow geological structure detection due to its advantages such as high resolution, high efficiency, and non-invasiveness. By emitting high-frequency electromagnetic waves and receiving the echo signals reflected by the underground medium, it can invert the location, shape, and physical properties of underground targets (such as pipelines, cavities, faults, etc.), and is widely used in various fields such as underground pipeline network inspection, urban infrastructure inspection, and engineering safety assessment.

[0003] In actual detection processes, the quality of ground-penetrating radar (GPR) data directly determines the accuracy of underground target identification. However, due to various factors, a large amount of random noise inevitably gets mixed into the collected data: on the one hand, electromagnetic environmental interference (such as industrial electromagnetic radiation and radio waves) is directly superimposed on the echo signal; on the other hand, the system's own electronic noise (such as receiver thermal noise) and the non-uniform scattering of the underground medium (such as the irregular reflection of gravel strata) further exacerbate noise interference. These random noises exhibit irregular distribution characteristics, which can distort the waveform of the target echo, mask low-energy signals (such as the reflection of shallow small-scale cavities), reduce the data signal-to-noise ratio, and in severe cases, even cause misjudgment of underground structural features.

[0004] To address the aforementioned issues, various random noise suppression methods have been developed in the industry, among which the following categories are widely used: Bandpass filtering: Based on the frequency difference between signal and noise, a filter is designed to retain the target signal frequency band (usually 10MHz-1GHz) while filtering out high-frequency or low-frequency noise. However, this method is only suitable for scenarios where the frequency boundary between signal and noise is clear. For the common problem of "signal and noise frequency band overlap" in ground-penetrating radar signals (such as low-energy target signals mixed with high-frequency noise), it is prone to signal distortion.

[0005] Wavelet thresholding method: This method utilizes the sparsity of the signal in the wavelet domain to perform thresholding on the wavelet coefficients (such as hard thresholding and soft thresholding) to suppress small coefficients corresponding to noise. However, it still relies on the difference in coefficient amplitude between the signal and noise. For ground-penetrating radar signals with strong non-stationarity (such as frequency characteristics that change with the detection depth), it is difficult to balance noise suppression and signal detail preservation.

[0006] The moving average method smooths out noise fluctuations by calculating the mean of pixels within a sliding window of data. This method is simple to implement, but it can blur abrupt changes in signal characteristics (such as steep reflections at target boundaries), and is particularly prone to causing loss of detail in underground structures when processing high-resolution data.

[0007] Singular value decomposition (SVD) is a method that denoises by retaining the signal components corresponding to large singular values, based on the difference between signal and noise in the distribution of singular values ​​in a matrix. However, this method is sensitive to the matrix construction method. When the noise energy is strong, some noise may be misjudged as singular value components of the valid signal, resulting in incomplete denoising.

[0008] Empirical Mode Decomposition (EMD) and its variants decompose a signal into several intrinsic mode functions (EMFs) and distinguish signal and noise components based on the statistical properties of the EMFs. However, while this method is suitable for non-stationary signal processing, it suffers from mode aliasing (i.e., the same EMF contains both signal and noise), making it difficult to accurately separate the target components, especially in low signal-to-noise ratio scenarios.

[0009] Sparse representation-based methods: These methods sparsely encode the signal by constructing an overcomplete dictionary, leveraging the poor sparsity of noise within the dictionary to achieve suppression. While this approach can preserve signal details well, its high computational complexity makes it unsuitable for real-time processing of large-scale ground-penetrating radar data.

[0010] Based on the above, it can be seen that although existing random noise suppression methods for ground-penetrating radar (GPR) data can achieve a certain denoising effect in specific scenarios, they are limited by the limitations of the technical approach and cannot adapt to the complex characteristics of GPR signals such as non-stationarity and spatial correlation. The main shortcomings are as follows: The reliance on a single feature of signal and noise leads to insufficient adaptability. Methods such as bandpass filtering and wavelet thresholding design processing strategies based solely on the differences between signal and noise in a single dimension (frequency or amplitude), failing to address scenarios where their features overlap. For example, when processing low-energy target signals mixed with high-frequency noise, bandpass filtering can cause signal distortion due to frequency band overlap; wavelet thresholding methods lack the ability to adaptively adjust to frequency characteristics that dynamically change with depth, easily resulting in "oversuppression" (loss of effective signal) or "undersuppression" (residual noise).

