Underground facility detection method and device, electronic equipment and readable medium

Abstract

The embodiment of the invention provides an underground facility detection method and device, electronic equipment and a readable medium, and the method comprises the steps: dividing an underground space related to a to-be-detected underground facility into at least one three-dimensional grid; obtaining at least one of the preset incident angle of at least one muon entering the three-dimensional grid, the energy attenuation information after the muon passes through the at least one three-dimensional grid and the size information of the three-dimensional grid; inputting at least one of the incident angle, the energy attenuation information and the size information into a preset grid density reasoning model to obtain reasoning density information of at least one three-dimensional grid; and determining the anomaly detection result of the three-dimensional grid based on the reasoning density information of the three-dimensional grid and / or the standard density information of the preset standard three-dimensional grid, thereby realizing centimeter-level detection of the underground facility without damaging the ground, improving the detection precision and detection efficiency of the underground facility, and reducing the detection cost.

Description

Technical Field

[0001] This application relates to the field of facility testing technology, and in particular to a method, apparatus, electronic device and readable medium for testing underground facilities. Background Technology

[0002] In related technologies, various pipelines and facilities, among other infrastructure, can be buried underground in urban areas. To inspect underground infrastructure, methods such as manual inspection, borehole drilling, or ground-penetrating radar can be employed. However, these methods suffer from drawbacks including high inspection costs, damage to road surfaces, and relatively low accuracy. Summary of the Invention

[0003] This application provides a method, apparatus, electronic device, and computer-readable storage medium for detecting underground facilities, in order to solve the problems of high detection costs, road surface damage, and low detection accuracy in the related technologies that use manual inspection, borehole detection, or ground-penetrating radar to detect underground infrastructure.

[0004] This application discloses a method for detecting underground facilities, including:

[0005] The underground space involved in the underground facility to be inspected is divided into at least one three-dimensional grid;

[0006] Obtain at least one of the following: the incident angle of at least one muon into the three-dimensional grid, the energy attenuation information of the muon after passing through at least one of the three-dimensional grids, and the size information of the three-dimensional grid;

[0007] Input at least one of the incident angle, the energy attenuation information, and the size information into a preset mesh density inference model to obtain the inference density information of the at least one three-dimensional mesh;

[0008] Based on the inference density information of the three-dimensional mesh and / or the standard density information of a preset standard three-dimensional mesh, the anomaly detection result of the three-dimensional mesh is determined.

[0009] Optionally, determining the anomaly detection result of the three-dimensional mesh based on the inference density information of the three-dimensional mesh and / or the standard density information of a preset standard three-dimensional mesh includes:

[0010] Extract the reasoning feature information from the reasoning density information;

[0011] Based on the inference feature information, at least one three-dimensional mesh is reconstructed to obtain at least one reconstructed three-dimensional mesh;

[0012] Based on at least one of the reconstructed 3D mesh, the standard density information, and the inference density information, the anomaly detection result of the 3D mesh is determined.

[0013] Optionally, determining the anomaly detection result of the three-dimensional mesh based on at least one of the reconstructed three-dimensional mesh, the standard density information, and the inference density information includes:

[0014] Extract the standard feature information from the standard density information;

[0015] The standard feature information and the inference feature information are compared to determine the feature difference data between the standard three-dimensional mesh and the three-dimensional mesh;

[0016] Based on the reconstructed 3D mesh and / or the feature difference data, the anomaly detection result of the 3D mesh is determined.

[0017] Optionally, determining the anomaly detection result of the three-dimensional mesh based on the reconstructed three-dimensional mesh and / or the feature difference data includes:

[0018] The three-dimensional mesh is compared with the reconstructed three-dimensional mesh to determine the reconstruction error data of the three-dimensional mesh;

[0019] For any of the three-dimensional meshes, if the reconstruction error data of the three-dimensional mesh is greater than a preset reconstruction error threshold, and / or the feature difference data of the three-dimensional mesh is greater than a preset feature difference threshold, then the three-dimensional mesh is determined to be abnormal.

[0020] Optionally, the method includes:

[0021] Obtain at least one of the following: the training incident angle of at least one training muon entering a preset training 3D grid, the training energy decay information of the training muon after passing through at least one of the training 3D grids, and the training size information of the training 3D grid;

[0022] Based on at least one of the training incident angle, the training size information, and the training energy decay information, the training standard density information of the training three-dimensional mesh is determined;

[0023] The preset generator is trained using at least one of the training incident angle, the training size information, the training energy decay information, and the training standard density information to obtain the grid density inference model.

[0024] Optionally, training the preset generator using at least one of the training incident angle, the training size information, the training energy decay information, and the training standard density information to obtain the grid density inference model includes:

[0025] Input at least one of the training incident angle, the training size information, and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh;

[0026] The training inference density information and / or the training standard density information are input into a preset discriminator to obtain similarity information between the training inference density information and the training standard density information.

[0027] Based on the training standard density information and / or the similarity information, it is determined whether the generator meets the preset training objective;

[0028] If the generator satisfies the training objective, then the generator is used as the grid density inference model;

[0029] If the generator does not meet the training objective, the steps of inputting at least one of the training incident angle, the training size information, and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh, and inputting the training inference density information and / or the training standard density information into a preset discriminator to obtain the similarity information between the training inference density information and the training standard density information are repeated until the generator meets the training objective.

[0030] Optionally, determining whether the generator meets the preset training objective based on the training standard density information and / or the similarity information includes:

[0031] Using the similarity information, the first loss function value of the generator is determined;

[0032] The second loss function value of the discriminator is determined using the first loss function value and / or the training standard density information;

[0033] Based on the first loss function value and / or the second loss function value, determine whether the generator satisfies the training objective.

[0034] Optionally, determining whether the generator satisfies the training objective based on the first loss function value and / or the second loss function value includes:

[0035] Determine whether the value of the first loss function is less than a preset threshold value for the first function;

[0036] Determine whether the value of the second loss function is greater than a preset threshold value for the second function.

[0037] If the first loss function value is less than the first function value threshold and the second loss function value is greater than the second function value threshold, then the generator is confirmed to meet the training objective.

[0038] Optionally, determining the training standard density information of the training 3D mesh based on at least one of the training incident angle, the training size information, and the training energy decay information includes:

[0039] Based on the training incident angle and / or the training size information, determine the path length of the training muon in the training 3D mesh;

[0040] Based on the path length and / or the training energy decay information, the training standard density information of the training 3D mesh is determined.

[0041] This application also discloses a detection device for underground facilities, comprising:

[0042] The grid division module is used to divide the underground space involved in the underground facility to be inspected into at least one three-dimensional grid.

[0043] The information acquisition module is used to acquire at least one of the following: the incident angle of at least one muon into the three-dimensional grid, the energy attenuation information of the muon after passing through at least one of the three-dimensional grids, and the size information of the three-dimensional grid;

[0044] The inference density information acquisition module is used to input at least one of the incident angle, the energy attenuation information and the size information into a preset grid density inference model to obtain the inference density information of the at least one three-dimensional grid.

