Underground target dynamic electromagnetic detection method and system based on graph neural network

Abstract

The invention provides an underground target dynamic electromagnetic detection method and system based on a graph neural network, and relates to the technical field of detection, and the method comprises the steps: obtaining the pose information of a detector at a plurality of detection points and a secondary field feedback signal; constructing a graph neural network by taking pose information of each detection point location and a secondary field feedback signal as input features; acquiring an underground target detection data set; training the graph neural network by using the underground target detection data set; and performing dynamic electromagnetic detection on the underground target according to the trained graph neural network. According to the method, the information sequence of the dynamic pose and the feedback signal is utilized to help the graph neural network to learn the continuity and time sequence characteristics in the motion process of the detector, so that the dynamic change is captured, and the requirement of real-time detection is met; meanwhile, the graph neural network can effectively process feedback signal deviation caused by movement or inclination of the detector, so that the positioning precision of the underground target is improved.

Description

Technical Field

[0001] The present invention relates to the field of detection technology, and particularly to a method and system for dynamic electromagnetic detection of underground targets based on graph neural networks. Background Art

[0002] Electromagnetic detection of underground targets has wide applications in many fields because it can effectively obtain the spatial information of underground metal targets. However, traditional underground target detection methods face many challenges. With the continuous improvement of detection requirements, the requirements for the accuracy and real-time performance of target detection are gradually increasing. How to improve the detection accuracy under complex geographical environments and dynamic detection conditions has become an urgent problem to be solved.

[0003] Currently, the main process of underground target electromagnetic detection includes moving an electromagnetic detector within the target area and collecting data through the induced secondary field electromagnetic signals. During the entire detection process, the detector usually collects data at different positions and heights. Combining with the electromagnetic response model of the target body, information such as the position, size, and material of the target can be inversely deduced. In terms of data processing, traditional methods mostly adopt offline processing, that is, after the detection is completed, all the collected data are uniformly analyzed and processed, and then the relevant parameters of the underground target are deduced.

[0004] Although the above traditional methods can achieve certain detection effects under certain conditions, they still have many limitations. First, due to the attitude changes of the detector during the detection process, the secondary field induction data will be affected by factors such as the movement and tilt of the detector, resulting in deviations in the measured data, which in turn affects the positioning accuracy of the target. In addition, the temporal characteristics between the detection data are usually ignored during data processing, and the dynamic information during the detection process is not fully utilized, which makes the detection results unable to reflect the real-time changes of the target and difficult to meet the requirements of real-time detection. Summary of the Invention

[0005] In order to solve the technical problems that in the existing methods, due to the attitude changes of the detector during the detection process, the secondary field induction data will be affected by the movement and tilt of the detector, resulting in deviations in the measured data, which in turn affects the positioning accuracy of the target; in addition, the temporal characteristics between the detection data are usually ignored during data processing, and the dynamic information during the detection process is not fully utilized, which makes the detection results unable to reflect the real-time changes of the target and difficult to meet the requirements of real-time detection, the present invention provides a method and system for dynamic electromagnetic detection of underground targets based on graph neural networks.

[0006] The technical solutions provided by the embodiments of the present invention are as follows:

[0007] First aspect:

[0008] A dynamic electromagnetic detection method for underground targets based on a graph neural network provided by an embodiment of the present invention includes:

[0009] S1: Obtain the pose information of the detector at multiple detection points and the secondary field feedback signal;

[0010] S2: Construct a graph neural network with the pose information and secondary field feedback signal of each detection point as input features;

[0011] S3: Obtain an underground target detection data set;

[0012] S4: Use the underground target detection data set to train the graph neural network;

[0013] S5: Perform dynamic electromagnetic detection of underground targets according to the trained graph neural network.

[0014] Second aspect:

[0015] A dynamic electromagnetic detection system for underground targets based on a graph neural network provided by an embodiment of the present invention includes:

[0016] A processor;

[0017] A memory, on which computer-readable instructions are stored, and when the computer-readable instructions are executed by the processor, the dynamic electromagnetic detection method for underground targets based on a graph neural network as described in the first aspect is implemented.

