Automatic extraction method and system for surface wave frequency dispersion curve, and storage medium

Abstract

The invention relates to the technical field of seismic detection, in particular to a surface wave frequency dispersion curve automatic extraction method and system and a storage medium. According to the surface wave frequency dispersion curve automatic extraction method provided by the invention, data point partitioning is performed on a surface wave frequency dispersion energy diagram based on data point extraction and clustering, then a central point in each region is extracted as a pick-up point through a mean value taking mode, and a frequency dispersion curve is constructed based on the pick-up point. The method breaks the current cognition of screening the frequency scatter points based on the principle of maximum energy, comprehensively considers the speed, energy and frequency to cluster the frequency scatter points, achieves the recognition of an energy concentration region mode, then extracts the frequency scatter points based on the clustering result, can accurately capture local features, avoids manual operation errors, and greatly improves the extraction precision of the frequency scatter points. The consistency and stability of extraction rules of frequency dispersion points in different areas are guaranteed, the height generality of a frequency dispersion curve to a surface wave frequency dispersion energy diagram is guaranteed, and high-precision and high-efficiency extraction of the frequency dispersion curve is guaranteed.

Description

Technical Field

[0001] The invention relates to the technical field of earthquake detection, and in particular to a method, system and storage medium for automatically extracting a surface wave dispersion curve. Background Art

[0002] Dispersion curves are an important means of seismic wave analysis. In geophysical seismic exploration, when using surface wave near-surface inversion to establish a velocity model, it is necessary to establish a dispersion spectrum, pick up discrete dispersion curves, and then invert the near-surface velocity model based on the discrete dispersion curves.

[0003] At present, the picking of dispersion curves is divided into manual picking and automatic picking; manual picking relies on manual experience to select the maximum energy points to form dispersion curves, which is inefficient and has a long cycle. There are two main directions for automatic picking:

[0004] 1) Select the maximum energy point and connect it to form a dispersion curve; this method only considers the energy size and has a single feature; and there may be multiple maximum energy points at the same frequency, causing interference;

[0005] 2) Use the prediction algorithm to predict the energy point on the next frequency as the prediction point, then find the dispersion point near the prediction point and select the picking point from the dispersion points near the prediction point according to the set optimization goal; this method takes into account the correlation of the picking points in the frequency dimension, but when the seismic wave changes suddenly, the error is extremely large.

[0006] The above two methods have great analytical errors in complex geological environments. Summary of the invention

[0007] In order to overcome the defect of the above-mentioned prior art that there is a lack of surface wave dispersion curve extraction method suitable for complex geology, the present invention proposes an automatic extraction method for surface wave dispersion curve, which comprehensively considers the influence of frequency, speed and energy to pick up the dispersion curve, has high accuracy, and is suitable for any geology and mode.

[0008] The present invention proposes a method for automatically extracting a surface wave dispersion curve, which converts a surface wave dispersion energy map into a set of data points; divides the data points into regions, calculates the mean point of each region, and the mean point is the average value of the data points in the region; and constructs a dispersion curve by combining the mean points of all regions.

[0009] The conditions for segmenting data points into regions are as follows: each data point in each region has at least num neighboring points in the set of data points, and each data point in each region has at least one neighboring point in the same region; if the distance between two data points is less than or equal to L, the two data points are defined as neighbors of each other; num is the minimum number of data points in the cluster, and L is the maximum distance of the neighborhood.

[0010] Preferably, a core point is defined as a data point whose number of neighborhood points is greater than or equal to num; and the method for performing region segmentation on the data point comprises the following steps:

[0011] Get the starting point as the core point and construct the corresponding data class;

[0012] Add the starting point to the data class, and add the neighborhood points of the starting point to the set candidate point set. The initial state of the candidate point set is an empty set.

[0013] Traverse the data points in the candidate point set. When it is a core point, move the data point to the data class and add the neighborhood points of the data point to the candidate point set. When it is not a core point, delete it from the candidate point set until the candidate point set is an empty set.

[0014] The above steps are repeated until all core points are segmented into data classes.

[0015] Preferably, the data points include velocity, frequency and energy.

[0016] Preferably, the surface roll dispersion energy map is converted into a set of data points by first extracting the original data points of the surface roll dispersion energy map, normalizing each data dimension of all the original data points, and converting the original data points into standardized data points.

[0017] Preferably, the distance between two data points is represented by a weighted Euclidean distance.

[0018] Preferably, the frequency weight, speed weight and energy weight decrease in sequence.

[0019] Preferably, the frequency weight is 1, the speed weight is 0.5, and the energy weight is 0.2.

[0020] Preferably, when constructing the dispersion curve, firstly, the mean points of all regions are collected to construct the discrete shape of the dispersion curve, and then smoothing and denoising are performed to obtain the final dispersion curve.

