Method, equipment and medium for interpreting geological radar intelligent data by integrating multi-source data

Abstract

The invention discloses a geological radar intelligent data interpretation method, device and medium for fusing multi-source data, and relates to the technical field of geological radar intelligent data interpretation. The method comprises: acquiring detection data of a geological radar in a test area, and inputting the detection data into an interpretation model for processing to obtain first velocity data of electromagnetic waves in an underground medium in the test area, wherein the detection data comprises a-scan data; using the first velocity data and historical seismic data of the test area, calculating correlation coefficients between the first velocity data and each data in the historical seismic data; based on each correlation coefficient, performing multi-source fusion processing on each data in the first velocity data and the historical seismic data, and determining the multi-source fusion result as a target propagation velocity; through the multi-source fusion processing of multiple data, the propagation velocity of the electromagnetic wave finally obtained can be made more accurate, so as to provide a reference for on-site construction, etc.

Description

Technical Field

[0001] The present invention relates to the technical field of geological radar intelligent data interpretation, and more specifically, to a geological radar intelligent data interpretation method, equipment and medium that fuses multi-source data. Background Art

[0002] As the main engineering geophysical exploration method currently used for urban road disease detection, geological radar has the advantages of high resolution, non-destructive detection, continuous measurement, short data acquisition and processing time, and real-time imaging in the detection of urban road diseases such as underground cavities. In addition, geological radar can also help engineers understand the characteristics and structure of map layers. It has been widely used in the detection of underground cavity diseases in roads.

[0003] Although geological radar has many advantages, it also has certain limitations. Due to the complexity of the underground environment and the propagation mechanism of electromagnetic waves, it is difficult for the target echo profile to intuitively reflect the underground detection object. The interpretation mainly relies on the personal experience of the data interpreter. The data interpretation results are usually very subjective and may lead to erroneous interpretation results. As geological radar is widely used in various fields, the processing, analysis and interpretation of geological radar data and images have become increasingly important. At the same time, when interpreting geological radar data, there is often no data reference and only this single data is interpreted, which also has the problem of low interpretation accuracy. Summary of the invention

[0004] The purpose of the present invention is to provide a geological radar intelligent data interpretation method, device and medium that fuses multi-source data to solve the problems existing in the above-mentioned background technology.

[0005] The above technical objectives of the present invention are achieved through the following technical solutions:

[0006] In the first aspect, the present application provides a method for intelligent data interpretation of geological radar by fusing multi-source data, including the following specific steps:

[0007] Construct an initial model based on deep learning. The input of the initial model is the a-scan data of the geological radar, and the output of the initial model is the propagation speed of electromagnetic waves.

[0008] The initial model is trained using preset training samples until the initial model reaches a preset training end condition, and the initial model that reaches the training end condition is determined as the interpretation model;

[0009] Acquire the detection data of the geological radar in the area to be measured, and input the detection data into the interpretation model for processing to obtain the first velocity data of the electromagnetic wave in the underground medium in the area to be measured, the detection data includes a-scan data;

[0010] Using the first velocity data and the historical seismic data of the area to be measured, a correlation coefficient between each data in the first velocity data and the historical seismic data is calculated;

[0011] Based on each correlation coefficient, the first velocity data and each data in the historical seismic data are subjected to multi-source fusion processing, and the multi-source fusion result is determined as the target propagation velocity.

[0012] The beneficial effects of the present invention are as follows: in this scheme, firstly, the initial model based on deep learning is trained using the preset training samples, and when the training is completed, the initial model is determined as the interpretation model, and the interpretation model is mainly used to interpret the propagation speed of electromagnetic waves from the a-scan data of radar data. The propagation speed of electromagnetic waves is the propagation speed of electromagnetic waves in underground structures composed of different media. The propagation speed can be used to analyze the water content and other information in the underground structure, which can better meet the needs of radar in the field of water content detection; secondly, the detection data of the geological radar in the test area is obtained, and the detection data includes a-scan data. an data, and then the a-scan data in the detection data is input into the interpretation model for processing, so as to obtain the first velocity data predicted by the interpretation model, that is, the propagation velocity of electromagnetic waves in the area to be measured; finally, the various data in the seismic data are used with the first velocity data for multi-source fusion processing, and finally the accurate target propagation velocity of electromagnetic waves in the area to be measured is obtained; the seismic data mainly uses the propagation velocity data of longitudinal waves, shear waves and surface waves, that is, the same type of data; the multi-source fusion processing of multiple data can make the final electromagnetic wave propagation velocity more accurate, which can provide a reference for on-site construction, etc.