[0011] Ignoring the spatial correlation of signals leads to structural damage. Methods such as moving average and singular value decomposition do not fully utilize the spatial redundancy of ground penetrating radar data: moving average smooths noise globally, which can blur the abrupt features of underground structures (such as steep reflections at target boundaries); singular value decomposition depends on the matrix construction method, and when the noise energy is strong, it is difficult to distinguish the singular value components of the signal and the noise, and it cannot preserve the continuity of the in-phase axis of adjacent signals, resulting in the destruction of the spatial correlation of underground structures.

[0012] The processing capabilities for non-stationary signals are insufficient, easily leading to mode aliasing or loss of detail. Empirical Mode Decomposition (EMD) and its variants suffer from mode aliasing when processing non-stationary signals, resulting in the simultaneous inclusion of signal and noise in the same intrinsic mode function (IMF). This is particularly problematic in low signal-to-noise ratio scenarios, making it difficult to accurately separate target components. While sparse representation-based methods can preserve details, they rely on the construction of an overcomplete dictionary, resulting in poor adaptability to the dynamic characteristics of ground-penetrating radar signals as the detection environment changes. Furthermore, their high computational complexity fails to meet the demands of real-time processing of large-scale data.

[0013] The ability to extract low-energy target signals is weak. Existing methods generally suffer from insufficient sensitivity to low-energy signals (such as shallow small-scale cavities and weakly reflective interfaces). While fixed thresholds or filtering strategies can suppress noise, they are prone to misclassifying effective low-energy signals as noise components and filtering them out, leading to the loss of detailed subsurface structural information and affecting the accuracy of subsequent geological interpretation.

[0014] The core reason for the above-mentioned defects is that the existing technology has not fully integrated the multiple inherent characteristics of ground penetrating radar data, such as self-similarity, time-frequency sparsity, and low-rank structure. It is difficult to achieve accurate separation of effective signals and random noise in complex noise environments. Therefore, there is an urgent need for a high-fidelity noise suppression method that can comprehensively utilize the multi-dimensional characteristics of data. Summary of the Invention

[0015] This invention provides a method and apparatus for suppressing random noise in ground-penetrating radar (GPR) data, in order to solve the problems of insufficient adaptability, signal structure destruction, weak non-stationary processing capability, and poor low-energy target extraction in existing GPR data random noise suppression methods.

[0016] A first aspect of the present invention provides a method for suppressing random noise in ground-penetrating radar data, comprising the following steps: The original ground-penetrating radar (GPR) data is subjected to self-similar block matching to obtain a self-similar block matrix; the self-similar block matrix is ​​subjected to multiple synchronous squeezing transformation to obtain a three-dimensional matrix of the result; a low-rank constraint is applied to the three-dimensional matrix of the result of the multiple synchronous squeezing transformation to establish a low-rank matrix and a sparse matrix; the low-rank matrices are superimposed to obtain a time-domain denoising result, and the original GPR data is reconstructed effectively based on the time-domain denoising result to generate ground-penetrating radar data with random noise suppression.

[0017] Optionally, the step of performing self-similar block matching processing on the original ground-penetrating radar data to obtain a self-similar block matrix includes: The raw ground-penetrating radar data is divided into multiple reference blocks. Taking each reference block as the center, multiple sub-blocks with similar waveform shapes are searched in its neighborhood, and the multiple sub-blocks are constructed as their corresponding local self-similar block sets. Each local self-similar block set is vectorized to obtain multiple vectorized local self-similar block sets. The multiple vectorized local self-similar block sets are arranged to obtain a self-similar block matrix.

[0018] Optionally, the step of performing multiple synchronous squeezing transformation on the self-similar block matrix to obtain a three-dimensional matrix of the multiple synchronous squeezing transformation result includes: The fundamental time spectrum of each column of signal in the self-similar block matrix is ​​obtained by using short-time Fourier transform; a multi-layer synchronous compression strategy is used to repeatedly compress and relocate the fundamental time spectrum of multiple columns of signals to obtain the multi-synchronous squeezing transformation result, and the multi-synchronous squeezing transformation result is converted into a three-dimensional matrix of multi-synchronous squeezing transformation result.