[0018] Third aspect:

[0019] A computer-readable storage medium provided by an embodiment of the present invention, on which a computer program is stored, and when the program is executed by a processor, the dynamic electromagnetic detection method for underground targets based on a graph neural network as described in the first aspect is implemented.

[0020] The beneficial effects brought by the technical solution provided by the embodiment of the present invention at least include:

[0021] (1) In the present invention, the pose information of the detector at multiple detection points and the secondary field feedback signal are obtained, and a graph neural network is constructed with the pose information and secondary field feedback signal of each detection point as input features. The graph neural network can effectively process the feedback signal deviation caused by the movement or tilt of the detector, thereby improving the positioning accuracy of underground targets.

[0022] (2) In the present invention, an underground target detection data set is used to train a graph neural network. Based on the trained graph neural network, dynamic electromagnetic detection of underground targets is carried out. The graph neural network can capture the relationships between nodes. By using the information sequence of dynamic poses and feedback signals, the network is helped to learn the continuity and temporal characteristics during the movement of the detector, so as to capture dynamic changes and meet the requirements of real-time detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0024] Figure 1 It is a schematic flowchart of a method for dynamic electromagnetic detection of underground targets based on a graph neural network provided by an embodiment of the present invention;

[0025] Figure 2 It is a schematic structural diagram of a system for dynamic electromagnetic detection of underground targets based on a graph neural network provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The following will describe the technical solutions in the present invention with reference to the drawings.

[0027] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, the use of the word "example" is intended to present concepts in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0028] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0029] In the embodiments of the present invention, sometimes subscripts such as W 1 may be miswritten as non-subscript forms such as W1. When the difference is not emphasized, the meanings they express are the same.

[0030] To make the technical problems, technical solutions, and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0031] Referring to the attached drawings of the specification Figure 1 , a schematic flowchart of a method for dynamic electromagnetic detection of underground targets based on a graph neural network provided by an embodiment of the present invention is shown.

[0032] An embodiment of the present invention provides a method for dynamic electromagnetic detection of underground targets based on a graph neural network. This method can be implemented by a device for dynamic electromagnetic detection of underground targets based on a graph neural network, and this device for dynamic electromagnetic detection of underground targets based on a graph neural network can be a terminal or a server. The processing flow of the method for dynamic electromagnetic detection of underground targets based on a graph neural network can include the following steps:

[0033] S1: Obtain the pose information of the detector at multiple detection points and the secondary field feedback signal.

[0034] Among them, the secondary field feedback signal refers to the reflection or scattering signal received by the detector after emitting electromagnetic waves (primary field), which is generated by the interaction between the electromagnetic waves and the underground targets. These signals reflect the electromagnetic characteristics and positions of the targets.

[0035] It should be noted that the dynamic detection path of the underground targets consists of N detection points. The detector moves sequentially at the N detection points and obtains the secondary field feedback signal output by the corresponding sensor at each point.

[0036] In the present invention, moving sequentially at multiple detection points and collecting the secondary field feedback signal can improve the detection and recognition ability of the targets and meet the detection requirements in dynamic and complex environments.

[0037] S2: Use the pose information of each detection point and the secondary field feedback signal as input features to construct a graph neural network.

[0038] Among them, graph neural networks (GNNs) are a type of deep learning model for processing graph-structured data. Graph-structured data widely exists in many practical applications, such as social networks, molecular structures, knowledge graphs, and transportation networks. The main purpose of GNNs is to perform effective node classification, graph classification, link prediction, etc. tasks by using the structural information of the graph.

[0039] In a possible implementation manner, S2 specifically includes:

[0040] S201: Use each detection point as a node of the graph neural network, and construct a corresponding node feature matrix for each detection point according to the pose information and the secondary field feedback signal at each detection point.

[0041] Among them, the node feature matrix is ​​a core concept in the graph neural network, which is used to represent the feature information of each node in the graph. In the graph neural network, the feature matrix of each node provides the attributes and related data of the node and is the input of the graph structure data.

[0042] In a possible implementation, S201 specifically includes:

[0043] S2011: splicing the position information of the current detection point obtained by each detector with the secondary field feedback signal to obtain a one-dimensional vector corresponding to each detector.