[0045] The result determination module is used to determine the anomaly detection result of the three-dimensional mesh based on the inference density information of the three-dimensional mesh and / or the standard density information of a preset standard three-dimensional mesh.

[0046] Optionally, the result determination module includes:

[0047] The reasoning feature information extraction submodule is used to extract the reasoning feature information of the reasoning density information;

[0048] The mesh reconstruction submodule is used to reconstruct at least one three-dimensional mesh based on the inference feature information, so as to obtain at least one reconstructed three-dimensional mesh;

[0049] The result determination submodule is used to determine the anomaly detection result of the three-dimensional mesh based on at least one of the reconstructed three-dimensional mesh, the standard density information, and the inference density information.

[0050] Optionally, the result determination submodule includes:

[0051] A standard feature information extraction unit is used to extract standard feature information from the standard density information; a feature difference data determination unit is used to compare the standard feature information and the inference feature information to determine the feature difference data between the standard three-dimensional mesh and the three-dimensional mesh.

[0052] The result determination unit is used to determine the anomaly detection result of the three-dimensional mesh based on the reconstructed three-dimensional mesh and / or the feature difference data.

[0053] Optionally, the result determination unit includes:

[0054] The reconstruction error data determination sub-unit is used to compare the three-dimensional mesh with the reconstructed three-dimensional mesh to determine the reconstruction error data of the three-dimensional mesh;

[0055] An anomaly determination subunit is used to determine that the three-dimensional mesh is abnormal if, for any of the three-dimensional meshes, the reconstruction error data of the three-dimensional mesh is greater than a preset reconstruction error threshold, and / or the feature difference data of the three-dimensional mesh is greater than a preset feature difference threshold.

[0056] Optionally, the device includes:

[0057] The training information acquisition module is used to acquire at least one of the following: the training incident angle of at least one training muon entering a preset training three-dimensional grid, the training energy decay information of the training muon after passing through at least one of the training three-dimensional grids, and the training size information of the training three-dimensional grid.

[0058] The training standard density information determination module is used to determine the training standard density information of the training three-dimensional mesh based on at least one of the training incident angle, the training size information, and the training energy decay information.

[0059] The training module is used to train a preset generator using at least one of the training incident angle, the training size information, the training energy decay information, and the training standard density information to obtain the grid density inference model.

[0060] Optionally, the training module includes:

[0061] The training inference density information acquisition submodule is used to input at least one of the training incident angle, the training size information and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh.

[0062] The similarity information acquisition submodule is used to input the training inference density information and / or the training standard density information into a preset discriminator to obtain similarity information between the training inference density information and the training standard density information;

[0063] The judgment submodule is used to determine whether the generator meets the preset training objective based on the training standard density information and / or the similarity information.

[0064] As a submodule, it is used to use the generator as the grid density inference model if the generator satisfies the training objective.

[0065] The repeat submodule is used to repeat the steps of inputting at least one of the training incident angle, the training size information, and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh, and inputting the training inference density information and / or the training standard density information into a preset discriminator to obtain the similarity information between the training inference density information and the training standard density information, if the generator does not meet the training objective, until the generator meets the training objective.

[0066] Optionally, the determination submodule includes:

[0067] The first loss function value determination unit is used to determine the first loss function value of the generator using the similarity information;

[0068] The second loss function value determination unit is used to determine the second loss function value of the discriminator using the first loss function value and / or the training standard density information;

[0069] The judgment unit is used to determine whether the generator satisfies the training objective based on the first loss function value and / or the second loss function value.

[0070] Optionally, the determining unit includes:

[0071] The first judgment subunit is used to determine whether the first loss function value is less than a preset first function value threshold.

[0072] The second judgment subunit is used to determine whether the value of the second loss function is greater than a preset second function value threshold.

[0073] The confirmation subunit is used to confirm that the generator satisfies the training objective if the first loss function value is less than the first function value threshold and the second loss function value is greater than the second function value threshold.

[0074] Optionally, the training standard density information determination module includes:

[0075] The path length determination submodule is used to determine the path length of the training muon in the training 3D mesh based on the training incident angle and / or the training size information.

[0076] The training standard density information determination submodule is used to determine the training standard density information of the training 3D mesh based on the path length and / or the training energy decay information.

[0077] This application also discloses an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0078] The memory is used to store computer programs;

[0079] When the processor executes a program stored in the memory, it implements the method described in the embodiments of this application.

[0080] This application also discloses one or more computer-readable media storing instructions that, when executed by one or more processors, cause the processors to perform the methods described in this application.

[0081] The embodiments of this application have the following advantages:

[0082] In this embodiment, the underground space involved in the underground facility to be detected is divided into at least one three-dimensional grid; at least one of the following is obtained: the incident angle of at least one muon entering the three-dimensional grid, the energy attenuation information of the muon after passing through at least one three-dimensional grid, and the size information of the three-dimensional grid; at least one of the incident angle, energy attenuation information, and size information is input into a preset grid density inference model to obtain the inference density information of at least one three-dimensional grid; based on the inference density information of the three-dimensional grid and / or the standard density information of a preset standard three-dimensional grid, the anomaly detection result of the three-dimensional grid is determined, thereby achieving centimeter-level detection of underground facilities without damaging the ground, improving the detection accuracy and efficiency of underground facilities, and reducing detection costs. Attached Figure Description

[0083] Figure 1 This is a flowchart illustrating the steps of a method for detecting underground facilities provided in an embodiment of this application;

[0084] Figure 2 This is a flowchart illustrating a method for detecting underground facilities provided in an embodiment of this application;

[0085] Figure 3 This is a structural block diagram of a detection device for underground facilities provided in an embodiment of this application;

[0086] Figure 4 This is a block diagram of an electronic device provided in an embodiment of this application;

[0087] Figure 5 This is a schematic diagram of a computer-readable medium provided in an embodiment of this application. Detailed Implementation

[0088] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0089] To facilitate understanding of the technical solutions and effects of the embodiments of this application, the relevant technologies of this application will be briefly described below.

[0090] In related technologies, various pipelines and facilities are typically buried underground in urban areas. To inspect these underground infrastructures, technologies usually employ methods such as manual inspection, borehole exploration, or ground-penetrating radar.

[0091] However, these methods suffer from high detection costs, damage to the road surface, and low detection accuracy. Detecting underground infrastructure by damaging the road surface not only increases the cost and difficulty of road maintenance but also may lead to cumbersome operations. The low detection accuracy refers to the difficulty in accurately determining the specific location of underground infrastructure, potentially resulting in significant discrepancies between the detection results and actual values—sometimes reaching several meters.