[0013] Based on the above technical solution, the present invention can also be improved as follows.

[0014] Furthermore, the above interpretation model is obtained in the following way:

[0015] Acquire simulation data of geological radar, the simulation data including a-scan data, geological simulation model data and simulation speed data, add preset noise to the a-scan data, and determine the simulation data with preset noise added as training samples;

[0016] The a-scan data in the training sample is input into the initial model for processing to obtain the predicted speed, and the loss function of the initial model is calculated based on the predicted speed corresponding to the a-scan data and the corresponding simulation speed data;

[0017] If the loss function does not exceed the threshold or the number of training iterations reaches the maximum number of iterations, the initial model is determined as the interpretation model, and the preset training end condition is that the loss function does not exceed the threshold or the number of training iterations reaches the maximum number of iterations.

[0018] The beneficial effect of adopting the above further scheme is that simulation data can be obtained through simulation experiments, and the corresponding simulation velocity data can be obtained by simulating the propagation of geological radar in the geological simulation model. The geological simulation model data includes the geological structure and component composition in the geological simulation model, and this data can be used as real data to effectively train the initial model.

[0019] Furthermore, the above method also includes: during the initial model training process, the weight parameters of the initial model are iteratively updated in the following manner:

[0020] ;

[0021] In the formula, Indicates t The weight parameters of the initial model at this iteration, Indicates t -1 is the weight parameter of the initial model at this iteration, Indicates t The learning rate of the initial model at this iteration, Indicates t -1 The value of the loss function of the initial model at this iteration.

[0022] The beneficial effect of adopting the above further scheme is that the static learning rate in the deep learning model is often not flexible enough and easily causes the model to fall into the local minimum. Therefore, in order to obtain better model performance, the learning rate can be dynamically adjusted during training to avoid problems such as premature convergence of the model or oscillation caused by excessive learning rate; the learning rate can be reduced by the cosine function, which can also avoid parameters such as weight parameters from falling into the local minimum and reduce the time loss of model training.

[0023] Furthermore, the multi-source fusion result of the multi-source fusion processing is specifically as follows:

[0024] ;

[0025] In the formula, represents the multi-source fusion result of multi-source fusion processing, Indicates the first speed data, Indicates the first data in the historical earthquake data. Representation data The correlation coefficient with the first velocity data, Representation data The exponential weight of n Represents the total number of data involved in multi-source fusion in historical seismic data.

[0026] Furthermore, the index weights of each data in the above historical earthquake data are obtained in the following way:

[0027] Based on the preset initial value of the exponential weight of each data, the first velocity data is preliminarily integrated with each data in the historical seismic data to obtain a first result;

[0028] Using the first result, calculate the loss values ​​corresponding to the exponential weights of each data in the historical earthquake data, and recalculate the corresponding exponential weights using the loss values ​​of each data in the historical earthquake data;

[0029] The recalculated exponential weights of each data are fused with the first velocity data and iterated until the loss values ​​corresponding to each exponential weight reach the iteration end condition. The exponential weights that reach the iteration end condition are used to perform multi-source fusion of each data in the historical seismic data and the first velocity data to obtain the multi-source fusion result.

[0030] Furthermore, the above correlation coefficient is specifically:

[0031] ;

[0032] In the formula, Indicates the historical earthquake data n The correlation coefficient between the data and the first speed data, Indicates the coordinates of the first velocity data in the preset spatial rectangular coordinate system ( x, y, z ), Indicates the historical earthquake data n The coordinates of each data in the preset spatial rectangular coordinate system ( x, y, z ) at .