[0019] Optionally, the step of superimposing the low-rank matrices to obtain a time-domain denoising result, and then performing effective signal reconstruction on the original ground-penetrating radar data based on the time-domain denoising result to generate ground-penetrating radar data with suppressed random noise, includes: The low-rank matrix is ​​superimposed along the frequency direction to obtain the denoising result in the time domain; the vectorized signal corresponding to each reference block in the self-similar block matrix is ​​selected from the denoising result in the time domain to obtain the denoising result of all reference blocks; according to the position of each reference block in the self-similar block matrix, the denoising result of each reference block is placed in the pre-denoised original ground-penetrating radar data to generate the ground-penetrating radar data after random noise suppression.

[0020] A second aspect of the present invention provides a random noise suppression device for ground-penetrating radar data, comprising: The system includes a matching module for performing self-similar block matching on the original ground-penetrating radar data to obtain a self-similar block matrix; a transformation module for performing multiple synchronous squeezing transformation on the self-similar block matrix to obtain a three-dimensional matrix of the result; a constraint module for applying low-rank constraints to the three-dimensional matrix of the result of the multiple synchronous squeezing transformation to establish a low-rank matrix and a sparse matrix; and a reconstruction module for superimposing the low-rank matrices to obtain a time-domain denoising result, and performing effective signal reconstruction on the original ground-penetrating radar data based on the time-domain denoising result to generate ground-penetrating radar data with random noise suppression.

[0021] Optionally, the matching processing module includes: The system comprises: a partitioning unit for dividing the raw ground-penetrating radar data into multiple reference blocks; a search unit for searching multiple sub-blocks with similar waveform shapes in the vicinity of each reference block, and constructing the multiple sub-blocks into their corresponding local self-similar block sets; a vectorization unit for vectorizing each local self-similar block set to obtain multiple vectorized local self-similar block sets; and an arrangement unit for arranging the multiple vectorized local self-similar block sets to obtain a self-similar block matrix.

[0022] Optionally, the transformation module includes: The acquisition unit is used to acquire the fundamental time spectrum of each column of the signal in the self-similar block matrix using short-time Fourier transform; the transformation unit is used to repeatedly compress and reposition the fundamental time spectrum of multiple columns of signals using a multi-layer synchronous compression strategy to obtain the multi-synchronous squeezing transformation result, and convert the multi-synchronous squeezing transformation result into a three-dimensional matrix of the multi-synchronous squeezing transformation result.

[0023] Optionally, the reconstruction module includes: The stacking unit is used to stack the low-rank matrix along the frequency direction to obtain the denoising result in the time domain; the selection unit is used to select the vectorized signal corresponding to each reference block in the self-similar block matrix from the denoising result in the time domain to obtain the denoising result of all reference blocks; the reconstruction unit is used to place the denoising result of each reference block into the pre-denoised original ground-penetrating radar data according to the position of each reference block in the self-similar block matrix to generate the ground-penetrating radar data after random noise suppression.

[0024] A third aspect of the present invention provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the random noise suppression method for ground penetrating radar data as described in the above embodiments.

[0025] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for suppressing random noise in ground-penetrating radar data.

[0026] The random noise suppression method and apparatus for ground penetrating radar data proposed in this invention integrate multiple inherent characteristics of ground penetrating radar data, such as self-similarity, time-frequency sparsity, and low-rank structure. It achieves accurate separation of effective signals and random noise in complex noise environments, improves the method's adaptability to non-stationary signals, enhances the sensitivity to low-energy target signals (such as shallow small-scale cavities and weakly reflective interfaces), solves the problem of low-energy effective signals being misjudged as noise and filtered out in existing methods, and improves the preservation effect of underground fine structure information.

[0027] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0028] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating a method for suppressing random noise in ground-penetrating radar data according to an embodiment of the present invention; Figure 2 This is a schematic diagram of raw ground-penetrating radar data provided according to an embodiment of the present invention; Figure 3 This is a schematic diagram of ground-penetrating radar data after random noise suppression, provided by an embodiment of the present invention. Figure 4 This is a block diagram of a random noise suppression device for ground-penetrating radar data according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention.

[0029] Explanation of reference numerals in the attached figures: 40-Random noise suppression device for ground penetrating radar data, 401-Matching processing module, 402-Transformation module, 403-Constraint module, 404-Reconstruction module, 501-Memory, 502-Processor and 503-Communication interface. Detailed Implementation

[0030] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0031] The random noise suppression method and apparatus for ground-penetrating radar data according to embodiments of the present invention are described below with reference to the accompanying drawings.