[0021] The present invention provides a system for automatically extracting surface wave dispersion curves, comprising a memory and a processor. The memory stores a computer program. The processor is connected to the memory and is used to execute the computer program to implement the method for automatically extracting surface wave dispersion curves.

[0022] The present invention provides a storage medium storing a computer program, which is used to implement the automatic extraction method of surface wave dispersion curve when executed.

[0023] The advantages of the present invention are:

[0024] (1) The present invention proposes an automatic extraction method for surface wave dispersion curves, which partitions the surface wave dispersion energy map based on data point extraction and clustering, then extracts the center point of each area as a pick-up point by taking the average, and constructs a dispersion curve based on the pick-up point. The present invention breaks the current cognition of screening dispersion points based on the maximum energy principle, clusters the dispersion points by comprehensively considering speed, energy and frequency, realizes the recognition of energy concentration area patterns, and then extracts dispersion points based on the clustering results, which can accurately capture local features, avoid human operation errors, and greatly improve the extraction accuracy of dispersion points; ensures the consistency and stability of the extraction rules of dispersion points in different regions, ensures the high generalization of the dispersion curve to the surface wave dispersion energy map, and ensures the high-precision and high-efficiency extraction of the dispersion curve.

[0025] (2) The present invention combines a clustering algorithm to perform data point region segmentation, thereby solving the problem of traditional methods relying on manually set thresholds and improving the automation and accuracy of dispersion curve extraction. In addition, noise data points can be automatically identified and excluded during the calculation process, thereby improving the robustness of dispersion curve extraction and reducing human intervention.

[0026] (3) The present invention proposes a weighted Euclidean distance method, which assigns different weights to each feature, thereby improving the accuracy of cluster analysis and making the contribution of different features reasonably reflected. The present invention combines standardized preprocessing and weighting mechanism, so that data can be clustered at the same scale, and the influence of each feature on the clustering result can be adjusted according to the actual importance.

[0027] (4) The present invention introduces post-processing methods, such as interpolation and smoothing, to further optimize the dispersion curve, ensuring that the final output dispersion curve is more realistic and consistent with the actual geological background. The method of the present invention has strong scalability and can be widely used in different types of seismic data analysis tasks to meet different exploration needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a flow chart of a method for automatically extracting surface wave dispersion curve;

[0029] Figure 2 is the original surface wave signal in the embodiment;

[0030] Figure 3 is the surface wave dispersion energy diagram;

[0031] Figure 4 It is a dispersion curve picked manually;

[0032] Figure 5 This is the dispersion curve obtained by the method of the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. 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.

[0034] The present invention proposes a method for automatically extracting a surface wave dispersion curve, comprising the following steps:

[0035] S1, obtain the surface wave dispersion energy map and extract the characteristic matrix of each data point, the characteristic matrix includes frequency, velocity and energy;

[0036] S2, standardize the data type in the feature representation to obtain a standardized matrix;

[0037] Let the characteristic matrix of the surface wave dispersion energy map be denoted as Data:

[0038]

[0039] Where (f n ,v n ,E n ) represents the frequency, velocity and energy of the nth data point on the surface wave dispersion energy diagram, and N represents the number of data points. It can be seen that (f1,v1,E1) represents the frequency, velocity and energy of the first data point on the surface wave dispersion energy diagram, (f2,v2,E2) represents the frequency, velocity and energy of the second data point on the surface wave dispersion energy diagram, and (f N ,v N ,E N ) represents the frequency, velocity and energy of the Nth data point on the surface wave dispersion energy diagram.

[0040] In specific implementation, the data can be normalized first, then (f n ,v n ,E n ) represents the frequency-normalized value, velocity-normalized value, and energy-normalized value of the nth data point on the surface wave dispersion energy diagram.

[0041] The data normalization process is as follows:

[0042] Let the original data be recorded as:

[0043]

[0044] Among them, (f 0,1 ,v 0,1 ,E 0,1) represents the original frequency, original velocity and original energy of the first data point on the surface wave dispersion energy diagram, (f 0,2 ,v 0,2 ,E 0,2 ) represents the original frequency, original velocity and original energy of the second data point on the surface wave dispersion energy diagram, (f 0,n ,v 0,n ,E 0,n ) represents the original frequency, original velocity and original energy of the nth data point on the surface wave dispersion energy diagram, N represents the number of data points, (f 0,N ,v 0,N ,E 0,N ) represents the original frequency, original velocity and original energy of the Nth data point on the surface wave dispersion energy diagram.