[0033] Furthermore, the loss values ​​corresponding to the exponential weights of the various data in the above historical earthquake data are specifically:

[0034] ;

[0035] In the formula, Indicates the historical earthquake data k The loss value corresponding to the exponential weight of each data point is: Indicates that the index weight is i The fusion result after the iteration update is: represents the simulation speed data corresponding to the first speed data, n Indicates the maximum number of iterations;

[0036] The corresponding index weights are recalculated using the loss values ​​of each data in the historical earthquake data, specifically:

[0037] ;

[0038] In the formula, represents the recalculated exponential weight, represents the exponential weight before updating, represents the learning rate of the interpretation model, Indicates the loss value corresponding to the exponential weight in the current iteration number.

[0039] In a second aspect, the present application provides a geological radar intelligent data interpretation system for fusing multi-source data, which is applied to any geological radar intelligent data interpretation method for fusing multi-source data in the first aspect, including:

[0040] The first module is used to build an initial model based on deep learning. The input of the initial model is the a-scan data of the geological radar, and the output of the initial model is the propagation speed of electromagnetic waves;

[0041] The second module is used to train the initial model using preset training samples until the initial model reaches a preset training end condition, and determine the initial model that reaches the training end condition as the interpretation model;

[0042] The third module is used to obtain the detection data of the geological radar in the area to be measured, and input the detection data into the interpretation model for processing to obtain the first velocity data of the electromagnetic wave in the underground medium in the area to be measured. The detection data includes a-scan data;

[0043] The fourth module is used to calculate the correlation coefficient between each data in the first velocity data and the historical seismic data by using the first velocity data and the historical seismic data of the area to be measured;

[0044] The fifth module is used to perform multi-source fusion processing on the first velocity data and each data in the historical seismic data based on each correlation coefficient, and determine the multi-source fusion result as the target propagation velocity.

[0045] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any one of the methods in the first aspect when executing the computer program.

[0046] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions enable a computer to execute any one of the methods in the first aspect.

[0047] Compared with the prior art, the present invention has at least the following beneficial effects:

[0048] In this application, firstly, the initial model based on deep learning is trained using the preset training samples. After the training is completed, the initial model is determined as the interpretation model. The interpretation model is mainly used to interpret the propagation speed of electromagnetic waves from the a-scan data of the radar data. The propagation speed of electromagnetic waves is the propagation speed of electromagnetic waves in underground structures composed of different media. The propagation speed can be used to analyze the water content and other information in the underground structure, which can better meet the needs of radar in the field of water content detection; secondly, the detection data of the geological radar in the test area is obtained, and this detection data includes the a-scan data , then the a-scan data in the detection data is input into the interpretation model for processing, so as to obtain the first velocity data predicted by the interpretation model, that is, the propagation velocity of electromagnetic waves in the area to be measured; finally, the various data in the seismic data are used for multi-source fusion processing with the first velocity data, and finally the accurate target propagation velocity of electromagnetic waves in the area to be measured is obtained; the seismic data mainly uses the propagation velocity data of longitudinal waves, shear waves and surface waves, that is, the same type of data; through the multi-source fusion processing of multiple data, the final electromagnetic wave propagation velocity can be made more accurate, providing a reference for on-site construction, etc.

[0049] In the present application, simulation data can be obtained through simulation experiments, and the corresponding simulation speed data can be obtained by simulating the propagation of geological radar in the geological simulation model. The geological simulation model data includes the geological structure and component composition in the geological simulation model, and this data can be used as real data to effectively train the initial model; at the same time, the static learning rate in the deep learning model is often not flexible enough and can easily cause the model to fall into a local minimum; in order to obtain better model performance, the learning rate can be dynamically adjusted during training to avoid problems such as premature convergence of the model or oscillation caused by excessive learning rate; the learning rate can be reduced by a cosine function, which can also avoid parameters such as weight parameters from falling into a local minimum while reducing the time loss of model training. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0051] Figure 1 A flowchart of a method for interpreting a method in an embodiment of the present invention;

[0052] Figure 2 A connection diagram of an interpretation system in an embodiment of the present invention;

[0053] Figure 3 Schematic diagram of the connection of electronic equipment in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0055] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0057] In the description of the embodiments of the present invention, "plurality" means at least 2.

