A method, system, equipment and medium for removing ground object interference from ground penetrating radar

Abstract

This invention belongs to the technical field of ground penetrating radar (GPR). To address the problem that existing GPR systems cannot completely remove ground object interference, this invention proposes a method, system, equipment, and medium for removing ground object interference. The method involves using the Stolt migration method to migrate the acquired GPR data at the air electromagnetic velocity. Then, based on a generative adversarial network (GAN), the ground object reflection energy cluster components of the migrated data are extracted. Next, the Stolt inverse migration method is used to inversely migrate these components at the air electromagnetic velocity. Finally, ground object reflections are removed from the original GPR data. By combining Stolt migration and Stolt inverse migration techniques with a combined GAN, the interference from ground object reflections during GPR detection is effectively and automatically removed.

Description

Technical Field

[0001] This invention belongs to the field of ground penetrating radar technology, and particularly relates to a method, system, device and medium for removing ground object interference in ground penetrating radar. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Ground-penetrating radar (GPR) is a radar system used for underground exploration. Unlike radar designed to detect dynamic targets in the air in a static state, GPR functions by emitting electromagnetic waves towards the ground while in motion and capturing reflections from underground targets. Compared to the underground environment, the difference in dielectric constant between the air environment and surrounding objects is often more significant, and electromagnetic waves attenuate less in the air. Therefore, objects above ground, including trees, lampposts, vehicles, buildings, and power lines, can cause significant reflections or diffractions, resulting in ground object interference noise. This phenomenon is particularly severe in low-frequency and unshielded antenna configurations. Furthermore, air-to-ground coupling in some GPR devices can cause a considerable amount of energy to be radiated into the air, leading to severe ground object reflection interference in vehicle-mounted or UAV-based GPR systems. Ground object reflections are morphologically very similar to those of underground targets, often resulting in misinterpretations for inexperienced engineers. Simultaneously, the high energy of ground object reflections can mask the true underground reflection information, making effective suppression of ground object reflections crucial.

[0004] Currently, there is no mature automated technology to effectively suppress and remove ground object reflections. All existing methods require manual identification and adjustment, making a fully automated process impossible. Furthermore, existing methods primarily target numerical simulation data for interference removal, while real ground-penetrating radar (GPR) data differs significantly from numerically simulated GPR data. In field acquisition, GPR acquisition intervals are typically not strictly consistent, meaning ground object reflection events are not strictly hyperbolic reflections. Due to the limitations of extraction methods, most existing removal methods generally fail to achieve ideal results on field data. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, the present invention provides a method, system, device and medium for removing ground object interference from ground penetrating radar, which can effectively remove interference reflected by ground objects during the ground penetrating radar detection process.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a method for removing ground object interference from ground penetrating radar, comprising:

[0008] Ground penetrating radar data is acquired, and the ground penetrating radar data is offset using the Stolt migration method under the air electromagnetic wave velocity to obtain the focused energy cluster reflected by ground objects;

[0009] For the focused ground object reflected energy cluster, the ground object reflected energy cluster component is extracted using a trained generative adversarial network. The Stolt inverse migration method is used to inversely migrate the ground object reflected energy cluster component under the air electromagnetic wave velocity to obtain the inversely migrated ground object reflected data.

[0010] By subtracting the ground object reflection data after inverse offset from the ground penetrating radar data, ground penetrating radar data without ground object reflection is obtained.

[0011] Secondly, the present invention provides a ground-penetrating radar ground object interference removal system, comprising:

[0012] The offset processing module is configured to: acquire ground penetrating radar data and perform offset processing on the ground penetrating radar data at the air electromagnetic wave velocity using the Stolt offset method to obtain offset data;

[0013] The reverse migration processing module is configured to: for the focused ground object reflected energy cluster, use a trained generative adversarial network to extract the ground object reflected energy cluster components, and use the Stolt reverse migration method to perform reverse migration processing on the ground object reflected energy cluster components under the air electromagnetic wave velocity to obtain the reverse-migrated ground object reflection data.

[0014] The ground object interference module is configured to subtract the ground object reflection data after inverse offset from the ground penetrating radar data to obtain ground penetrating radar data without ground object reflection.

[0015] Thirdly, the present invention provides an electronic device including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0016] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in the first aspect.