[0092] Reference Figure 1 The diagram illustrates a flowchart of a method for detecting underground facilities provided in an embodiment of this application, which may specifically include the following steps:

[0093] Step 101: Divide the underground space involved in the underground facility to be inspected into at least one three-dimensional grid;

[0094] In this embodiment of the application, various pipelines and facilities are typically buried underground in urban areas and other regions. In this embodiment, at least one muon detector can be deployed on the surface corresponding to the underground space involved in the underground facility to be detected. This muon detector is also known as a muon detector. Multiple muon detectors are integrated with a high-precision clock module to achieve time synchronization of data acquisition.

[0095] In one example, three muon detectors are deployed on the surface corresponding to the underground space involved in the underground facility to be detected. These three muon detectors are arranged in a straight line along the surface, covering the underground space involved in the facility to be detected, with a spacing of 50-100 meters between adjacent muon detectors. The spacing between adjacent muon detectors can be adjusted according to the actual terrain.

[0096] In yet another example, at least one muon detector is deployed in a grid array on the surface corresponding to the underground space involved in the underground facility to be inspected.

[0097] In this embodiment, the underground space involved in the underground facility to be inspected can be divided into at least one three-dimensional grid. The three-dimensional grid can be referred to as a voxel grid, and a voxel grid can have a volume of 1 cm³. 3 A cubic grid of (cubic centimeters).

[0098] Step 102: Obtain at least one of the following: the incident angle of at least one muon entering the three-dimensional grid, the energy attenuation information of the muon after passing through at least one of the three-dimensional grids, and the size information of the three-dimensional grid;

[0099] In this embodiment of the application, cosmic rays, such as muons, are emitted from space towards the ground and penetrate the ground and underground structures. Muons can be referred to as muons.

[0100] In this embodiment, the underground space involved in the underground facility to be detected can be divided into at least one three-dimensional grid, and the size information of the three-dimensional grid can be obtained. A muon detector deployed on the surface corresponding to the underground space involved in the underground facility to be detected can collect the incident angle of at least one muon entering the three-dimensional grid and / or the energy attenuation information of the muon after passing through at least one three-dimensional grid.

[0101] The incident angle can refer to the direction of motion of the muon entering the 3D grid, such as the pitch angle or azimuth angle. Energy decay information refers to the flux decay ratio between the flux of any muon before it passes through at least one 3D grid and the flux of that muon after it passes through at least one 3D grid. If Φ0 represents the flux of any muon before it passes through at least one 3D grid, and Φ represents the flux of that muon after it passes through at least one 3D grid, then the flux decay ratio can be ln(Φ0 / Φ). The muon flux can refer to the muon intensity.

[0102] Step 103: Input at least one of the incident angle, the energy attenuation information and the size information into a preset grid density inference model to obtain the inference density information of the at least one three-dimensional grid.

[0103] In this embodiment, the incident angle of at least one muon entering the three-dimensional grid and / or the energy attenuation information of the muon after passing through at least one three-dimensional grid, collected by the muon detector, can be referred to as initial muon data. In this embodiment, the initial muon data can be subjected to noise filtering to remove sharp noise or outliers and retain effective signal characteristics.

[0104] In this embodiment, a median filter can be used to detect sharp noise or outlier data in the initial muon data. Then, for the data S(t) corresponding to sharp noise or outlier, a filtering window can be selected, and the median of the data within the filtering window can be used to replace the data corresponding to sharp noise or outlier, thereby converting the initial muon data into target muon data. The window length can be represented by w, which can be 3 points or 5 points.

[0105] In one example, at the i-th data point of the initial muon data, neighboring points S(i-1), S(i), and S(i+1) are selected with a window length w = 3. The median of the data within the window is used to replace the i-th data point. The formula for calculating the median of the data within the window is:

[0106] Sdenoised(i)=median(S(i-1),S(i),S(i+1))

[0107] Where Sdenoised(i) is the median of the data within the window.

[0108] In this embodiment, the incident angle, energy attenuation information, and 3D mesh size information from the target muon data are input into a preset mesh density inference model to obtain inference density information for at least one 3D mesh. The inference density information of the 3D mesh obtained by dividing the underground space involved in the underground facility to be detected can be referred to as the material density projection.

[0109] In some embodiments of this application, the method includes:

[0110] Obtain at least one of the following: the training incident angle of at least one training muon entering a preset training 3D grid, the training energy decay information of the training muon after passing through at least one of the training 3D grids, and the training size information of the training 3D grid;

[0111] Based on at least one of the training incident angle, the training size information, and the training energy decay information, the training standard density information of the training three-dimensional mesh is determined;

[0112] The preset generator is trained using at least one of the training incident angle, the training size information, the training energy decay information, and the training standard density information to obtain the grid density inference model.

[0113] In this embodiment, at least one of the following can be obtained: the training incident angle of at least one preset training muon entering a preset training 3D grid; the training energy attenuation information of the training muon after passing through at least one training 3D grid; and the training size information of the training 3D grid. Here, the training muon refers to a muon used for training, and the training 3D grid is a 3D grid obtained by dividing the underground space involved in the underground facility to be detected. The training energy attenuation information is the energy attenuation information used for training, and the training size information is the size information used for training.

[0114] In this embodiment, the training standard density information of the training 3D mesh can be determined based on at least one of the training incident angle, training size information, and training energy attenuation information. Alternatively, the training standard density information of the training 3D mesh can also be obtained by performing a CT (Computed Tomography) scan on the underground space involved in the underground facility to be detected for training.

[0115] In this embodiment of the application, a grid density inference model can be obtained by training a preset generator using at least one of the following: training incident angle, training size information, training energy decay information, and training standard density information.

[0116] In this embodiment of the application, the training standard density information of the training three-dimensional mesh is determined by training at least one of the incident angle, training size information, and training energy attenuation information; the preset generator is trained by using at least one of the incident angle, training size information, training energy attenuation information, and training standard density information to obtain a mesh density inference model, and the density information of the underground space involved in the underground facility to be detected can be determined by using the mesh density inference model.

[0117] In some embodiments of this application, determining the training standard density information of the training 3D mesh based on at least one of the training incident angle, the training size information, and the training energy decay information includes:

[0118] Based on the training incident angle and / or the training size information, determine the path length of the training muon in the training 3D mesh;

[0119] Based on the path length and / or the training energy decay information, the training standard density information of the training 3D mesh is determined.

[0120] In this embodiment, the path length of the training muon in the training 3D grid can be determined using the training incident angle of the training muon entering the training 3D grid and / or the training size information of the training 3D grid. Then, the training standard density information of the training 3D grid can be determined using the path length of the training muon in the training 3D grid and / or the training energy decay information after the training muon passes through at least one training 3D grid.

[0121] Each training 3D grid has a voxel index, which serves as the identifier for the training 3D grid. Training energy decay information represents the training flux decay ratio after the training muon passes through at least one training 3D grid. The training incident angle is the direction of motion of the training muon entering the training 3D grid, such as pitch or azimuth.