[0044] S2012: vertically concatenate the one-dimensional vectors corresponding to each detector to obtain a node feature matrix.

[0045] Specifically, each detection point is used as a node of the graph neural network, and the node initial feature X consists of two parts: posture information and the corresponding Bs feedback signal. The posture information is the spatial coordinate D of the detection point. x , D y , D z and attitude angle D γ , D θ , The secondary field feedback signal is composed of array signals from different sensors. The array signal value output by each sensor is preprocessed into a one-dimensional vector of size 1*Q. The signal obtained by the sensor is the Q-time sampling of the secondary field feedback signal within the detection time T (B s,1 ,B s,2 ,…,B s,Q ), which can characterize the intensity and attenuation characteristics of the secondary field feedback signal. Assuming the number of sensors on the detector is M, the posture information is spliced ​​with the array signal vector of each sensor, and M one-dimensional vectors of size 1*(6+Q) are obtained. These M one-dimensional vectors are spliced ​​vertically to obtain a matrix of size M*(6+Q), which is the node initial feature X_(M×(6+Q)). The node initial feature X contains the posture information of the detector at the detection point and the secondary field feedback signals corresponding to each sensor. By generating a node initial feature for each detection point, N node initial features can be obtained.

[0046] In the present invention, by integrating the detection points and sensor feedback signals into a feature matrix, the network can make comprehensive use of various information for learning and prediction, further optimizing the performance of the model. At the same time, by splicing the posture information of the detection points and the secondary field feedback signals into the node feature matrix, the spatial position and electromagnetic response of the target are comprehensively considered. This multi-dimensional feature integration provides a more comprehensive description of the target and helps to improve the accuracy of detection.

[0047] S202: Construct corresponding adjacency feature matrices for each detection point according to the node feature matrix.

[0048] Among them, the adjacency feature matrix is an important concept in graph neural networks, used to represent the relationships or connection methods between nodes in a graph. Its main role is to help graph neural networks capture the correlation information between nodes and support the effective processing and analysis of graph-structured data.

[0049] In a possible implementation, S202 specifically includes:

[0050] S2021: Input the node feature matrix of the current node into the convolutional layer to extract the feature information of the current node itself.

[0051] S2022: Input the feature information of the current node itself into the activation function to obtain the node feature:

[0052] Y = σ(W a ·X + b a )

[0053] Among them, Y represents the node feature, σ represents the activation function, X represents the node feature matrix, W a represents the weight matrix of its own feature information, and b a represents the bias term of its own feature information.

[0054] S2023: Generate a relationship matrix by multiplying the transpose of the node feature of the current node with the node feature of the i-th node, input the relationship matrix into the convolutional layer, extract the correlation information between the current node and the i-th node, and perform standard normalization and activation operations on the correlation information to obtain the updated relationship matrix:

[0055]

[0056] Among them, represents the updated relationship matrix, σ represents the activation function, BN represents the normalization layer, Y 1 represents the node feature of the current node, Y i represents the node feature of the i-th node, () T represents the transpose matrix, W b represents the node relationship modeling weight matrix, and b b represents the node relationship modeling bias term.

[0057] S2024: Use one-dimensional convolution to compress the dimension of the updated relationship matrix to generate the adjacency feature matrix:

[0058]

[0059] Among them, represents the adjacency feature matrix between the current node and the i-th node, represents the convolutional kernel weight matrix, represents the relationship matrix, represents the bias term in the one-dimensional convolution.

[0060] For example, taking the first node as an example, the construction process of the adjacency feature matrix is shown:

[0061] To extract the feature information of each node itself, first output the node feature information by passing the initial feature X of the node through a convolutional layer, and then input this feature into the activation function sigmoid to remove redundant information, obtaining the node feature Y. This operation does not change the feature size, which is still M*(6 + Q). Let the node feature Y of the i-th node be Y i , generate a relationship matrix with a dimension of M*M by multiplying the transpose of the node feature of the first node with the node feature of the i-th node; then input this relationship matrix into a brand-new convolutional layer to extract the correlation information between the two; then perform standard normalization and activation operations on the correlation information in sequence to obtain a refined relationship matrix with a dimension of M*M To facilitate the subsequent input into the graph convolutional layer to generate new node information, one-dimensional convolution is used for dimension compression to generate a with a dimension of 1*M, which is the edge vector representation between the first node and the i-th node. With the same operation, calculate the correlation relationships between the first node and the remaining N - 1 nodes respectively, and then N - 1 edge vector representations of the first node are obtained.