[0045] Let μ f and σ f Respectively represent the frequency sequence {f 0,1 ;f 0,2 ;…;f 0,n ;…;f 0,N} The mean and standard deviation of v and σ v Respectively represent the frequency sequence {v 0,1 ;v 0,2 ;…;v 0,n ;…;v 0,N} The mean and standard deviation of E and σ E Respectively represent the frequency sequence {E 0,1 ; E 0,2 ;…;E 0,n ;…;E 0,N} The average value and standard deviation of the original data Data0 are recorded as Data' after normalization;

[0046]

[0047] Among them, (f 1,1 ,v 1,1 ,E 1,1 ) represents the original frequency, original velocity and original energy of the first data point on the surface wave dispersion energy diagram, (f 1,2 ,v 1,2 ,E 1,2 ) represents the original frequency, original velocity and original energy of the second data point on the surface wave dispersion energy diagram, (f 1,n ,v 1,n ,E 1,n ) represents the original frequency, original velocity and original energy of the nth data point on the surface wave dispersion energy diagram, N represents the number of data points, (f 1,N ,v1,N ,E 1,N ) represents the original frequency, original velocity and original energy of the Nth data point on the surface wave dispersion energy diagram.

[0048] When Data uses the original data of the surface roll dispersion energy map, Data=Data0; when Data uses the normalized processing result of the original data of the surface roll dispersion energy map, Data=Data'.

[0049] S3. Define the maximum neighborhood distance L and the minimum number of cluster data points num, and classify the data points. The classification method is as follows:

[0050] S31, extracting an unclassified data point as a starting point;

[0051] S32, determine whether the number of data points in the neighborhood of the starting point is greater than the minimum number of data points in the cluster; if not, return to step S31 to re-extract the starting point; if yes, update the parameter m to m+1, the initial value of m is 0, and then execute step S33;

[0052] S33, constructing a data class Y(m+1), adding the starting point to the data class Y(m+1), and adding the neighborhood points of the starting point to the set candidate point set, where the initial state of the candidate point set is an empty set;

[0053] S34. Extract data point Q from the candidate point set without replacement. If data point Q is a core point, add data point Q to data class Y(m+1), and add the neighborhood points of data point Q to the candidate point set; loop step S34 until the candidate point set is an empty set.

[0054] For the convenience of subsequent description, an invalid point set is set here. If the data point Q in step S34 is not a core point, it will be added to the invalid point set.

[0055] When the number of neighborhood points of a data point is greater than or equal to the minimum number of cluster data points num, the data point is a core point; the neighborhood points of a data point are other data points in the area with the data point as the center and the maximum neighborhood distance L as the radius; that is, the distance between the data point and any of its domain points is less than or equal to L;

[0056] In the present invention, the distance between data points is represented by weighted Euclidean distance. Taking data point x1(f1, v1, E1) and data point x2(f2, v2, E2) as an example, the distance calculation formula between the two is as follows:

[0057]

[0058] Among them, w f represents the frequency weight, w v represents the speed weight, wE represents the energy weight;

[0059] Referring to D(x1,x2), if D(x1,x2)≤L, then data point x1 and data point x2 are neighboring points of each other.

[0060] S35. Determine whether all data points have been traversed; if not, extract an uncalculated data point as the starting point, and then return to step S32, that is, the new starting point does not belong to any existing data class and does not belong to the invalid point set; if yes, complete the data classification and count all data classes Y(1), Y(2), ...Y(m+1).

[0061] S4. Calculate the mean point of each data class. The set of all mean points constitutes the discrete shape of the dispersion curve corresponding to the Data.

[0062] The mean point of a data class is the mean of the data points within the data class; assuming that the data class Y(j) contains data points (f j,1 ,v j,1 ,E j,1 )、(f j,2 ,v j,2 ,E j,2 ),……、(f j,k ,v j,k ,E j,k ), then the calculation formula for the mean point Y(j,ave) of the data class Y(j) is: Y(j,ave)=(f j,ave ,v j,ave ,E j,ave );

[0063]

[0064] S5. Smoothing the dispersion curve obtained in step S4, i.e., the point set {Y(j,ave); 1≤j≤m+1}, and then removing noise to obtain a continuous dispersion curve corresponding to Data.

[0065] The smoothing process may specifically adopt an interpolation algorithm, such as spline interpolation.

[0066] The above-mentioned method for automatically extracting the surface wave dispersion curve is verified in combination with specific embodiments below.

[0067] In this embodiment, the original data points of the surface wave dispersion energy map are first normalized, and then the data points are segmented into regions to calculate the mean value points of each region.