[0058] Embodiment 1: In order to solve the problem that the current interpretation of geological radar mainly relies on the personal experience of data interpreters, the data interpretation results are usually highly subjective, which may lead to erroneous interpretation results, and when interpreting geological radar data, there is often no data reference and only this single data is interpreted, which also has the problem of low interpretation accuracy; this embodiment provides a geological radar intelligent data interpretation method that integrates multi-source data, such as Figure 1 As shown, the specific steps include:

[0059] S1, build an initial model based on deep learning. The input of the initial model is the a-scan data of the geological radar, and the output of the initial model is the propagation speed of electromagnetic waves.

[0060] Among them, the initial model based on deep learning can be a U-net model, in which the multi-scale features in the geological radar a-scan data can be fully mined by combining with spatial pyramid pooling, so as to achieve the purpose of accurately predicting the propagation speed of electromagnetic waves from the a-scan data; specifically, the a-scan data of the geological radar is a single-channel waveform diagram.

[0061] S2, using preset training samples to train the initial model until the initial model reaches a preset training end condition, and determining the initial model that reaches the training end condition as the interpretation model.

[0062] Optionally, the above interpretation model is obtained by:

[0063] S21, obtaining simulation data of geological radar, the simulation data including a-scan data, geological simulation model data and simulation speed data, adding preset noise to the a-scan data, and determining the simulation data with preset noise added as training samples.

[0064] Among them, due to the complexity of the actual detection environment, the data collected by the actual geological radar has a lot of clutter interference compared with the radar simulation data. Therefore, 0.35-0.85 Gaussian noise can be randomly added to the constructed training samples.

[0065] Specifically, simulation data can be obtained through simulation experiments, and the corresponding simulation velocity data can be obtained by simulating the propagation of geological radar in the geological simulation model. The geological simulation model data includes the geological structure and component composition in the geological simulation model. This data can be used as real data to effectively train the initial model.

[0066] S22, input the a-scan data in the training sample into the initial model for processing to obtain the predicted speed, and calculate the loss function of the initial model based on the predicted speed corresponding to the a-scan data and the corresponding simulation speed data. The loss function can be an MSE function.

[0067] Optionally, the weight parameters of the initial model are iteratively updated in the following way:

[0068] ;

[0069] In the formula, Indicates t The weight parameters of the initial model at this iteration, Indicates t -1 is the weight parameter of the initial model at this iteration, Indicates t The learning rate of the initial model at this iteration, Indicates t -1 The value of the loss function of the initial model at this iteration.

[0070] Specifically, in deep learning models, static learning rates are often not flexible enough and can easily cause the model to fall into a local minimum. Therefore, in order to obtain better model performance, the learning rate can be adjusted dynamically during training to avoid problems such as premature model convergence or oscillation caused by excessive learning rates; the learning rate can be reduced through a cosine function, which can also prevent parameters such as weight parameters from falling into a local minimum and reduce the time loss of model training.

[0071] S23, if the loss function does not exceed the threshold or the number of training times reaches the maximum number of iterations, the initial model is determined as the interpretation model, and the preset training end condition is that the loss function does not exceed the threshold or the number of training times reaches the maximum number of iterations.

[0072] S3, acquiring detection data of the geological radar in the area to be measured, and inputting the detection data into the interpretation model for processing to obtain the first velocity data of the electromagnetic wave in the underground medium in the area to be measured, the detection data including a-scan data.

[0073] Among them, when processing the detection data of the geological radar in the measured area, that is, when processing the measured data of the measured area, it can first be processed by removing direct waves, removing background noise, bandpass filtering, etc., and then the a-scan data in the detection data is input into the interpretation model. The output of the interpretation model is the propagation speed of electromagnetic waves in various underground media, that is, the first speed data.