[0017] The above one or more technical solutions have the following beneficial effects:

[0018] This invention performs Stolt migration on the acquired ground-penetrating radar data at the air electromagnetic velocity, then extracts the ground object reflection energy cluster components from the migrated data based on a generative adversarial network, and further performs inverse migration on the ground object reflection energy cluster components at the air electromagnetic velocity using the Stolt inverse migration method. Finally, ground object reflections are removed from the original ground-penetrating radar data. By combining Stolt migration and Stolt inverse migration techniques with a joint generative adversarial network, interference from ground object reflections during the ground-penetrating radar detection process is removed, achieving effective and fully automated removal of ground object reflections.

[0019] Advantages of additional aspects 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

[0020] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0021] Figure 1 This is a flowchart of the ground-penetrating radar ground object interference removal method in Embodiment 1 of the present invention;

[0022] Figure 2 This is a schematic diagram of the original radar data collected and after data verification and zero-point drift removal in Embodiment 1 of the present invention;

[0023] Figure 3 This is a schematic diagram of the processing result after performing a stop offset on the original data under the electromagnetic wave velocity in an air environment in Embodiment 1 of the present invention.

[0024] Figure 4 This is a schematic diagram illustrating the processing results of the offset data using the energy cluster extractor of a generative adversarial network in Embodiment 1 of the present invention.

[0025] Figure 5 In Embodiment 1 of the present invention, the electromagnetic wave velocity in an air environment is... Figure 4 A schematic diagram of the results after the extracted data underwent a stop-off inverse shift.

[0026] Figure 6 This is a schematic diagram of the processing results of separating ground feature reflections from the original data in Embodiment 1 of the present invention. Detailed Implementation

[0027] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0028] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0029] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0030] Example 1

[0031] This embodiment discloses a method for removing ground object interference from ground penetrating radar, including:

[0032] Ground penetrating radar data is acquired, and the ground penetrating radar data is offset using the Stolt migration method under the air electromagnetic wave velocity to obtain the offset data;

[0033] For the focused ground object reflected energy cluster, the ground object reflected energy cluster component is extracted using a trained generative adversarial network. The Stolt inverse migration method is used to inversely migrate the ground object reflected energy cluster component under the air electromagnetic wave velocity to obtain the inversely migrated ground object reflected data.

[0034] By subtracting the ground object reflection data after inverse offset from the ground penetrating radar data, ground penetrating radar data without ground object reflection is obtained.

[0035] This embodiment combines Stolt offset and Stolt inverse offset techniques, along with a joint generative adversarial network, to remove surface object reflection interference during ground penetrating radar detection, achieving effective and fully automated removal.

[0036] The following is combined Figure 1 This embodiment provides a detailed description of a ground-penetrating radar (GPR) ground object interference removal method.

[0037] Step 1: Acquire ground penetrating radar data and perform offset processing on the ground penetrating radar data at the air electromagnetic wave velocity using the Stolt offset method to obtain the offset data.

[0038] This embodiment imposes certain requirements on the ground-penetrating radar (GPR) data acquisition process. The Stolt offset method used in this embodiment is based on the fundamental assumption of an explosive reflective interface, which presupposes that the distance between the transmitting and receiving antennas should be zero. Therefore, during GPR data acquisition, the transmitting and receiving antennas should be self-excited and self-receiving antennas, or a GPR antenna deployment mode with a small transmit / receive distance should be adopted. In principle, the transmit / receive distance for high-frequency antennas with a center frequency greater than 100MHz should not exceed 0.5m, and the transmit / receive distance for low-frequency antennas with a center frequency less than or equal to 100MHz should not exceed 1m.

[0039] It is important to note that in processing acquired ground-penetrating radar (GPR) data, accurate travel time and measurement point spacing are crucial for accurately migrating ground feature reflections within the GPR data. Therefore, data verification is necessary during GPR data preprocessing to ensure that all travel time information and measurement point spacing are correct, facilitating the subsequent accurate migration process.

[0040] The collected ground-penetrating radar data also needs to be preprocessed to ensure data quality. Specifically, the collected ground-penetrating radar data is subjected to zero-point drift removal, DC removal, mean removal, deconvolution, and signal enhancement to ensure that the signal has sufficient strength and amplitude.

[0041] like Figure 2 The diagram shown is a schematic of ground-penetrating radar data after data verification, including two parts: ground object reflection and underground reflection.

[0042] High-precision migration is crucial for ensuring accurate restoration and removal of ground feature reflections. In both Stolt and Stolt inverse migrations, interpolation bias in different spatial regions is a significant cause of error. Sine function interpolation is used for spatial mapping during the transformation from the time-depth domain to the frequency-wavenumber domain in both Stolt and Stolt inverse migrations to avoid spurious phase axes and artifacts, and to improve image smoothness.