[0062] In the present invention, by calculating the relationship matrix, the interaction and correlation between nodes can be effectively captured. This helps the network understand the complex relationships between nodes and improves the expressiveness of the graph neural network. At the same time, this method ensures that the graph neural network can make full use of the feature information of nodes and edges, thereby improving the overall performance and accuracy of the underground target detection system.

[0063] S203: Combine the node feature matrix and the adjacency feature matrix using the graph convolutional layer, and output the final node features.

[0064] In a possible implementation manner, S203 specifically includes:

[0065] S2031: Concatenate the dimensions of the adjacency feature matrices of multiple nodes to obtain the global edge features.

[0066] Among them, Global Edge Features is an important concept in graph neural networks for representing the feature information of edges in a graph. When processing graph-structured data, Global Edge Features are used to capture the comprehensive information of all edges in the graph, which helps the model better understand the relationships between nodes and the overall structure of the graph.

[0067] In the present invention, the Global Edge Features provide the connection information between nodes and the characteristics of the edges, and effectively capture these relationships through graph convolution operations, which helps to construct more accurate node representations.

[0068] S2032: Multiply the Global Edge Features by the weight matrix to obtain an intermediate matrix:

[0069] Z (6+Q)*M = W (6+Q)*(N-1) · F (N-1)*M

[0070] Among them, Z (6+Q)*M represents an intermediate matrix with dimensions (6 + Q) * M, W (6+Q)*(N-1) represents a weight matrix with dimensions (6 + Q) * (N - 1), and F (N-1)*M represents the Global Edge Features with dimensions (N - 1) * M.

[0071] S2033: Multiply the node features by the intermediate matrix to obtain a feature matrix containing the node's own feature information and the relationship information between nodes, and perform feature activation through an activation function. Multiply the node features by their own transpose to obtain the self-feature matrix. Perform element-wise summation of the activated features and the self-feature matrix to obtain the intermediate node features:

[0072]

[0073] Among them, Y M ′ *M represents the intermediate node features with dimensions M * M, Y M*(6+Q) represents the node features with dimensions (6 + Q) * M, () T represents the transpose matrix, σ represents the activation function, and Z (6+Q)*M represents the intermediate matrix with dimensions (6 + Q) * M.

[0074] In the present invention, combining the node features and the intermediate matrix can effectively fuse the relationship information between nodes and the features of the nodes themselves. By performing element-wise summation and activation processing on the features, more rich and refined node feature representations can be generated.

[0075] S2034: Input the intermediate node features into a value convolution layer and a Relu activation layer in sequence, and output the final node features:

[0076] Z M*M = Relu(We·Y' M*M + b e )

[0077] Among them, Z M*M represents the final node feature with a dimension of M*M, Relu() represents the Relu activation function, and W e represents the node feature aggregation weight matrix, and b e represents the node feature aggregation bias term.

[0078] Optionally, the same operation is performed on N detection points, and multiple node features can be obtained.

[0079] In the present invention, by combining the node feature matrix and the adjacency feature matrix, the self-features of the nodes and the relationship information between the nodes can be comprehensively considered. This comprehensive feature representation enables the graph neural network to more comprehensively understand the characteristics of the nodes and their positions in the graph.

[0080] Furthermore, through detailed feature processing and fusion, rich input features are provided for the graph neural network, significantly improving the ability to process graph-structured data.

[0081] S204: According to the final node features, perform one-dimensional graph convolution operations and splicing operations in sequence to obtain the graph global feature.