[0122] In this embodiment, a system of linear equations can be constructed. Using this system, and employing the path length of the training muon within the training 3D grid and / or the training energy decay information after the training muon traverses at least one training 3D grid, the standard density information of the training 3D grid can be determined. The system of linear equations is as follows:

[0123] A*x=b

[0124] Where A is the path matrix, each row of the path matrix corresponds to a training muon, and each column corresponds to a training 3D grid. The element corresponding to the training muon in the i-th row and the training 3D grid in the j-th column is the path length L of the training muon in that training 3D grid. ij .

[0125] b is the observation vector, and each element b in the observation vector b is... i Let Φ represent the training flux decay ratio of the training muon in the i-th row. If Φ0 represents the flux of the training muon before it passes through at least one training 3D grid, and Φ represents the flux of the training muon after it passes through at least one training 3D grid, then the training flux decay ratio of the training muon can be ln(Φ0 / Φ).

[0126] x includes the standard training density of at least one training 3D grid. x = [μ1ρ1, μ2ρ2, ..., μ n ρ n ], where ρ1,ρ2,…,ρ n The training standard density, μ1,μ2,…,μ, is used to train the 3D mesh. n This is the preset attenuation coefficient.

[0127] In one example, a training muon traverses three training 3D grids, with path lengths of L1 = 10 cm, L2 = 5 cm, and L3 = 8 cm, respectively. The training flux attenuation ratio for this training muon traversing these three training 3D grids is ln(Φ0 / Φ) = 0.5.

[0128] μ1ρ1*10+μ2ρ2*5+μ3ρ3*8=0.5

[0129] By solving this formula, the standard training density of the training 3D mesh can be obtained.

[0130] In the embodiments of this application, the path length of the training muon in the training three-dimensional mesh is determined based on the training incident angle and / or training size information, and the accurate training standard density information of the training three-dimensional mesh is obtained based on the path length and / or training energy decay information.

[0131] In some embodiments of this application, training a preset generator using at least one of the training incident angle, the training size information, the training energy decay information, and the training standard density information to obtain the grid density inference model includes:

[0132] Input at least one of the training incident angle, the training size information, and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh;

[0133] The training inference density information and / or the training standard density information are input into a preset discriminator to obtain similarity information between the training inference density information and the training standard density information.

[0134] Based on the training standard density information and / or the similarity information, it is determined whether the generator meets the preset training objective;

[0135] If the generator satisfies the training objective, then the generator is used as the grid density inference model;

[0136] If the generator does not meet the training objective, the steps of inputting at least one of the training incident angle, the training size information, and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh, and inputting the training inference density information and / or the training standard density information into a preset discriminator to obtain the similarity information between the training inference density information and the training standard density information are repeated until the generator meets the training objective.

[0137] In the embodiments of this application, the constructed GAN (Generative Adversarial Network) includes a generator and a discriminator.

[0138] The generator network structure is as follows:

[0139] Input layer: Used to receive input data, including: training incident angle, training energy decay information, and training size information.

[0140] Hidden layers: Employ a multi-layered convolutional neural network (CNN), specifically including: convolutional layers for extracting features from the input data; batch normalization for stabilizing the training process and accelerating convergence; and activation functions such as ReLU (Rectified Linear Unit) for introducing non-linear transformations and enhancing the model's expressive power.

[0141] Output layer: Used to generate training inference density information G(p) for training 3D mesh.

[0142] The formula for calculating z in the core convolutional layer is:

[0143] z = ReLU(Wk*x + bk)

[0144] Where Wk and bk are the convolution kernel parameters, and x is the input data.

[0145] The discriminator network structure is as follows:

[0146] Input layer: Used to receive the training inference density information and the training standard density information of the training 3D mesh generated by the generator.

[0147] Hidden layers: These are used to extract features from the input data of the input discriminator through multiple CNN layers. Specifically, they include: convolutional layers, used to extract local features from the input data; batch normalization layers, used to ensure the stability of the training process; and activation functions, such as Leaky ReLU, used to avoid neuron death problems and improve robustness.

[0148] Output layer: Used to output a probability score D(x), representing the probability that the training inference density information of the training 3D grid generated by the generator is the training standard density information of the training 3D grid.

[0149] In this embodiment, at least one of the following can be input into the generator: the training incident angle of the training muon into the training 3D grid, the training size information of the training 3D grid, and the training energy decay information after the training muon passes through at least one training 3D grid, to obtain the training inference density information of the training 3D grid.

[0150] Then, the training inference density information and the training standard density information are input into the discriminator to obtain the similarity information between the training inference density information and the training standard density information. This similarity information indicates the probability that the training inference density information of the training 3D mesh generated by the generator is the training standard density information of the training 3D mesh; the higher the similarity in the similarity information, the greater the probability; the lower the similarity in the similarity information, the smaller the probability.

[0151] In this embodiment, it can be determined whether the generator meets the preset training objective based on the training standard density information and / or similarity information. If the generator meets the training objective, then the generator is used as a grid density inference model;

[0152] If the generator does not meet the training objective, the process of inputting at least one of the training incident angle, training size information, and training energy decay information into the generator to obtain the training inference density information of the training 3D mesh, and inputting the training inference density information and / or training standard density information into a preset discriminator to obtain the similarity information between the training inference density information and the training standard density information is repeated until the generator meets the training objective.

[0153] In this embodiment of the application, the generator and discriminator are trained alternately so that the generator and discriminator respectively meet the corresponding training objectives. When the generator meets the training objectives, a grid density inference model is obtained.

[0154] In some embodiments of this application, determining whether the generator meets a preset training objective based on the training standard density information and / or the similarity information includes:

[0155] Using the similarity information, the first loss function value of the generator is determined;

[0156] The second loss function value of the discriminator is determined using the first loss function value and / or the training standard density information;

[0157] Based on the first loss function value and / or the second loss function value, determine whether the generator satisfies the training objective.

[0158] In this embodiment, the similarity information between the training inference density information output by the discriminator and the training standard density information is used, and the generator's loss function is employed to determine the generator's first loss function value. The generator's loss function is:

[0159] L G =E[log(1-D(G(p))]

[0160] Where G refers to the generator, D refers to the discriminator, G(p) is the training inference density information, D(G(p)) is the similarity information output by the discriminator, and L... G It is the loss function of the generator.

[0161] Using the generator's first loss function value and / or training standard density information, and leveraging the discriminator's loss function, the discriminator's second loss function value can be determined. The discriminator's loss function is:

[0162] L D =-E[logD(Xreal)]-E[log(1-D(G(p)))]

[0163] Among them, L D Xreal is the loss function of the discriminator, and Xreal is the training standard density information.

[0164] Based on the first loss function value and / or the second loss function value, it can be determined whether the generator meets the training objective.