[0043] The processed ground-penetrating radar data is processed using Stolt offset at the electromagnetic wave velocity in air (0.3 m / ns) to reorient diffracted waves, focus reflected waves, and effectively separate them from other underground reflected signals.

[0044] Specifically, in the Stolt migration process, ground-penetrating radar data with ground object reflection events can be represented as:

[0045]

[0046] in, This represents 2D ground-penetrating radar observation field data collected along horizontal survey lines; ASD and These represent observed ground reflection events and actual underground reflections, respectively.

[0047] Therefore, the Stolt offset process can be expressed as:

[0048]

[0049] in,

[0050] Where c is the speed of light in air. , These represent the wave numbers in the two directions, respectively; It is the original observation field The Fourier transform result, . It is the Stolt offset operator.

[0051] In this process, this embodiment uses a migration speed of 0.3 m / ns. Therefore, the image field It consists of two parts, among which, Represents the image of an object on a real surface; It is the result of the offset of underground reflections, which is considered to be clutter here.

[0052] like Figure 3 As shown, this is the result of processing after Stolt offset under the electromagnetic wave velocity in the air environment, where ground object reflections are repositioned and focused.

[0053] Step 2: For the focused ground object reflected energy cluster, the ground object reflected energy cluster components are extracted using a trained generative adversarial network. The Stolt inverse migration method is used to inversely migrate the ground object reflected energy cluster components under the air electromagnetic wave velocity to obtain the inversely migrated ground object reflected data.

[0054] To perform ground object reflection energy cluster extraction based on generative adversarial networks, a deep learning-based ground-penetrating radar (GPR) migration energy cluster extractor needs to be built, thus requiring an effective deep learning dataset. This necessitates a sufficiently large dataset, a robust generalization method, and data highly similar to real-world scenarios. Therefore, synthetic data is used as the dataset, comprising two parts: a subsurface reflection dataset and a ground object reflection dataset. For the subsurface reflection data, actual long-line GPR data is used. For the ground object reflection data, numerical simulation data is used.

[0055] Specifically, the construction of the underground reflection dataset includes:

[0056] (1) Screening of underground reflection datasets.

[0057] Examine all data in the subsurface reflection dataset and remove abandoned roads and low-quality data. Additionally, delete all data segments containing ground feature reflection events to ensure the accuracy of the dataset.

[0058] (2) Processing and creation of underground reflection datasets.

[0059] The underground reflection dataset used includes long-line ground-penetrating radar data ranging from hundreds of meters to several kilometers. The acquired long-line ground-penetrating radar data should first undergo necessary processing, including data verification, zero-point drift processing, DC removal processing, mean removal processing, deconvolution processing, and signal enhancement processing.

[0060] (3) Expansion of underground reflection dataset.

[0061] Due to the limited availability of existing underground reflectance datasets, a series of methods were employed to extend them. Two data segments of equal length were randomly cut from the long-line ground-penetrating radar data, with the segment lengths randomly set between 50 meters and 100 meters. These two data segments were then overlaid to achieve data augmentation.

[0062] Data augmentation methods were introduced, including adding Gaussian noise with random variance and performing horizontal and phase flipping on the subsurface reflection data.

[0063] The construction of the ground object reflection dataset is as follows: numerical simulation is performed using random target models randomly distributed in the atmospheric background to simulate B-SCAN data of ground object reflection in a 3D scene.

[0064] The random target is created by adding random Berlin noise to random major and minor axis ellipses, thereby establishing a random target model.

[0065] Processing the ground feature reflection dataset: The simulated ground feature reflection data, which is the same length as the underground reflection data, is cropped for subsequent combination and synthesis of the dataset.

[0066] In actual ground-penetrating radar (GPR) data, ground object reflections are often not strictly hyperbolic due to uneven motion or uneven point spacing, which differs from the results of numerical simulations. Therefore, the Stolt migration results of real ground object reflection data are usually not strictly clustered. To make the simulated ground object reflection data more similar to field data, certain data segments are scaled and adjusted to simulate uneven velocity or uneven point spacing. In this process, the range of each scaled data segment is randomly sampled from 2 to 10 m, the scaling factor is randomly sampled from 0.75 to 1.25, and the scaling frequency is randomly sampled between 5 and 15 times.

[0067] The underground hyperbolic reflection dataset was constructed by using random target models randomly distributed in an underground background for numerical simulation to simulate the B-SCAN data of underground reflection in a 3D scene.