[0082] Among them, the one-dimensional graph convolution operation (1D Graph Convolution) is an important operation in graph neural networks (GNNs). The one-dimensional graph convolution is mainly used to process the relationship between nodes and edges in graph data, especially when dealing with feature propagation and update in the graph.

[0083] In a possible implementation manner, S204 specifically includes:

[0084] S2041: Pass the final node features through the adjacency feature matrix and the graph convolution operation for the second time to obtain the second node features.

[0085] S2042: Perform one-dimensional convolution operations on the second node features to obtain the third node features.

[0086] S2043: Splice multiple third node features to obtain the graph global feature containing the underground target feature information obtained from all detection points:

[0087] H N*M = concat[G 1 , G 2 , …, G N

[0088] Among them, H​N*M Represents the global graph feature of dimension N*M, where N represents the total number of nodes, and G i Represents the i-th third node feature, i = 1, 2... N, and concat[] represents the concatenation operation.

[0089] In the present invention, by subjecting the final node feature to a one-dimensional graph convolution and a concatenation operation, the feature information of each node can be effectively integrated. This enables the model to comprehensively capture the features of each node in the graph, thereby obtaining richer global information. At the same time, the global feature can be used as the feature input of the graph for graph classification tasks, which can accurately capture the overall information of the graph and perform classification or regression tasks to improve the detection accuracy.

[0090] S205: Output the detection result of the underground target according to the global graph feature.

[0091] In a possible implementation manner, S205 is specifically:

[0092] Input the global graph feature into the convolutional layer for feature extraction, and sequentially input the extracted features into the fully connected layer to output the detection result of the underground target:

[0093] H N*M = Fc(σ(W f ·H N*M + b f ))

[0094] where H N*M represents the global graph feature of dimension N*M, σ represents the activation function, Fc() represents the fully connected layer operation, W f represents the output result mapping weight matrix, and b f represents the output result mapping bias term.

[0095] In a possible implementation manner, the detection result includes: centroid spatial coordinates, material, and morphology.

[0096] In a possible implementation manner, the centroid spatial coordinates are specifically: (T x , T y , T z ), where T x represents the spatial position coordinate of the target on the x-axis, T y represents the spatial position coordinate of the target on the y-axis, and T z represents the spatial position coordinate of the target on the z-axis.

[0097] The categories of materials include: no target, metal target, and non-metal target.

[0098] The morphology includes: the bottom radius, height, pitch angle, and yaw angle of the cylindrical target.

[0099] Specifically, a result output module is constructed. The result output module for the underground target detection task consists of three modules: a centroid positioning module, a material judgment module, and a morphology output module. As shown in Equation 8, the construction process of the centroid positioning module is as follows: The global feature H of the graph is N*M input into a convolutional layer for feature extraction, and then successively input into a fully connected layer. The output dimension of this fully connected layer is set to 3, which corresponds to the centroid space coordinates of the target (T x , T y , T z ).

[0100] The construction processes of the material judgment module and the morphology output module are similar to that of the centroid positioning module. The difference is that the output dimension of the material judgment module is set to 1, which corresponds to three categories: (no target, metal target, non-metal target); the output dimension of the morphology output module is set to 4, which corresponds to the bottom radius, height, pitch angle, and yaw angle of the cylindrical target.

[0101] In the present invention, by inputting the global feature of the graph into the convolutional layer and the fully connected layer, accurate features of the target can be extracted, thereby accurately calculating the centroid space coordinates of the underground target and improving the positioning accuracy and recognition ability of the underground target.

[0102] S206: Complete the construction of the graph neural network.

[0103] S3: Obtain the underground target detection dataset.

[0104] Optionally, the way to obtain the underground target detection dataset is as follows:[[]]

[0105] Take a batch of targets with different materials and shapes as detection samples, place them in different underground positions in different postures, and record the material, shape, and pose information of the targets, that is, the target information. Use the detector to conduct multiple underground target detection experiments according to different detection points, and record the corresponding detector pose information and the array signals output by different sensors, that is, the detector information, to obtain the underground target detection dataset.

[0106] In the present invention, the target detection datasets with different materials and shapes ensure that the dataset can cover a wide range of target types and features. This diversity enables the dataset to more comprehensively reflect different situations in actual detection. At the same time, placing the targets in different underground positions in different postures to simulate the target distribution in the real environment enhances the representativeness and authenticity of the data.