[0032] Figure 1 This is a flowchart illustrating a method for suppressing random noise in ground-penetrating radar data provided in an embodiment of the present invention.

[0033] like Figure 1 As shown, the random noise suppression method for this ground-penetrating radar data includes the following steps: In step S101, the original ground-penetrating radar data is subjected to self-similar block matching processing to obtain a self-similar block matrix.

[0034] In some embodiments, self-similar block matching is performed on the raw ground-penetrating radar data to obtain a self-similar block matrix, including: The raw ground-penetrating radar data is divided into multiple reference blocks; Centered on each reference block, search for multiple sub-blocks in its neighborhood that are similar to its waveform shape, and construct the set of local self-similar blocks corresponding to the multiple sub-blocks. Each set of local self-similar blocks is vectorized to obtain multiple vectorized sets of local self-similar blocks; Arrange multiple vectorized sets of local self-similar blocks to obtain a self-similar block matrix.

[0035] In actual implementation, the embodiments of the present invention will collect the raw two-dimensional ground-penetrating radar data. Divided into several non-overlapping small windows Also known as a reference block; with each reference block Centered on, in its neighboring area Internal search is similar to its waveform structure Each sub-block, construct The corresponding set of locally self-similar blocks The waveform structure similarity is evaluated using Euclidean distance to ensure consistency in key features such as amplitude and phase in the matching results. The formula is as follows: , The smaller, the better and The more similar they are.

[0036] Furthermore, for each set of locally self-similar blocks Vectorize the data and then arrange the vectorized set to obtain a self-similar block matrix. ,in, This represents matrix vectorization.

[0037] In step S102, the self-similar block matrix is ​​subjected to multiple synchronous extrusion transformation to obtain a three-dimensional matrix of the result of the multiple synchronous extrusion transformation.

[0038] In some embodiments, a multi-synchronous squash transformation is performed on the self-similar block matrix to obtain a three-dimensional matrix of the multi-synchronous squash transformation result, including: The fundamental time spectrum of each column of signal in the self-similar block matrix is ​​obtained by using short-time Fourier transform; A multi-level synchronous compression strategy is adopted to repeatedly compress and relocate the fundamental time spectrum of multiple signals to obtain the multi-level synchronous squeezing transformation result, and then the multi-level synchronous squeezing transformation result is converted into a three-dimensional matrix of the multi-level synchronous squeezing transformation result.

[0039] In actual implementation, in order to further improve the self-similar block matrix To ensure the separability of effective signals and random noise, this invention introduces a multiple synchronous squeezing transform pair. High-precision time-frequency decomposition significantly improves the instantaneous frequency characterization capability of ground-penetrating radar signals, enhances the separability of signals and noise in the time-frequency domain, and effectively improves the noise suppression effect of subsequent low-rank constraint processing. The specific processing procedure is as follows: The self-similar block matrix is ​​obtained using the short-time Fourier transform. Each column of signals The fundamental time spectrum can be expressed mathematically as follows:

[0040] in, For window functions, Indicates frequency, Indicates time, This represents the time shift factor.

[0041] Furthermore, based on the initial spectrum, a multi-layer synchronous compression strategy is employed to repeatedly compress and relocate the energy within different frequency scales to obtain the synchronous squeezing transformation result, which can be expressed as:

[0042] in, It is the Dirac function.

[0043] By repeating the above synchronous squeezing operation on the fundamental time spectrum of multiple signals, a more concentrated energy representation (i.e., the result of multiple synchronous squeezing transformation) can be obtained, which can be expressed mathematically as follows:

[0044] in, This indicates the number of iterations in the synchronous squeezing transformation. This is the result of multiple synchronous extrusion transformations.

[0045] Furthermore, the self-similar block matrix The three-dimensional matrix of the result of multiple synchronous extrusion transformation .

[0046] In step S103, a low-rank constraint is applied to the three-dimensional matrix of the result of the multiple synchronous extrusion transformation to establish a low-rank matrix and a sparse matrix.

[0047] In actual implementation, in order to transform the three-dimensional matrix from the multi-synchronous extrusion transformation result of the self-similar block matrix... To further separate the main signal components and suppress random noise, this embodiment of the invention introduces a low-rank constraint modeling method. Apply low-rank constraints, Modeled as a low-rank matrix sparse matrix The solution is This can be considered as the effective signal after suppressing random noise. The optimization problem is as follows:

[0048] in, The lower-rank term represents a valid signal. The sparse term represents random noise. It is the Frobeius norm. for Norm, These are the weighting coefficients. Let be the rank of the lower-rank term.