[0068] Each subsurface reflection data point was added to a simulated subsurface reflection hyperbola representing the velocity of the subsurface environment. This step expanded the dataset, prevented clutter within the network, and ensured that all reflection hyperbolas were correctly identified.

[0069] The underground reflection dataset, underground hyperbolic reflection data, and simulated random ground feature reflection data obtained after the above processing are randomly combined by directly superimposing them on the amplitude.

[0070] The randomized dataset (i.e., underground reflection dataset, underground hyperbolic reflection data, and simulated random ground object reflection data) is shifted using the electromagnetic wave velocity in air and used as the input to the generative adversarial network (GAN). The corresponding ground object reflection dataset is shifted using the electromagnetic wave velocity in air and used as the output of the generator in the GAN, thus training the GAN.

[0071] In this embodiment, a generative adversarial network is used as the ground penetrating radar offset energy cluster extractor. The generative adversarial network includes a generator and a discriminator. The generator's job is to extract focused ground object reflection data based on the input randomly combined dataset, so that the extracted focused ground object reflection data is indistinguishable from the real focused ground object reflection.

[0072] The discriminator's role is to distinguish between real ground feature reflectance data and focused ground feature reflectance data generated by the generator, thereby improving the discrimination capability. The generator and discriminator compete with each other, with both networks iterating to improve performance.

[0073] The generator takes the input data and uses three dilated convolutions to transform it into feature maps of three different scales. Then, an ensemble residual convolution block processes the smallest scale feature map and fuses it with the feature maps of other scales. The fused feature map is further processed by the residual convolution block. Finally, the offset ground feature reflection data can be generated.

[0074] For the feature fusion part in the generative adversarial network, the attention score from one feature map to another is first calculated and saved; then, multiple convolution operations are performed on one of the feature maps to obtain two modulation parameters a and b; the trained modulation parameters are then embedded into the other feature map.

[0075] The constructed generative adversarial network is trained using the training dataset. The Adam optimizer is used, and the learning rate starts from 0.001 and decays exponentially. The number of iterations during training should be no less than 150,000 to obtain a trained ground-penetrating radar offset energy cluster extractor.

[0076] After offsetting the actual ground-penetrating radar data, normalization and standardization were performed. Data of different sizes were scaled proportionally to the data range of -1 to 1, and first-order interpolation was used to interpolate the data to a standard 512×512 two-dimensional data size. The standardized data was used as the test set data input to the Generated Adversarial Network Extractor. After processing, the extracted data was obtained. The extracted data is the offset result of ground-penetrating radar data with only ground object reflections at the electromagnetic wave velocity in the air.

[0077] When extracting the ground feature reflection energy cluster components from the offset data using a ground-penetrating radar offset energy cluster extractor, the background values ​​are typically non-zero and contain weak clutter, which may affect subsequent amplitude alignment. Therefore, data elements with absolute values ​​less than 5% of the maximum absolute value should be set to zero to remove all low-energy background data.

[0078] like Figure 4 As shown, the processing results of the offset data using a ground-penetrating radar offset energy cluster extractor based on generative adversarial networks are shown. The offset signal reflected by ground objects is extracted, and the offset signal reflected from underground is filtered out.

[0079] Before entering the ground-penetrating radar (GPR) offset energy cluster extractor, the offset data is scaled to 512×512. Therefore, the scale of the extracted ground feature reflectance energy cluster components needs to be reconstructed to ensure consistency between the inverse offset results and the original GPR data. The two-dimensional data output from the GPR offset energy cluster extractor is reconstructed to the scale of the original GPR data using first-order interpolation.

[0080] Before entering the ground-penetrating radar (GPR) offset energy cluster extractor, the offset data is scaled to a value between -1 and 1. Therefore, the amplitude energy of the extracted ground feature reflection energy cluster components needs to be reconstructed to ensure consistency between the inverse offset results and the original GPR data. Energy alignment is achieved by replacing the amplitudes of all extracted ground feature reflection energy clusters with the amplitudes of the original GPR data. During this process, the background-removed extracted data serves as a mask for data recovery.