[0107] S4: Use the underground target detection dataset to train the graph neural network.

[0108] Specifically, based on the underground target detection dataset, the graph neural network is trained until the loss function converges, and the training of the graph neural network is completed.

[0109] Among them, the loss function is a function used in machine learning and deep learning to measure the gap between the model prediction and the actual target. It is the core component of training the model. By optimizing the loss function, the model parameters are adjusted to improve the prediction accuracy of the model.

[0110] In the present invention, through training based on the underground target detection dataset, the model can learn to extract useful features from the detection data, thereby improving the prediction accuracy of underground targets. At the same time, the loss function is used to quantify the gap between the model prediction and the actual target. Optimizing the loss function can make the prediction result of the model closer to the true value and improve the overall accuracy.

[0111] S5: According to the trained graph neural network, perform dynamic electromagnetic detection on underground targets.

[0112] The beneficial effects brought by the technical solution provided by the embodiment of the present invention at least include:

[0113] (1) In the present invention, the pose information of the detector at multiple detection points and the secondary field feedback signal are obtained. Using the pose information and the secondary field feedback signal of each detection point as input features, a graph neural network is constructed. Through the graph neural network, the feedback signal deviation caused by the movement or tilt of the detector can be effectively processed, thereby improving the positioning accuracy of underground targets.

[0114] (2) In the present invention, using the underground target detection dataset, the graph neural network is trained. According to the trained graph neural network, dynamic electromagnetic detection is performed on underground targets. Through the graph neural network, the relationship between nodes can be captured, and the information sequence of the dynamic pose and the feedback signal is used to help the network learn the continuity and timing characteristics during the movement of the detector, so as to capture dynamic changes and meet the requirements of real-time detection.

[0115] Refer to the attached Figure 2 illustrates the structural schematic diagram of a dynamic electromagnetic detection system for underground targets based on a graph neural network provided by the present invention.

[0116] A dynamic electromagnetic detection system 20 for underground targets based on a graph neural network provided by an embodiment of the present invention is applied to the above-mentioned dynamic electromagnetic detection method for underground targets based on a graph neural network, and includes:

[0117] A processor 201;

[0118] A memory 202 stores computer-readable instructions thereon. When the computer-readable instructions are executed by the processor 201, the method for dynamically detecting underground targets based on a graph neural network as described in the first aspect is implemented.

[0119] The underground target dynamic electromagnetic detection system 20 provided by the present invention can execute the above-mentioned method for dynamically detecting underground targets based on a graph neural network and achieve the same or similar technical effects. To avoid repetition, the present invention will not elaborate further.

[0120] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include:

[0121] (1) In the present invention, the pose information and the secondary field feedback signal of the detector at multiple detection points are obtained. Using the pose information and the secondary field feedback signal of each detection point as input features, a graph neural network is constructed. The graph neural network can effectively process the feedback signal deviation caused by the movement or tilt of the detector, thereby improving the positioning accuracy of underground targets.

[0122] (2) In the present invention, the graph neural network is trained using an underground target detection data set. According to the trained graph neural network, dynamic electromagnetic detection of underground targets is performed. The graph neural network can capture the relationships between nodes and utilize the information sequence of the dynamic pose and the feedback signal to help the network learn the continuity and temporal characteristics during the movement of the detector, thereby capturing dynamic changes and meeting the requirements of real-time detection.

[0123] It should be understood that the processor in the embodiments of the present invention may be a central processing unit (CPU), and this processor may also be other general-purpose processors, 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, discrete hardware components, etc. The general-purpose processor may be a microprocessor or this processor may also be any conventional processor, etc.

[0124] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0125] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. 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 by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, or magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0126] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood with reference to the context before and after.

[0127] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0128] It should be understood that in various embodiments of the present invention, the magnitudes of the sequence numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0129] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled professionals may use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0130] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated herein.

[0131] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0132] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0133] In addition, the functional units in each embodiment of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0134] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0135] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for dynamically detecting underground targets based on a graph neural network as described in the method embodiment.