[0165] In the embodiments of this application, similarity information is used to determine the first loss function value of the generator; the first loss function value and / or training standard density information are used to determine the second loss function value of the discriminator; based on the first loss function value and / or the second loss function value, it is determined whether the generator meets the training objective, thus providing a criterion for determining whether the generator meets the training objective.

[0166] In some embodiments of this application, determining whether the generator satisfies the training objective based on the first loss function value and / or the second loss function value includes:

[0167] Determine whether the value of the first loss function is less than a preset threshold value for the first function;

[0168] Determine whether the value of the second loss function is greater than a preset threshold value for the second function.

[0169] If the first loss function value is less than the first function value threshold and the second loss function value is greater than the second function value threshold, then the generator is confirmed to meet the training objective.

[0170] In this embodiment, the training objective of the generator is to generate the most accurate training inference density information for the training 3D mesh, making it impossible for the discriminator to distinguish between the training inference density information generated by the generator and the standard training density information. The discriminator assumes that the training inference density information generated by the generator is, as far as possible, the standard training density information. The training objective of the discriminator is to distinguish between the training inference density information generated by the generator and the standard training density information as accurately as possible, thereby improving the discriminator's recognition capability.

[0171] The training objectives for the generator and discriminator can be expressed as:

[0172] min G max D E[logD(Xreal)]+E[log(1-D(G(p)))]

[0173] In this embodiment, it can be determined whether the first loss function value of the generator is less than a preset first function value threshold; and whether the second loss function value of the discriminator is greater than a preset second function value threshold. If the first loss function value is less than the first function value threshold and the second loss function value is greater than the second function value threshold, then it is confirmed that the generator and the discriminator meet the training objective.

[0174] In this embodiment of the application, if the value of the first loss function is less than the first function value threshold and the value of the second loss function is greater than the second function value threshold, then the generator is confirmed to meet the training objective, and a grid density inference model is obtained.

[0175] Step 104: Based on the inference density information of the three-dimensional mesh and / or the standard density information of the preset standard three-dimensional mesh, determine the anomaly detection result of the three-dimensional mesh.

[0176] In this embodiment, based on the inference density information of the 3D mesh and / or the standard density information of a preset standard 3D mesh, it is possible to determine whether the 3D mesh is abnormal and to determine the anomaly detection result of the 3D mesh. A standard 3D mesh can be a 3D mesh that has not historically exhibited any anomalies. An abnormal 3D mesh can refer to deformed or hollow areas.

[0177] In this embodiment of the application, a visualization platform is developed. The real-time monitoring system can display the inference density information of the 3D mesh and abnormal 3D meshes. The detection results are displayed intuitively through the interface, and an alarm is triggered when abnormal 3D meshes are detected. The real-time monitoring system can automatically notify maintenance personnel.

[0178] In some embodiments of this application, determining the anomaly detection result of the three-dimensional mesh based on the inference density information of the three-dimensional mesh and / or the standard density information of a preset standard three-dimensional mesh includes:

[0179] Extract the reasoning feature information from the reasoning density information;

[0180] Based on the inference feature information, at least one three-dimensional mesh is reconstructed to obtain at least one reconstructed three-dimensional mesh;

[0181] Based on at least one of the reconstructed 3D mesh, the standard density information, and the inference density information, the anomaly detection result of the 3D mesh is determined.

[0182] In this embodiment, inference feature information of the inference density information of a 3D mesh can be extracted using a convolutional autoencoder (CAE). Then, based on the inference feature information, at least one 3D mesh is reconstructed to obtain at least one reconstructed 3D mesh. Based on at least one of the reconstructed 3D mesh, standard density information, and inference density information, it can be determined whether the 3D mesh is abnormal, and the anomaly detection result of the 3D mesh can be determined.

[0183] In this embodiment, inference feature information of inference density information of three-dimensional mesh is extracted; at least one three-dimensional mesh is reconstructed based on the inference feature information to obtain at least one reconstructed three-dimensional mesh; based on at least one of the reconstructed three-dimensional mesh, standard density information and inference density information, it is determined whether the three-dimensional mesh is abnormal, and the abnormality detection result of the three-dimensional mesh is determined, thereby realizing the detection of underground facilities without damaging the ground.

[0184] In some embodiments of this application, determining the anomaly detection result of the three-dimensional mesh based on at least one of the reconstructed three-dimensional mesh, the standard density information, and the inference density information includes:

[0185] Extract the standard feature information from the standard density information;

[0186] The standard feature information and the inference feature information are compared to determine the feature difference data between the standard three-dimensional mesh and the three-dimensional mesh;

[0187] Based on the reconstructed 3D mesh and / or the feature difference data, the anomaly detection result of the 3D mesh is determined.

[0188] In this embodiment, standard feature information of the standard density information of a standard 3D mesh can be extracted using a convolutional autoencoder (CAE). Then, by comparing the standard feature information with the inferred feature information, the feature difference data between the standard 3D mesh and the 3D mesh can be determined. Based on the reconstructed 3D mesh and / or the feature difference data, it can be determined whether the 3D mesh is abnormal, thus determining the anomaly detection result of the 3D mesh.

[0189] In this embodiment, standard feature information of standard density information of standard three-dimensional mesh is extracted; standard feature information and inferred feature information are compared to determine feature difference data between standard three-dimensional mesh and three-dimensional mesh; based on the reconstructed three-dimensional mesh and / or feature difference data, the judgment of whether the three-dimensional mesh is abnormal is realized.

[0190] In some embodiments of this application, determining the anomaly detection result of the three-dimensional mesh based on the reconstructed three-dimensional mesh and / or the feature difference data includes:

[0191] The three-dimensional mesh is compared with the reconstructed three-dimensional mesh to determine the reconstruction error data of the three-dimensional mesh;

[0192] For any of the three-dimensional meshes, if the reconstruction error data of the three-dimensional mesh is greater than a preset reconstruction error threshold, and / or the feature difference data of the three-dimensional mesh is greater than a preset feature difference threshold, then the three-dimensional mesh is determined to be abnormal.

[0193] In this embodiment, a 3D mesh can be compared with a reconstructed 3D mesh to determine the reconstruction error data of the 3D mesh. For any 3D mesh, if the reconstruction error data of the 3D mesh is greater than a preset reconstruction error threshold, and / or the feature difference data of the 3D mesh is greater than a preset feature difference threshold, then the 3D mesh is determined to be abnormal.

[0194] In this embodiment of the application, the determination of whether a 3D mesh is abnormal is achieved by using the reconstruction error data of the 3D mesh and the feature difference data between the 3D mesh and the standard 3D mesh.