[0081] The extracted ground feature reflectance energy cluster components are processed using Stolt inverse migration at an electromagnetic wave velocity in air, such as 0.3 m / ns, to restore all ground feature reflectances to their original state. This inverse migration process can be represented as:

[0082]

[0083] in, Indicates the frequency of the radar signal. It can be replaced by the following equation, These represent the wave number in the x-direction and the wave number in the z-direction, respectively. Represents the ASD data after Fourier transform:

[0084]

[0085] Where c is the speed of light in air, 0.3 m / ns. This is the Fourier transform result of focused ground object reflections. After offset removal, an observation field containing only the original ground object reflection events can be obtained.

[0086] High-precision migration and inverse migration are crucial for ensuring accurate reconstruction of ground feature reflections. In inverse migration, data boundaries need to be expanded, and zero-padding is applied to the left, right, and bottom boundaries to eliminate boundary artifacts and improve reconstruction accuracy. The results after inverse migration are cropped and truncated using the original data dimensions and positions, ensuring a one-to-one correspondence between the truncated and original results.

[0087] like Figure 5 As shown, at the speed of electromagnetic waves in an air environment, for Figure 4 The extracted data was processed by stolt inverse offset, and the ground reflections were restored to the shape and amplitude of the original data.

[0088] Step 3: Subtract the ground object reflection data after inverse offset from the ground penetrating radar data to obtain ground penetrating radar data without ground object reflection.

[0089] By directly subtracting the ground object reflection separation result after inverse offset from the original ground penetrating radar data, preliminary data without ground object reflection is obtained.

[0090] A spike removal filter is set up to smooth and remove spikes from the initially obtained data without ground object reflections.

[0091] Peak removal filters include moving average filters and bandpass filters. Moving average filters smooth the data to eliminate spikes and jagged noise in the processed results. For each data point, the smoothed data is the average of the three data points surrounding it.

[0092] The bandpass filter smooths the data, aiming to eliminate high-frequency spikes and harsh noise in the processed results. Each data track is processed using a separate bandpass filter. Since this bandpass filter targets only ultra-high frequency spike noise and low-frequency background noise, its bandpass range is 5MHz-800MHz to avoid affecting the fully automated processing flow of ground-penetrating radar equipment at different frequencies.

[0093] like Figure 6 As shown, the processing result of separating ground object reflections from the original ground-penetrating radar data leaves only the underground reflection signal in the data.

[0094] This embodiment addresses ground-penetrating radar (GPR) data interfered with by echo signals from ground objects in the field. First, the data undergoes rigorous verification. Then, the Stolt migration method is used to migrate the data at the air electromagnetic wave velocity, thereby obtaining the focused energy clusters of ground object reflections. A GPR migration energy cluster extractor based on generative adversarial networks (GANs) receives the migrated data and completely extracts the ground object reflection energy cluster components. Next, the Stolt inverse migration method is used to recover the original ground object reflections from the energy clusters. Finally, the ground object reflections are separated from the original data. This implementation combines Stolt and inverse migration techniques, GAN technology, and various filter techniques to achieve automated and effective removal of surface object reflection interference during GPR detection.

[0095] Example 2

[0096] The purpose of this embodiment is to provide a ground-penetrating radar ground object interference removal system, including:

[0097] The offset processing module is configured to: acquire ground penetrating radar data and perform offset processing on the ground penetrating radar data at the air electromagnetic wave velocity using the Stolt offset method to obtain offset data;

[0098] The reverse migration processing module is configured to: for the focused ground object reflected energy cluster, use a trained generative adversarial network to extract the ground object reflected energy cluster components, and use the Stolt reverse migration method to perform reverse migration processing on the ground object reflected energy cluster components under the air electromagnetic wave velocity to obtain the reverse-migrated ground object reflection data.

[0099] The ground object interference module is configured to subtract the ground object reflection data after inverse offset from the ground penetrating radar data to obtain ground penetrating radar data without ground object reflection.

[0100] In further embodiments, the following is also provided:

[0101] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When executed by the processor, the computer instructions perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0102] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0103] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0104] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in Embodiment 1.

[0105] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0106] A computer program product includes a computer program that, when executed by a processor, implements the method described in Embodiment 1.

[0107] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions included in program modules, which execute in a device on a target real or virtual processor to perform the processes / methods described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc., that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided among program modules as needed. The machine-executable instructions for the program modules can execute within a local or distributed device. In a distributed device, the program modules can reside in both local and remote storage media.

[0108] The computer program code used to implement the methods of the present invention may be written in one or more programming languages. This computer program code may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the computer or other programmable data processing device, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a computer, partially on a computer, as a stand-alone software package, partially on a computer and partially on a remote computer, or entirely on a remote computer or server.