[0049] In step S104, the low-rank matrices are superimposed to obtain the denoising result in the time domain, and the original ground-penetrating radar data is reconstructed effectively based on the denoising result in the time domain to generate ground-penetrating radar data after random noise suppression.

[0050] In some embodiments, low-rank matrices are superimposed to obtain a denoised result in the time domain, and the original ground-penetrating radar data is reconstructed effectively based on the denoised result in the time domain to generate ground-penetrating radar data with suppressed random noise, including: The low-rank matrices are superimposed along the frequency direction to obtain the denoising result in the time domain; In the denoising results in the time domain, the vectorized signal corresponding to each reference block in the similarity block matrix is ​​selected to obtain the denoising results of all reference blocks; According to the position of each reference block in the self-similar block matrix, the denoising result of each reference block is placed in the pre-denoised original ground-penetrating radar data to generate ground-penetrating radar data with random noise suppression.

[0051] In actual implementation, after completing the low-rank constraint modeling and optimization problem solving for each group of self-similar blocks, in order to restore the denoising results to complete ground-penetrating radar profile data, this embodiment of the invention designs an effective signal reconstruction and stitching integration process to ensure the integrity and usability of the processed data. The specific steps are as follows: The low-rank terms to be solved The denoising results are obtained by superimposing along the frequency direction. ; Denoising results in the time domain Select reference blocks from The corresponding vectorized signal And rearrange them as Reference block The denoising results; Denoising results for all reference blocks According to the reference block The location is placed in the pre-denoised raw ground-penetrating radar data matrix In, that is To obtain ground-penetrating radar data after random noise suppression. .

[0052] In summary, the random noise suppression method for ground-penetrating radar data proposed in the embodiments of the present invention has the following beneficial effects: (1) By mining the spatial redundancy of data through self-similar block matching, high-precision time-frequency decomposition is achieved using multiple synchronous squeezing transformation, and effective signals and noise are accurately separated by low-rank constraint modeling. This solves the problem of poor adaptability of fixed-mode filtering methods to non-stationary signals and improves the adaptability to non-stationary signals and the accuracy of noise separation. Experiments show that this method can significantly improve the signal-to-noise ratio in complex noise environments, providing a more reliable data foundation for subsequent geological interpretation. (2) The energy concentration of low-energy signals in the time-frequency domain can be enhanced by multiple synchronous compression transformations, and the low-rank constraint model can preferentially extract the main components of the signal. The combination of the two can effectively capture the reflection characteristics of low-energy targets, solving the problem of insufficient sensitivity of existing technologies to low-energy targets (such as shallow small-scale cavities and weak reflective interfaces), which are easily misjudged as noise. In actual data processing, this method can identify shallow cavities with a diameter of less than 0.5 meters that are difficult to detect by traditional methods, thus improving the accuracy of geological interpretation.

[0053] Next, with reference to the accompanying drawings, a random noise suppression device for ground-penetrating radar data according to an embodiment of the present invention is described.

[0054] Figure 4 This is a block diagram of a random noise suppression device for ground-penetrating radar data provided in an embodiment of the present invention.

[0055] like Figure 4 As shown, the random noise suppression device 40 for the ground penetrating radar data includes: a matching processing module 401, a transformation module 404, a constraint module 403, and a reconstruction module 404.

[0056] The matching processing module 401 performs self-similar block matching on the original ground-penetrating radar data to obtain a self-similar block matrix. The transformation module 404 performs multiple synchronous squeezing transformation on the self-similar block matrix to obtain a three-dimensional matrix of the result. The constraint module 403 applies low-rank constraints to the three-dimensional matrix of the result of the multiple synchronous squeezing transformation to establish a low-rank matrix and a sparse matrix. The reconstruction module 404 superimposes the low-rank matrices to obtain a denoised result in the time domain, and performs effective signal reconstruction on the original ground-penetrating radar data based on the denoised result in the time domain to generate ground-penetrating radar data with random noise suppression.

[0057] In some embodiments, the matching processing module 401 includes: The partitioning unit is used to divide the raw ground-penetrating radar data into multiple reference blocks; The search unit is used to search for multiple sub-blocks with similar waveform shapes in the neighborhood of each reference block, and to construct the multiple sub-blocks into a set of local self-similar blocks corresponding to each reference block. The vectorization unit is used to vectorize each set of local self-similar blocks to obtain multiple vectorized sets of local self-similar blocks. The permutation unit is used to arrange multiple vectorized sets of local self-similar blocks to obtain a self-similar block matrix.