[0074] S4, using the first velocity data and the historical seismic data of the area to be measured, calculate the correlation coefficient between the first velocity data and each data in the historical seismic data.

[0075] Among them, the seismic data mainly uses the data of the propagation speed of longitudinal waves, transverse waves and surface waves, that is, the same type of velocity data; seismic data can be obtained through various detection equipment when an earthquake occurs.

[0076] Optionally, the above correlation coefficient is specifically:

[0077] ;

[0078] In the formula, Indicates the historical earthquake data n The correlation coefficient between the data and the first speed data, Indicates the coordinates of the first velocity data in the preset spatial rectangular coordinate system ( x, y, z ), Indicates the historical earthquake data n The coordinates of each data in the preset spatial rectangular coordinate system ( x, y, z ) at .

[0079] Among them, the preset spatial rectangular coordinate system can take the position of the geological radar as the coordinate origin, set the x and y axes in the horizontal direction, and set the z axis in the vertical direction; the first velocity data records the velocity value of each coordinate point in the coordinate system; similarly, the historical seismic data also records the velocity value of each coordinate point in the coordinate system (the velocity values ​​of longitudinal waves, shear waves and surface waves).

[0080] S5, based on each correlation coefficient, performing multi-source fusion processing on each data in the first velocity data and the historical seismic data, and determining the multi-source fusion result as the target propagation velocity.

[0081] Optionally, the multi-source fusion result of the multi-source fusion processing is specifically:

[0082] ;

[0083] In the formula, represents the multi-source fusion result of multi-source fusion processing, Indicates the first speed data, Indicates the first data in the historical earthquake data. Representation data The correlation coefficient with the first velocity data, Representation data The exponential weight of n Represents the total number of data involved in multi-source fusion in historical seismic data.

[0084] Among them, it can be seen from the above steps that in this embodiment, three types of longitudinal wave propagation velocity, shear wave propagation velocity and surface wave propagation velocity in historical seismic data are used. n is 3, that is, the above multi-source fusion result is expressed as: ;Right now They represent the longitudinal wave propagation velocity, shear wave propagation velocity and surface wave propagation velocity respectively.

[0085] Optionally, the exponential weight of each data in the above historical earthquake data is obtained by:

[0086] S51, based on the preset initial value of the exponential weight of each data, preliminarily fuse the first velocity data with each data in the historical seismic data to obtain a first result.

[0087] S52, using the first result to calculate the loss values ​​corresponding to the exponential weights of each data in the historical earthquake data, and using the loss values ​​of each data in the historical earthquake data to recalculate the corresponding exponential weights.

[0088] Optionally, the loss values ​​corresponding to the exponential weights of the various data in the above historical earthquake data are specifically:

[0089] ;

[0090] In the formula, Indicates the historical earthquake data k The loss value corresponding to the exponential weight of each data point is: Indicates that the index weight is iThe fusion result after the iteration update is: represents the simulation speed data corresponding to the first speed data, n Indicates the maximum number of iterations.

[0091] S53, using the recalculated exponential weights of each data to fuse with the first velocity data and iterate until the loss value corresponding to each exponential weight reaches the iteration end condition, using the exponential weights that reach the iteration end condition to perform multi-source fusion of each data in the historical seismic data and the first velocity data and obtain a multi-source fusion result.

[0092] Among them, the above-mentioned iterative process is: 1. presetting the initial value of the exponential weight; 2. fusing with the first speed data to obtain the first result; 3. using the first result to calculate the loss value of each exponential weight. When the loss value does not reach the iteration end condition, repeat 2-3 until the iteration end condition is reached; specifically, the iteration end condition can be reaching the maximum number of iterations.

[0093] Optionally, the corresponding index weights are recalculated using the loss values ​​of each data in the historical earthquake data, specifically:

[0094] ;

[0095] In the formula, represents the recalculated exponential weight, represents the exponential weight before updating, represents the learning rate of the interpretation model, Indicates the loss value corresponding to the exponential weight in the current iteration number.