[0109] In the context of this invention, computer program code or related data may be carried by any suitable carrier to enable a device, apparatus, or processor to perform the various processes and operations described above. Examples of carriers include signals, computer-readable media, and the like. Examples of signals may include electrical, optical, radio, sound, or other forms of propagation signals, such as carrier waves, infrared signals, etc.

[0110] Those skilled in the art will recognize that the units and algorithm steps described in conjunction with the embodiments herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0111] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for removing ground object interference from ground penetrating radar, characterized in that, The method comprises the following steps: Obtaining ground penetrating radar data, and performing migration processing on the ground penetrating radar data at an air electromagnetic wave velocity by using a Stolt migration method to obtain post-migration data; For focused ground object reflection energy clusters, extracting ground object reflection energy cluster components by using a trained generative adversarial network, and performing de-migration processing on the ground object reflection energy cluster components at the air electromagnetic wave velocity by using a Stolt de-migration method to obtain de-migrated ground object reflection data; Subtracting the de-migrated ground object reflection data from the ground penetrating radar data to obtain ground object reflection-free ground penetrating radar data; The training of the generative adversarial network comprises the following steps: The randomly combined data set is migrated at the air electromagnetic wave velocity and used as the input of the generative adversarial network, the corresponding ground object reflection data set is migrated at the air electromagnetic wave velocity and used as the output of the generator in the generative adversarial network, and the generative adversarial network is trained; wherein the randomly combined data set comprises a subsurface reflection data set, a subsurface hyperbolic reflection data set and a simulated random ground object reflection data set.

2. The method of removing ground object interference from ground penetrating radar data of claim 1, wherein, In the Stolt migration and the Stolt de-migration, spline interpolation or sine function interpolation is used for spatial mapping.

3. The method of claim 1, wherein the step of removing the ground object interference from the ground penetrating radar data comprises the steps of: The construction of the random ground object reflection data set comprises the following steps: ​ Random Berlin noise is added to a random long-short-axis ellipse to establish a random target model; Numerical simulation is performed on the random target model randomly distributed in the air background to obtain simulated random ground object reflection data.

4. The method of claim 1, wherein the step of removing the ground object interference from the ground penetrating radar data comprises the steps of: determining a ground object interference signal; and removing the ground object interference signal from the ground penetrating radar data. 5 The ground penetrating radar data is processed at the air electromagnetic wave velocity by using the Stolt migration method, so that the diffraction wave is homed, the reflection wave is focused, and the reflection wave is effectively separated from other subsurface reflection signals.

5. The method of claim 1, wherein the step of removing the ground object interference from the ground penetrating radar data comprises the steps of: determining a ground object interference signal; and removing the ground object interference signal from the ground penetrating radar data. 5 The scale and energy of the ground object reflection energy cluster components are restored, and the restored ground object reflection energy cluster components are de-migrated by using the Stolt de-migration method.

6. The method of removing ground objects interference from ground penetrating radar data of claim 1, wherein, Further comprising: The obtained ground object reflection-free ground penetrating radar data is smoothed and peak removed.

7. A ground penetrating radar ground clutter removal system characterized by, The method comprises the following steps: An offset processing module is configured to obtain ground penetrating radar data, and perform migration processing on the ground penetrating radar data at an air electromagnetic wave velocity by using a Stolt migration method to obtain post-migration data; A de-migration processing module is configured to, for focused ground object reflection energy clusters, extract ground object reflection energy cluster components by using a trained generative adversarial network, and perform de-migration processing on the ground object reflection energy cluster components at the air electromagnetic wave velocity by using a Stolt de-migration method to obtain de-migrated ground object reflection data; wherein the training of the generative adversarial network comprises the following steps: the randomly combined data set is migrated at the air electromagnetic wave velocity and used as the input of the generative adversarial network, the corresponding ground object reflection data set is migrated at the air electromagnetic wave velocity and used as the output of the generator in the generative adversarial network, and the generative adversarial network is trained; wherein the randomly combined data set comprises a subsurface reflection data set, a subsurface hyperbolic reflection data set and a simulated random ground object reflection data set; A ground object interference module is configured to subtract the de-migrated ground object reflection data from the ground penetrating radar data to obtain ground object reflection-free ground penetrating radar data.

8. An electronic device, comprising: A computer program product comprising a memory and a processor, and computer instructions stored on the memory and running on the processor, which, when run by the processor, perform the method of any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, A computer program product for storing computer instructions, which, when executed by a processor, perform the method of any one of claims 1-6.

Images