[0058] In some embodiments, the transformation module 404 includes: The acquisition unit is used to acquire the fundamental time spectrum of each column of signal in the self-similar block matrix using short-time Fourier transform; The transformation unit is used to repeatedly compress and reposition the fundamental time spectrum of multiple signals using a multi-layer synchronous compression strategy to obtain the multi-synchronous squeezing transformation result, and then convert the multi-synchronous squeezing transformation result into a three-dimensional matrix of the multi-synchronous squeezing transformation result.

[0059] In some embodiments, the reconstruction module 404 includes: The stacking unit is used to stack low-rank matrices along the frequency direction to obtain the denoising result in the time domain. The selection unit is used to select the vectorized signal corresponding to each reference block in the similarity block matrix from the denoising results in the time domain, so as to obtain the denoising results of all reference blocks; The reconstruction unit is used to place the denoising result of each reference block into the pre-denoised original ground-penetrating radar data according to the position of each reference block in the self-similar block matrix, so as to generate ground-penetrating radar data with random noise suppression.

[0060] It should be noted that the explanation of the above-described embodiment of the random noise suppression method for ground penetrating radar data also applies to the random noise suppression device for ground penetrating radar data in this embodiment, and will not be repeated here.

[0061] The random noise suppression device for ground-penetrating radar data proposed according to embodiments of the present invention has the following beneficial effects: (1) By mining the spatial redundancy of data through self-similar block matching, high-precision time-frequency decomposition is achieved using multiple synchronous squeezing transformation, and effective signals and noise are accurately separated by low-rank constraint modeling. This solves the problem of poor adaptability of fixed-mode filtering methods to non-stationary signals and improves the adaptability to non-stationary signals and the accuracy of noise separation. Experiments show that this method can significantly improve the signal-to-noise ratio in complex noise environments, providing a more reliable data foundation for subsequent geological interpretation. (2) The energy concentration of low-energy signals in the time-frequency domain can be enhanced by multiple synchronous compression transformations, and the low-rank constraint model can preferentially extract the main components of the signal. The combination of the two can effectively capture the reflection characteristics of low-energy targets, solving the problem of insufficient sensitivity of existing technologies to low-energy targets (such as shallow small-scale cavities and weak reflective interfaces), which are easily misjudged as noise. In actual data processing, this method can identify shallow cavities with a diameter of less than 0.5 meters that are difficult to detect by traditional methods, thus improving the accuracy of geological interpretation.

[0062] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0063] The electronic device may include: a memory 501, a processor 502, and a computer program stored on the memory 501 and capable of running on the processor 502.

[0064] When the processor 502 executes the program, it implements the random noise suppression method for ground-penetrating radar data provided in the above embodiments.

[0065] Furthermore, electronic devices also include: Communication interface 505 is used for communication between memory 501 and processor 502.

[0066] The memory 501 is used to store computer programs that can run on the processor 502.

[0067] Memory 501 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0068] If the memory 501, processor 502, and communication interface 505 are implemented independently, then the communication interface 505, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0069] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 505 are integrated on a single chip, then the memory 501, processor 502, and communication interface 505 can communicate with each other through an internal interface.

[0070] Processor 502 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0071] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for suppressing random noise in ground-penetrating radar data.

[0072] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0073] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0074] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0075] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0076] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0077] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0078] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0079] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A random noise suppression method of ground penetrating radar data, characterized by, The method comprises the following steps: performing self-similar block matching processing on the original ground penetrating radar data to obtain a self-similar block matrix; performing multiple synchronous squeezing transformation on the self-similar block matrix to obtain a multiple synchronous squeezing transformation result three-dimensional matrix; applying low-rank constraint to the multiple synchronous squeezing transformation result three-dimensional matrix to establish a low-rank matrix and a sparse matrix; stacking the low-rank matrix to obtain a time domain denoising result, and reconstructing effective signals according to the time domain denoising result to generate random noise suppressed ground penetrating radar data.