[0096] Specifically, in this embodiment, first, the initial model based on deep learning is trained using preset training samples. After the training is completed, the initial model is determined as an interpretation model. The interpretation model is mainly used to interpret the propagation speed of electromagnetic waves from the a-scan data of radar data. The propagation speed of electromagnetic waves is the propagation speed of electromagnetic waves in underground structures composed of different media. The propagation speed can be used to analyze information such as water content in underground structures, which can better meet the needs of radar in the field of water content detection. Secondly, the detection data of the geological radar in the test area is obtained, and this detection data includes a-scan Data is obtained, and then the a-scan data in the detection data is input into the interpretation model for processing, so as to obtain the first velocity data predicted by the interpretation model, that is, the propagation velocity of electromagnetic waves in the area to be measured; finally, the various data in the seismic data are used with the first velocity data for multi-source fusion processing, and finally the accurate target propagation velocity of electromagnetic waves in the area to be measured is obtained; the seismic data mainly use the propagation velocity data of longitudinal waves, shear waves and surface waves, that is, the same type of data; the multi-source fusion processing of multiple data can make the final electromagnetic wave propagation velocity more accurate, which can provide a reference for on-site construction, etc.

[0097] Embodiment 2: The embodiment of the present application provides a geological radar intelligent data interpretation system integrating multi-source data, which is applied to any geological radar intelligent data interpretation method integrating multi-source data in the first aspect, such as Figure 2 As shown, including:

[0098] The first module is used to build an initial model based on deep learning. The input of the initial model is the a-scan data of the geological radar, and the output of the initial model is the propagation speed of electromagnetic waves.

[0099] The second module is used to train the initial model using preset training samples until the initial model reaches a preset training end condition, and the initial model that reaches the training end condition is determined as the interpretation model.

[0100] The third module is used to obtain the detection data of the geological radar in the area to be tested, and input the detection data into the interpretation model for processing to obtain the first velocity data of the electromagnetic wave in the underground medium in the area to be tested. The detection data includes a-scan data.

[0101] The fourth module is used to calculate the correlation coefficient between each data in the first velocity data and the historical seismic data by using the first velocity data and the historical seismic data of the area to be measured.

[0102] The fifth module is used to perform multi-source fusion processing on the first velocity data and each data in the historical seismic data based on each correlation coefficient, and determine the multi-source fusion result as the target propagation velocity.

[0103] Embodiment 3: This embodiment of the present application provides an electronic device, such as Figure 3 As shown, it includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any method in Embodiment 1 is implemented.

[0104] Embodiment 4: The embodiment of the present application provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions enable a computer to execute any method in Embodiment 1.

[0105] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for interpreting geological radar intelligent data by integrating multi-source data, characterized in that: The specific steps include: Constructing an initial model based on deep learning, wherein the input of the initial model is a-scan data of geological radar, and the output of the initial model is the propagation speed of electromagnetic waves; The initial model is trained using preset training samples until the initial model reaches a preset training end condition, and the initial model that reaches the training end condition is determined as the interpretation model; Acquire detection data of geological radar in the area to be measured, and input the detection data into the interpretation model for processing to obtain first velocity data of electromagnetic waves in the underground medium in the area to be measured, wherein the detection data includes a-scan data; Using the first velocity data and historical seismic data of the area to be measured, calculate the correlation coefficient between the first velocity data and each data in the historical seismic data; Based on each of the correlation coefficients, the first velocity data and each of the historical seismic data are subjected to multi-source fusion processing, and the multi-source fusion result is determined as the target propagation velocity; the interpretation model is obtained by: Acquire simulation data of geological radar, the simulation data including a-scan data, geological simulation model data and simulation speed data, add preset noise to the a-scan data, and determine the simulation data with preset noise added as training samples; Inputting the a-scan data in the training sample into the initial model for processing to obtain a predicted speed, and calculating the loss function of the initial model based on the predicted speed corresponding to the a-scan data and the corresponding simulation speed data; If the loss function does not exceed a threshold or the number of training times reaches a maximum number of iterations, the initial model is determined as the interpretation model, and the preset training end condition is that the loss function does not exceed a threshold or the number of training times reaches a maximum number of iterations; The correlation coefficient is specifically: Where θ represents the correlation coefficient between the nth data in the historical seismic data and the first velocity data, N(x, y, z) represents the value of the first velocity data at the coordinate (x, y, z) in the preset spatial rectangular coordinate system, and m n (x, y, z) represents the value of the nth data in the historical earthquake data at the coordinate (x, y, z) in the preset spatial rectangular coordinate system; The loss values ​​corresponding to the exponential weights of the various data in the historical earthquake data are specifically: In the formula, μ k represents the loss value corresponding to the exponential weight of the kth data in the historical seismic data, m(i) represents the fusion result after the exponential weight is updated for the ith time, and y i represents the simulation speed data corresponding to the first speed data, and n represents the maximum number of iterations; The corresponding index weight is recalculated by using the loss value of each data in the historical earthquake data, specifically: Where ω represents the recalculated exponential weight, ω a Represents the exponential weight before updating, α represents the learning rate of the interpretation model, and μ represents the loss value corresponding to the exponential weight in the current number of iterations.