[0136] The computer-readable storage medium provided by the present invention can implement the steps and effects of the method for dynamically detecting underground targets based on a graph neural network in the above method embodiment. To avoid repetition, the present invention will not elaborate further.

[0137] The beneficial effects brought by the technical solution provided by the embodiment of the present invention at least include:

[0138] (1) In the present invention, the pose information of the detector at multiple detection points and the secondary field feedback signal are obtained. Using the pose information of each detection point and the secondary field feedback signal as input features, a graph neural network is constructed. Through the graph neural network, the deviation of the feedback signal caused by the movement or tilt of the detector can be effectively processed, thereby improving the positioning accuracy of underground targets.

[0139] (2) In the present invention, the graph neural network is trained using an underground target detection data set. According to the trained graph neural network, dynamic electromagnetic detection of underground targets is performed. Through the graph neural network, the relationship between nodes can be captured, and the information sequence of the dynamic pose and feedback signal is used to help the network learn the continuity and temporal characteristics during the movement of the detector, thereby capturing dynamic changes and meeting the requirements of real-time detection.

[0140] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

[0141] The following points need to be explained:

[0142] (1) The accompanying drawings of the embodiments of the present invention only relate to the structures involved in the embodiments of the present invention, and other structures can refer to the general design.

[0143] (2) For clarity, in the accompanying drawings used to describe the embodiments of the present invention, the thickness of layers or regions is enlarged or reduced, that is, these drawings are not drawn to actual scale. It can be understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element can be "directly" on or under the other element or there can be intervening elements.

[0144] (3) Without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0145] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for dynamic electromagnetic detection of underground targets based on graph neural network, characterized in that: include: S1: Obtain the position information of the detector at multiple detection points and the secondary field feedback signal; S2: Using the posture information of each detection point and the secondary field feedback signal as input features, constructing a graph neural network; S3: Acquire underground target detection dataset; S4: Using the underground target detection dataset, training the graph neural network; S5: Perform dynamic electromagnetic detection of underground targets based on the trained graph neural network.

2. The underground target dynamic electromagnetic detection method based on graph neural network according to claim 1 is characterized in that: The S2 specifically includes: S201: Taking each detection point as a node of the graph neural network, and constructing a corresponding node feature matrix for each detection point according to the posture information and secondary field feedback signal at each detection point; S202: constructing a corresponding adjacency feature matrix for each of the detection points according to the node feature matrix; S203: Use the graph convolution layer to combine the node feature matrix and the adjacency feature matrix to output the final node features; S204: performing one-dimensional graph convolution operation and concatenation operation in sequence according to the final node features to obtain graph global features; S205: outputting the detection result of the underground target according to the global features of the graph; S206: Complete the construction of the graph neural network.

3. The underground target dynamic electromagnetic detection method based on graph neural network according to claim 2 is characterized in that: The S201 specifically includes: S2011: splicing the position information of the current detection point obtained by each detector with the secondary field feedback signal to obtain a one-dimensional vector corresponding to each detector; S2012: vertically concatenate the one-dimensional vectors corresponding to each detector to obtain a node feature matrix.

4. The underground target dynamic electromagnetic detection method based on graph neural network according to claim 3 is characterized in that: The S202 specifically includes: S2021: Inputting the node feature matrix of the current node into the convolution layer to extract feature information of the current node itself; S2022: Input the feature information of the current node itself into the activation function to obtain the node feature: Y=σ(W a ·X+b a ) Among them, Y represents the node feature, σ represents the activation function, X represents the node feature matrix, and W a represents the weight matrix of its own feature information, b a Represents the bias item of its own feature information; S2023: Generate a relationship matrix by multiplying the node feature of the current node by the transpose of the node feature of the i-th node, input the relationship matrix into the convolution layer, extract the correlation information between the current node and the i-th node, and perform standard normalization and activation operations on the correlation information to obtain an updated relationship matrix: in, represents the updated relationship matrix, σ represents the activation function, BN represents the normalization layer, Y1 represents the node feature of the current node, and Y i Represents the node feature of the i-th node, () T represents the transposed matrix, W b represents the node relationship modeling weight matrix, b b Represents the node relationship modeling bias; S2024: Use one-dimensional convolution to compress the dimension of the updated relationship matrix to generate an adjacency feature matrix: in, represents the adjacency feature matrix between the current node and the i-th node, represents the convolution kernel weight matrix, represents the relationship matrix, Represents the bias term in one-dimensional convolution.