[0195] In this embodiment, the underground space involved in the underground facility to be detected is divided into at least one three-dimensional grid; at least one of the following is obtained: the incident angle of at least one muon entering the three-dimensional grid, the energy attenuation information of the muon after passing through at least one three-dimensional grid, and the size information of the three-dimensional grid; at least one of the incident angle, energy attenuation information, and size information is input into a preset grid density inference model to obtain the inference density information of at least one three-dimensional grid; based on the inference density information of the three-dimensional grid and / or the standard density information of a preset standard three-dimensional grid, the anomaly detection result of the three-dimensional grid is determined, thereby achieving centimeter-level detection of underground facilities without damaging the ground, improving the detection accuracy and efficiency of underground facilities, and reducing detection costs.

[0196] In this embodiment, muon imaging combined with GAN deep learning enables centimeter-level detection of underground facilities without damaging the ground. GAN enhances the resolution of muon imaging, improving the accuracy of underground facility detection and achieving high-precision non-destructive testing, capable of identifying minute deformations and cavities in underground facilities. Detection is performed using natural cosmic rays from muons, without damaging the road surface. Anomalies in underground facilities are determined with relatively little muon data, reducing the frequency of manual inspections and lowering municipal maintenance costs.

[0197] In this application embodiment, different types of detectors can be used to replace the muon detector, for example, introducing a higher-precision particle detector to improve detection sensitivity and accuracy. In this application embodiment, the integration module of the detector array can be enhanced, for example, by integrating more advanced clock synchronization technology to improve the accuracy of data acquisition. In this application embodiment, other deep learning algorithms such as variational autoencoders (VAEs) and attention mechanisms can be used to optimize the 3D reconstruction effect. Different noise processing algorithms can be used to improve the efficiency of signal filtering during the data preprocessing stage, thereby improving the accuracy of detection results. In this application embodiment, a multi-platform compatible visualization system can be developed to adapt to different operating environments and devices, improving the user experience. In this application embodiment, augmented reality (AR) technology can be introduced to achieve real-time visualization of on-site detection results, enhancing the interactivity and intuitiveness for maintenance personnel. In this application embodiment, optimized wireless communication protocols or Internet of Things (IoT) technologies (such as NB-IoT, LoRa) can also be used to achieve data transmission between detectors, improving the system's flexibility and scalability. Security protocols are introduced to ensure the security of data transmission and storage, preventing information leakage or tampering.

[0198] Reference Figure 2 The diagram illustrates a flowchart of a method for detecting underground facilities provided in an embodiment of this application. First, the underground space involved in the underground facility to be detected is divided into at least one three-dimensional grid. Data is collected using muon detectors 1, 2, and 3, including at least one of the following: the incident angle of at least one muon entering the three-dimensional grid, energy attenuation information of the muon after passing through at least one three-dimensional grid, and the size information of the three-dimensional grid. Then, the data collected by the three muon detectors is preprocessed.

[0199] The 3D reconstruction processing of GAN involves training the generator and discriminator to transform the generator into a mesh density inference model. At least one of the incident angle, energy attenuation information, and size information is input into the pre-defined mesh density inference model to obtain the inference density information of at least one 3D mesh, and the 3D structure of the underground pipeline network is output.

[0200] Inference feature information of the inference density information of the 3D mesh is extracted using a convolutional autoencoder (CAE), and standard feature information of the standard density information of the standard 3D mesh is extracted using a convolutional autoencoder (CAE). Then, the standard feature information and the inference feature information are compared to determine the feature difference data between the standard 3D mesh and the 3D mesh.

[0201] Based on the inference feature information, at least one 3D mesh is reconstructed to obtain at least one reconstructed 3D mesh. The 3D mesh is compared with the reconstructed 3D mesh to determine the reconstruction error data of the 3D mesh. For any 3D mesh, if the reconstruction error data of the 3D mesh is greater than a preset reconstruction error threshold, and / or the feature difference data of the 3D mesh is greater than a preset feature difference threshold, then the 3D mesh is determined to be abnormal. The real-time monitoring system can display the inference density information of the 3D mesh and the abnormal 3D mesh, intuitively display the detection results through the interface, and alarm when an abnormal 3D mesh is detected. The real-time monitoring system can automatically notify maintenance personnel.

[0202] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of this application.

[0203] Reference Figure 3 The diagram shows a structural block diagram of a detection device for underground facilities provided in an embodiment of this application, which may specifically include the following modules:

[0204] The grid division module 301 is used to divide the underground space involved in the underground facility to be inspected into at least one three-dimensional grid.

[0205] Information acquisition module 302 is used to acquire at least one of the following: the incident angle of at least one muon entering the three-dimensional grid, the energy attenuation information of the muon after passing through at least one of the three-dimensional grids, and the size information of the three-dimensional grid;

[0206] The inference density information acquisition module 303 is used to input at least one of the incident angle, the energy attenuation information and the size information into a preset grid density inference model to obtain the inference density information of the at least one three-dimensional grid.

[0207] The result determination module 304 is used to determine the anomaly detection result of the three-dimensional mesh based on the inference density information of the three-dimensional mesh and / or the standard density information of a preset standard three-dimensional mesh.

[0208] In one optional embodiment of this application, the result determination module includes:

[0209] The reasoning feature information extraction submodule is used to extract the reasoning feature information of the reasoning density information;

[0210] The mesh reconstruction submodule is used to reconstruct at least one three-dimensional mesh based on the inference feature information, so as to obtain at least one reconstructed three-dimensional mesh;

[0211] The result determination submodule is used to determine the anomaly detection result of the three-dimensional mesh based on at least one of the reconstructed three-dimensional mesh, the standard density information, and the inference density information.

[0212] In one optional embodiment of this application, the result determination submodule includes:

[0213] A standard feature information extraction unit is used to extract the standard feature information of the standard density information;

[0214] The feature difference data determination unit is used to compare the standard feature information and the inferred feature information to determine the feature difference data between the standard three-dimensional mesh and the three-dimensional mesh;

[0215] The result determination unit is used to determine the anomaly detection result of the three-dimensional mesh based on the reconstructed three-dimensional mesh and / or the feature difference data.

[0216] In one optional embodiment of this application, the result determination unit includes:

[0217] The reconstruction error data determination sub-unit is used to compare the three-dimensional mesh with the reconstructed three-dimensional mesh to determine the reconstruction error data of the three-dimensional mesh;

[0218] An anomaly determination subunit is used to determine that the three-dimensional mesh is abnormal if, for any of the three-dimensional meshes, the reconstruction error data of the three-dimensional mesh is greater than a preset reconstruction error threshold, and / or the feature difference data of the three-dimensional mesh is greater than a preset feature difference threshold.

[0219] In one optional embodiment of this application, the apparatus includes:

[0220] The training information acquisition module is used to acquire at least one of the following: the training incident angle of at least one training muon entering a preset training three-dimensional grid, the training energy decay information of the training muon after passing through at least one of the training three-dimensional grids, and the training size information of the training three-dimensional grid.