[0068] In this embodiment, the distance between two data points is represented by the weighted Euclidean distance. During the calculation process, the frequency weight w f The value is 1, the speed weight is w vThe value is 0.5, and the energy weight w E The value is 0.2

[0069] In this embodiment, the original surface wave signal is as follows: Figure 2 As shown in Figure 2, the surface wave dispersion energy diagram is shown in Figure 2. Figure 3 shown.

[0070] In this embodiment, the traditional method of manually picking up the dispersion curve is used to obtain Figure 4 The black line in the middle shows the dispersion curve, where the black dots represent the manually clicked dispersion energy points, and the dispersion curve is formed by connecting the dispersion energy points. When picking manually, the dispersion energy diagram is manually observed, and the mouse is clicked on the part with the largest energy in the dispersion energy diagram, that is, the dark part, to form a dispersion curve.

[0071] Manual point-by-point selection is time-consuming, especially when the number is large. Manual picking is easily affected by the subjective experience of the operator, and different operators may have different picking results for the same dispersion energy diagram. Manual picking may be difficult to distinguish for complex signals with weak signal energy and multiple modes.

[0072] Figure 5 The black line shown is a dispersion curve obtained by using the automatic extraction method of surface wave dispersion curve proposed by the present invention. Figure 4 , Figure 5 By comparison, it can be seen that the method of the present invention can quickly extract more data points, and the obtained dispersion curve is consistent with the trend of the manually picked dispersion curve, especially the manually picked points almost completely fall on the dispersion curve obtained by the method of the present invention, which proves the reliability of the method of the present invention. In addition, the dispersion curve obtained by the present invention is smoother and more consistent with the energy diagram trend at the non-manually picked point position.

[0073] Of course, it is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, but also includes the same or similar structures that can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-limiting, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any reference numerals in the claims should not be regarded as limiting the claims involved.

[0074] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

[0075] The techniques, shapes, and structural parts not described in detail in the present invention are all well-known techniques.

Claims

1. A method for automatically extracting surface wave dispersion curves, characterized in that: Convert the surface wave dispersion energy map into a set of data points; divide the data points into regions, calculate the mean point of each region, and the mean point is the average value of the data points in the region; and construct a dispersion curve by combining the mean points of all regions; The conditions for segmenting data points into regions are as follows: each data point in each region has at least num neighboring points in the set of data points, and each data point in each region has at least one neighboring point in the same region; if the distance between two data points is less than or equal to L, the two data points are defined as neighbors of each other; num is the minimum number of data points in the cluster, and L is the maximum distance of the neighborhood.

2. The method for automatically extracting surface wave dispersion curves according to claim 1, characterized in that: A core point is defined as a data point whose number of neighboring points is greater than or equal to num; the method for segmenting the data point into regions comprises the following steps: Get the starting point as the core point and construct the corresponding data class; Add the starting point to the data class, and add the neighborhood points of the starting point to the set candidate point set. The initial state of the candidate point set is an empty set. Traverse the data points in the candidate point set. When it is a core point, move the data point to the data class and add the neighborhood points of the data point to the candidate point set. When it is not a core point, delete it from the candidate point set until the candidate point set is an empty set. The above steps are repeated until all core points are segmented into data classes.

3. The method for automatically extracting surface wave dispersion curves according to claim 1, characterized in that: Data points include speed, frequency and energy.

4. The method for automatically extracting surface wave dispersion curves according to claim 3, characterized in that: The method of converting the surface wave dispersion energy map into a set of data points is as follows: firstly, the original data points of the surface wave dispersion energy map are extracted, each data dimension of all the original data points is standardized, and the original data points are converted into standardized data points.

5. The method for automatically extracting surface wave dispersion curves according to claim 3 or 4, characterized in that: The distance between two data points is expressed as weighted Euclidean distance.

6. The method for automatically extracting surface wave dispersion curves according to claim 5, characterized in that: The frequency weight, speed weight, and energy weight decrease in turn.

7. The method for automatically extracting surface wave dispersion curves according to claim 6, characterized in that: The frequency weight is 1, the speed weight is 0.5, and the energy weight is 0.

2.

8. The method for automatically extracting surface wave dispersion curves according to any one of claims 1 to 4, characterized in that: When constructing a dispersion curve, the mean points of all regions are firstly collected to construct the discrete shape of the dispersion curve, and then smoothing and denoising are performed to obtain the final dispersion curve.

9. A surface wave dispersion curve automatic extraction system, characterized in that: The invention comprises a memory and a processor, wherein a computer program is stored in the memory, and the processor is connected to the memory, and the processor is used to execute the computer program to implement the method for automatically extracting the surface wave dispersion curve according to any one of claims 1 to 7.

10. A storage medium, characterized in that: A computer program is stored, and when the computer program is executed, it is used to implement the method for automatically extracting the surface wave dispersion curve according to any one of claims 1 to 7.

Images