5. The underground target dynamic electromagnetic detection method based on graph neural network according to claim 4 is characterized in that: The S203 specifically includes: S2031: Dimensionally concatenate the adjacency feature matrices of multiple nodes to obtain global edge features; S2032: Multiply the global edge feature and the weight matrix to obtain an intermediate matrix: Z (6+Q)*M =W (6+Q)*(N-1) ·F (N-1)* M Among them, Z (6+Q)*M represents the intermediate matrix with dimension (6+Q)*M, W (6+Q)*(N-1) represents a weight matrix of size (6+Q)*(N-1), F (N-1)*M Represents a global edge feature of size (N-1)*M; S2033: Multiply the node feature with the intermediate matrix to obtain a feature matrix containing the node's own feature information and the relationship information between nodes, and activate the feature through an activation function, multiply the node feature with its own transpose to obtain its own feature matrix; perform element-by-element summation of the activated feature and the own feature matrix to obtain the intermediate node feature: Among them, Y M ′ *M Represents the intermediate node features with dimension M*M, Y M*(6+Q) Represents node features with dimension (6+Q)*M, () T represents the transposed matrix, σ represents the activation function, Z (6+Q)*M Represents an intermediate matrix with dimension (6+Q)*M; S2034: Input the intermediate node features into the convolution layer and the Relu activation layer in sequence, and output the final node features: WITH M*M =Relu(W e ·Y′ M*M +b e ) Among them, Z M*M represents the final node feature with dimension M*M, Relu() represents the Relu activation function, W e represents the node feature aggregation weight matrix, b e Represents the node feature aggregation bias.

6. The underground target dynamic electromagnetic detection method based on graph neural network according to claim 5 is characterized in that: The S204 specifically includes: S2041: The final node feature is subjected to an adjacency feature matrix and a graph convolution operation for a second time to obtain a second node feature; S2042: performing a one-dimensional convolution operation on the second node feature to obtain a third node feature; S2043: Splicing the plurality of third node features to obtain a graph global feature containing underground target feature information obtained from all detection points: H N*M =concat[G 1 ,G 2 ,…,G N ] Among them, H N*M Represents the global features of a graph with dimension N*M, where N represents the total number of nodes and G i Represents the i-th third node feature, i=1,2…N, and concat[] represents the concatenation operation.

7. The underground target dynamic electromagnetic detection method based on graph neural network according to claim 6 is characterized in that: The S205 is specifically as follows: The global features of the graph are input into the convolutional layer for feature extraction, and the extracted features are input into the fully connected layer in turn to output the detection results of underground targets: H N*M =Fc(σ(W f ·H N*M +b f )) Among them, H N*M represents the global features of the graph with dimension N*M, σ represents the activation function, Fc() represents the fully connected layer operation, and W f Represents the output result mapping weight matrix, b f Represents the output result mapping bias item.

8. The underground target dynamic electromagnetic detection method based on graph neural network according to claim 7 is characterized in that: The detection results include: centroid space coordinates, material and shape.

9. The underground target dynamic electromagnetic detection method based on graph neural network according to claim 8 is characterized in that: The mass center spatial coordinates are specifically: x , T y , T z ), T x Represents the spatial position coordinate of the target on the x-axis, T y Represents the spatial position coordinate of the target on the y-axis, T z Indicates the spatial position coordinates of the target on the z-axis; The material categories include: no target, metal target and non-metal target; The shape includes: the bottom radius, height, pitch angle and yaw angle of the cylindrical target.

10. A dynamic electromagnetic detection system for underground targets based on graph neural network, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method for dynamic electromagnetic detection of underground targets based on a graph neural network as described in any one of claims 1 to 9 is implemented.

Images