[0221] The training standard density information determination module is used to determine the training standard density information of the training three-dimensional mesh based on at least one of the training incident angle, the training size information, and the training energy decay information.

[0222] The training module is used to train a preset generator using at least one of the training incident angle, the training size information, the training energy decay information, and the training standard density information to obtain the grid density inference model.

[0223] In one optional embodiment of this application, the training module includes:

[0224] The training inference density information acquisition submodule is used to input at least one of the training incident angle, the training size information and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh.

[0225] The similarity information acquisition submodule is used to input the training inference density information and / or the training standard density information into a preset discriminator to obtain similarity information between the training inference density information and the training standard density information;

[0226] The judgment submodule is used to determine whether the generator meets the preset training objective based on the training standard density information and / or the similarity information.

[0227] As a submodule, it is used to use the generator as the grid density inference model if the generator satisfies the training objective.

[0228] The repeat submodule is used to repeat the steps of inputting at least one of the training incident angle, the training size information, and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh, and inputting the training inference density information and / or the training standard density information into a preset discriminator to obtain the similarity information between the training inference density information and the training standard density information, if the generator does not meet the training objective, until the generator meets the training objective.

[0229] In one optional embodiment of this application, the determination submodule includes:

[0230] The first loss function value determination unit is used to determine the first loss function value of the generator using the similarity information;

[0231] The second loss function value determination unit is used to determine the second loss function value of the discriminator using the first loss function value and / or the training standard density information;

[0232] The judgment unit is used to determine whether the generator satisfies the training objective based on the first loss function value and / or the second loss function value.

[0233] In one optional embodiment of this application, the determining unit includes:

[0234] The first judgment subunit is used to determine whether the first loss function value is less than a preset first function value threshold.

[0235] The second judgment subunit is used to determine whether the value of the second loss function is greater than a preset second function value threshold.

[0236] The confirmation subunit is used to confirm that the generator satisfies the training objective if the first loss function value is less than the first function value threshold and the second loss function value is greater than the second function value threshold.

[0237] In one optional embodiment of this application, the training standard density information determination module includes:

[0238] The path length determination submodule is used to determine the path length of the training muon in the training 3D mesh based on the training incident angle and / or the training size information.

[0239] The training standard density information determination submodule is used to determine the training standard density information of the training 3D mesh based on the path length and / or the training energy decay information.

[0240] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0241] In addition, embodiments of this application also provide an electronic device, such as... Figure 4 As shown, it includes a processor 401, a communication interface 402, a memory 403, and a communication bus 404, wherein the processor 401, the communication interface 402, and the memory 403 communicate with each other through the communication bus 404.

[0242] Memory 403 is used to store computer programs;

[0243] When processor 401 executes the program stored in memory 403, it performs the following steps:

[0244] The underground space involved in the underground facility to be inspected is divided into at least one three-dimensional grid;

[0245] Obtain at least one of the following: the incident angle of at least one muon into the three-dimensional grid, the energy attenuation information of the muon after passing through at least one of the three-dimensional grids, and the size information of the three-dimensional grid;

[0246] Input at least one of the incident angle, the energy attenuation information, and the size information into a preset mesh density inference model to obtain the inference density information of the at least one three-dimensional mesh;

[0247] Based on the inference density information of the three-dimensional mesh and / or the standard density information of a preset standard three-dimensional mesh, the anomaly detection result of the three-dimensional mesh is determined.

[0248] In one optional embodiment of this application, determining the anomaly detection result of the three-dimensional mesh based on the inference density information of the three-dimensional mesh and / or the standard density information of a preset standard three-dimensional mesh includes:

[0249] Extract the reasoning feature information from the reasoning density information;

[0250] Based on the inference feature information, at least one three-dimensional mesh is reconstructed to obtain at least one reconstructed three-dimensional mesh;

[0251] Based on at least one of the reconstructed 3D mesh, the standard density information, and the inference density information, the anomaly detection result of the 3D mesh is determined.

[0252] In one optional embodiment of this application, determining the anomaly detection result of the three-dimensional mesh based on at least one of the reconstructed three-dimensional mesh, the standard density information, and the inference density information includes:

[0253] Extract the standard feature information from the standard density information;

[0254] The standard feature information and the inference feature information are compared to determine the feature difference data between the standard three-dimensional mesh and the three-dimensional mesh;

[0255] Based on the reconstructed 3D mesh and / or the feature difference data, the anomaly detection result of the 3D mesh is determined.

[0256] In one optional embodiment of this application, determining the anomaly detection result of the three-dimensional mesh based on the reconstructed three-dimensional mesh and / or the feature difference data includes:

[0257] The three-dimensional mesh is compared with the reconstructed three-dimensional mesh to determine the reconstruction error data of the three-dimensional mesh;

[0258] For any of the three-dimensional meshes, if the reconstruction error data of the three-dimensional mesh is greater than a preset reconstruction error threshold, and / or the feature difference data of the three-dimensional mesh is greater than a preset feature difference threshold, then the three-dimensional mesh is determined to be abnormal.

[0259] In one optional embodiment of this application, the method includes:

[0260] Obtain at least one of the following: the training incident angle of at least one training muon entering a preset training 3D grid, the training energy decay information of the training muon after passing through at least one of the training 3D grids, and the training size information of the training 3D grid;

[0261] Based on at least one of the training incident angle, the training size information, and the training energy decay information, the training standard density information of the training three-dimensional mesh is determined;

[0262] The preset generator is trained using at least one of the training incident angle, the training size information, the training energy decay information, and the training standard density information to obtain the grid density inference model.

[0263] In one optional embodiment of this application, training a preset generator using at least one of the training incident angle, the training size information, the training energy decay information, and the training standard density information to obtain the grid density inference model includes:

[0264] Input at least one of the training incident angle, the training size information, and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh;

[0265] The training inference density information and / or the training standard density information are input into a preset discriminator to obtain similarity information between the training inference density information and the training standard density information.

[0266] Based on the training standard density information and / or the similarity information, it is determined whether the generator meets the preset training objective;

[0267] If the generator satisfies the training objective, then the generator is used as the grid density inference model;

[0268] If the generator does not meet the training objective, the steps of inputting at least one of the training incident angle, the training size information, and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh, and inputting the training inference density information and / or the training standard density information into a preset discriminator to obtain the similarity information between the training inference density information and the training standard density information are repeated until the generator meets the training objective.

[0269] In one optional embodiment of this application, determining whether the generator meets the preset training objective based on the training standard density information and / or the similarity information includes:

[0270] Using the similarity information, the first loss function value of the generator is determined;

[0271] The second loss function value of the discriminator is determined using the first loss function value and / or the training standard density information;

[0272] Based on the first loss function value and / or the second loss function value, determine whether the generator satisfies the training objective.

[0273] In one optional embodiment of this application, determining whether the generator satisfies the training objective based on the first loss function value and / or the second loss function value includes:

[0274] Determine whether the value of the first loss function is less than a preset threshold value for the first function;

[0275] Determine whether the value of the second loss function is greater than a preset threshold value for the second function.