2. The random noise suppression method of ground penetrating radar data according to claim 1, wherein, The self-similar block matching processing on the original ground penetrating radar data to obtain a self-similar block matrix comprises: dividing the original ground penetrating radar data into a plurality of reference blocks; searching for a plurality of sub-blocks similar in waveform shape to each reference block in its adjacent area and constructing the plurality of sub-blocks into a corresponding local self-similar block set; vectorizing each local self-similar block set to obtain a plurality of vectorized local self-similar block sets; arranging the plurality of vectorized local self-similar block sets to obtain a self-similar block matrix.

3. The random noise suppression method of ground penetrating radar data according to claim 1, wherein, The multiple synchronous squeezing transformation on the self-similar block matrix to obtain a multiple synchronous squeezing transformation result three-dimensional matrix comprises: obtaining a basic time-frequency spectrum of each column signal in the self-similar block matrix by using short-time Fourier transform; repeatedly compressing and relocating the basic time-frequency spectrum of multiple column signals by using a multi-layer synchronous compression strategy to obtain a multiple synchronous squeezing transformation result, and converting the multiple synchronous squeezing transformation result into a multiple synchronous squeezing transformation result three-dimensional matrix.

4. The random noise suppression method of ground penetrating radar data according to claim 1, wherein, The stacking of the low-rank matrix to obtain a time domain denoising result, and the reconstruction of effective signals according to the time domain denoising result to generate random noise suppressed ground penetrating radar data comprises: stacking the low-rank matrix along the frequency direction to obtain a time domain denoising result; selecting vectorized signals corresponding to each reference block in the self-similar block matrix from the time domain denoising result to obtain denoising results of all reference blocks; placing the denoising results of each reference block in the pre-denoised original ground penetrating radar data according to the positions of each reference block in the self-similar block matrix to generate the random noise suppressed ground penetrating radar data.

5. An apparatus for random noise suppression of ground penetrating radar data, characterized by The method comprises: a matching processing module configured to perform self-similar block matching processing on the original ground penetrating radar data to obtain a self-similar block matrix; a transformation module configured to perform multiple synchronous squeezing transformation on the self-similar block matrix to obtain a multiple synchronous squeezing transformation result three-dimensional matrix; a constraint module configured to apply low-rank constraint to the multiple synchronous squeezing transformation result three-dimensional matrix to establish a low-rank matrix and a sparse matrix; a reconstruction module configured to stack the low-rank matrix to obtain a time domain denoising result, and reconstruct effective signals according to the time domain denoising result to generate random noise suppressed ground penetrating radar data.

6. The random noise suppression apparatus for ground penetrating radar data according to claim 5, wherein, The matching processing module comprises: The division unit is configured to divide original ground penetrating radar data into a plurality of reference blocks. The searching unit is configured to search for a plurality of sub-blocks similar in waveform shape to each reference block in a neighboring area thereof, and to construct the plurality of sub-blocks into a corresponding local self-similar block set. The vectorization unit is configured to vectorize each local self-similar block set to obtain a plurality of vectorized local self-similar block sets. The arrangement unit is configured to arrange the plurality of vectorized local self-similar block sets to obtain a self-similar block matrix.

7. The random noise suppression apparatus for ground penetrating radar data according to claim 5, wherein, The transformation module comprises: The acquisition unit is configured to acquire a basic time-frequency spectrum of each column signal in the self-similar block matrix by using a short-time Fourier transform. The transformation unit is configured to repeatedly compress and relocate the basic time-frequency spectrum of multiple column signals by using a multi-layer synchronous compression strategy to obtain a multi-layer synchronous squeezing transformation result, and convert the multi-layer synchronous squeezing transformation result into a multi-layer synchronous squeezing transformation result three-dimensional matrix.

8. The random noise suppression apparatus for ground penetrating radar data according to claim 5, wherein, The reconstruction module comprises: The superposition unit is configured to superpose the low-rank matrix along a frequency direction to obtain a denoising result in a time domain. The selection unit is configured to select a vectorized signal corresponding to each reference block in the self-similar block matrix from the denoising result in the time domain to obtain a denoising result of all reference blocks. The reconstruction unit is configured to place the denoising result of each reference block in the original ground penetrating radar data after pre-denoising according to a position of each reference block in the self-similar block matrix to generate the random noise suppressed ground penetrating radar data.

9. An electronic device, comprising: The memory, the processor and the computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the random noise suppression method for ground penetrating radar data according to any one of claims 1-4. The program is executed by the processor to implement the random noise suppression method for ground penetrating radar data according to any one of claims 1-4.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, ​

Images