2. The method for interpreting geological radar intelligent data by fusing multi-source data according to claim 1 is characterized in that: The method further includes: during the initial model training process, the weight parameters of the initial model are iteratively updated in the following manner: Where η t represents the weight parameter of the initial model at the tth iteration, η t-1 represents the weight parameter of the initial model at the t-1th iteration, α t represents the learning rate of the initial model at the tth iteration, LOSS t-1 Represents the value of the loss function of the initial model at the t-1th iteration.

3. The method for interpreting geological radar intelligent data by fusing multi-source data according to claim 1 is characterized in that: The multi-source fusion result of the multi-source fusion processing is specifically: Wherein, M represents the multi-source fusion result of multi-source fusion processing, N represents the first velocity data, m1 represents the first data in the historical seismic data, θ1 represents the correlation coefficient between data m1 and the first velocity data, ω1 represents the exponential weight of data m1, and n represents the total number of data participating in multi-source fusion in the historical seismic data.

4. The method for interpreting geological radar intelligent data by fusing multi-source data according to claim 3 is characterized in that: The exponential weights of each data in the historical earthquake data are obtained in the following way: Based on the preset initial value of the exponential weight of each data, the first velocity data is preliminarily integrated with each data in the historical seismic data to obtain a first result; Using the first result, calculate the loss values ​​corresponding to the exponential weights of each data in the historical seismic data, and recalculate the corresponding exponential weights using the loss values ​​of each data in the historical seismic data; The recalculated exponential weights of each data are fused with the first velocity data and iterated until the loss values ​​corresponding to each exponential weight reach the iteration end condition, and the exponential weights that reach the iteration end condition are used to perform multi-source fusion of each data in the historical seismic data and the first velocity data to obtain a multi-source fusion result.

5. A geological radar intelligent data interpretation system integrating multi-source data, applied to a geological radar intelligent data interpretation method integrating multi-source data according to any one of claims 1 to 4, characterized in that: include: The first module is used to construct an initial model based on deep learning, the input of the initial model is the a-scan data of the geological radar, and the output of the initial model is the propagation speed of electromagnetic waves; The second module is used to train the initial model using preset training samples until the initial model reaches a preset training end condition, and determine the initial model that reaches the training end condition as the interpretation model; The third module is used to obtain the detection data of the geological radar in the area to be measured, and input the detection data into the interpretation model for processing to obtain the first velocity data of the electromagnetic wave in the underground medium in the area to be measured, wherein the detection data includes a-scan data; A fourth module is used to calculate the correlation coefficient between each data in the first velocity data and the historical seismic data by using the first velocity data and the historical seismic data of the area to be measured; The fifth module is used to perform multi-source fusion processing on each data in the first velocity data and the historical seismic data based on each of the correlation coefficients, and determine the multi-source fusion result as the target propagation velocity.

6. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the method according to any one of claims 1 to 4 is implemented when the processor executes the computer program.

7. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions, which enable a computer to execute the method of any one of claims 1-4.

Images