[0276] If the first loss function value is less than the first function value threshold and the second loss function value is greater than the second function value threshold, then the generator is confirmed to meet the training objective.

[0277] In one optional embodiment of this application, determining the training standard density information of the training 3D mesh based on at least one of the training incident angle, the training size information, and the training energy decay information includes:

[0278] Based on the training incident angle and / or the training size information, determine the path length of the training muon in the training 3D mesh;

[0279] Based on the path length and / or the training energy decay information, the training standard density information of the training 3D mesh is determined.

[0280] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0281] The communication interface is used for communication between the aforementioned terminal and other devices.

[0282] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0283] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0284] like Figure 5 As shown, in another embodiment provided in this application, a computer-readable storage medium 501 is also provided, which stores instructions that, when executed on a computer, cause the computer to perform a method for detecting underground facilities as described in the above embodiments.

[0285] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute a method for detecting underground facilities as described in the above embodiments.

[0286] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk (SSD)).

[0287] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0288] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0289] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application are included within the scope of protection of this application.

Claims

1. A method for detecting underground facilities, characterized in that, include: The underground space involved in the underground facility to be inspected is divided into at least one three-dimensional grid; Obtain at least one of the following: the incident angle of at least one muon into the three-dimensional grid, the energy attenuation information of the muon after passing through at least one of the three-dimensional grids, and the size information of the three-dimensional grid; Input at least one of the incident angle, the energy attenuation information, and the size information into a preset mesh density inference model to obtain the inference density information of the at least one three-dimensional mesh; Based on the inference density information of the three-dimensional mesh and / or the standard density information of a preset standard three-dimensional mesh, the anomaly detection result of the three-dimensional mesh is determined.

2. The method according to claim 1, characterized in that, The determination of the anomaly detection result of the three-dimensional mesh based on the inference density information of the three-dimensional mesh and / or the standard density information of a preset standard three-dimensional mesh includes: Extract the reasoning feature information from the reasoning density information; Based on the inference feature information, at least one three-dimensional mesh is reconstructed to obtain at least one reconstructed three-dimensional mesh; Based on at least one of the reconstructed 3D mesh, the standard density information, and the inference density information, the anomaly detection result of the 3D mesh is determined.

3. The method according to claim 2, characterized in that, The step of determining the anomaly detection result of the three-dimensional mesh based on at least one of the reconstructed three-dimensional mesh, the standard density information, and the inference density information includes: Extract the standard feature information from the standard density information; The standard feature information and the inference feature information are compared to determine the feature difference data between the standard three-dimensional mesh and the three-dimensional mesh; Based on the reconstructed 3D mesh and / or the feature difference data, the anomaly detection result of the 3D mesh is determined.

4. The method according to claim 3, characterized in that, The step of determining the anomaly detection result of the three-dimensional mesh based on the reconstructed three-dimensional mesh and / or the feature difference data includes: The three-dimensional mesh is compared with the reconstructed three-dimensional mesh to determine the reconstruction error data of the three-dimensional mesh; For any of the three-dimensional meshes, if the reconstruction error data of the three-dimensional mesh is greater than a preset reconstruction error threshold, and / or the feature difference data of the three-dimensional mesh is greater than a preset feature difference threshold, then the three-dimensional mesh is determined to be abnormal.

5. The method according to claim 1, characterized in that, The method comprises: Obtain at least one of the following: the training incident angle of at least one training muon entering a preset training 3D grid, the training energy decay information of the training muon after passing through at least one of the training 3D grids, and the training size information of the training 3D grid; Based on at least one of the training incident angle, the training size information, and the training energy decay information, the training standard density information of the training three-dimensional mesh is determined; The preset generator is trained using at least one of the training incident angle, the training size information, the training energy decay information, and the training standard density information to obtain the grid density inference model.

6. The method according to claim 5, characterized in that, The step of training a preset generator using at least one of the training incident angle, the training size information, the training energy decay information, and the training standard density information to obtain the grid density inference model includes: Input at least one of the training incident angle, the training size information, and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh; The training inference density information and / or the training standard density information are input into a preset discriminator to obtain similarity information between the training inference density information and the training standard density information. Based on the training standard density information and / or the similarity information, it is determined whether the generator meets the preset training objective; If the generator satisfies the training objective, then the generator is used as the grid density inference model; If the generator does not meet the training objective, the steps of inputting at least one of the training incident angle, the training size information, and the training energy decay information into the generator to obtain the training inference density information of the training 3D mesh, and inputting the training inference density information and / or the training standard density information into a preset discriminator to obtain the similarity information between the training inference density information and the training standard density information are repeated until the generator meets the training objective.

7. The method according to claim 6, characterized in that, The step of determining whether the generator meets the preset training objective based on the training standard density information and / or the similarity information includes: Using the similarity information, the first loss function value of the generator is determined; The second loss function value of the discriminator is determined using the first loss function value and / or the training standard density information; Based on the first loss function value and / or the second loss function value, determine whether the generator satisfies the training objective.

8. The method according to claim 7, characterized in that, The step of determining whether the generator meets the training objective based on the first loss function value and / or the second loss function value includes: Determine whether the value of the first loss function is less than a preset threshold value for the first function; Determine whether the value of the second loss function is greater than a preset threshold value for the second function. If the first loss function value is less than the first function value threshold and the second loss function value is greater than the second function value threshold, then the generator is confirmed to meet the training objective.

9. The method according to claim 5, characterized in that, Determining the standard density information of the training 3D mesh based on at least one of the training incident angle, the training size information, and the training energy decay information includes: Based on the training incident angle and / or the training size information, determine the path length of the training muon in the training 3D mesh; Based on the path length and / or the training energy decay information, the training standard density information of the training 3D mesh is determined.

10. A detection device for underground facilities, characterized in that, include: The grid division module is used to divide the underground space involved in the underground facility to be inspected into at least one three-dimensional grid. The information acquisition module is used to acquire at least one of the following: the incident angle of at least one muon into the three-dimensional grid, the energy attenuation information of the muon after passing through at least one of the three-dimensional grids, and the size information of the three-dimensional grid; The inference density information acquisition module is used to input at least one of the incident angle, the energy attenuation information and the size information into a preset grid density inference model to obtain the inference density information of the at least one three-dimensional grid. The result determination module is used to determine the anomaly detection result of the three-dimensional mesh based on the inference density information of the three-dimensional mesh and / or the standard density information of a preset standard three-dimensional mesh.

11. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; The memory is used to store computer programs; When the processor executes a program stored in the memory, it implements the method as described in any one of claims 1-9.

12. One or more computer-readable media having instructions stored thereon that, when executed by one or more processors, cause the processors to perform the method as described in any one of claims 1-9.

Images