Unsupervised velocity spectrum automatic picking method based on k-means clustering, electronic device, storage medium and apparatus

Abstract

This invention provides an unsupervised automatic velocity spectrum picking method, electronic device, storage medium, and apparatus based on K-means clustering, belonging to the field of seismic data processing technology. The automatic picking method includes: manually picking multiple time-velocity pairs and generating prediction lines, as well as generating first and second constraint lines; cleaning the data and performing K-means clustering on the cleaned data to obtain multiple near-surface cluster centers, mid-level cluster centers, and deep cluster centers; performing least-squares linear fitting on the multiple near-surface cluster centers to obtain multiple near-surface time-velocity pairs; calculating the percentage of the average velocity of the multiple deep cluster centers to the velocity value of the corresponding point on the prediction line, and correcting the deep cluster centers based on this percentage to obtain multiple deep time-velocity pairs. This method can effectively reduce the density of velocity control points and manual picking time, and has picking accuracy comparable to manual picking, thereby reducing data processing costs.

Description

Technical Field

[0001] This invention belongs to the field of seismic data processing technology, and more specifically, relates to an unsupervised automatic velocity spectrum picking method, electronic device, storage medium and apparatus based on K-means clustering. Background Technology

[0002] Velocity analysis is one of the most important steps in seismic data processing. Accurate velocity is crucial for subsequent processing such as multiple wave suppression, migration, time-depth conversion, and inversion. Velocity is estimated using velocity spectra and then used for normal time difference (NMO) correction. In traditional methods, velocity analysis involves the following steps: (1) Dynamic correction of stacked velocity estimation is achieved by summing similarity along a hyperbolic time trajectory and generating a velocity spectrum; (2) Several stacking time points are manually selected from the velocity spectrum; (3) Layer velocities are calculated based on the selected stacking velocities to construct a velocity-time model. Velocity analysis is a time-consuming task, requiring visual inspection of a large number of velocity spectra by processing personnel. This is very time-consuming and highly dependent on the experience and subjectivity of seismic processing personnel. Nevertheless, since the velocity spectrum contains effective information for determining the underground velocity at the current location, the process of estimating background velocities based on velocity spectra is still widely used in production. Summary of the Invention

[0003] The purpose of this invention is to provide an unsupervised automatic velocity spectrum acquisition method, electronic device, storage medium, and apparatus based on K-means clustering, which solves the problem that existing technologies require personnel to visually inspect the velocity spectrum.

[0004] To achieve the above objectives, in a first aspect, the present invention provides an unsupervised automatic velocity spectrum picking method based on K-means clustering, comprising the following steps:

[0005] Multiple time-velocity pairs are picked up across the entire area, and prediction lines are generated using the multiple time-velocity pairs as control points.

[0006] A first constraint line and a second constraint line are generated based on the predicted line. The velocity trend of the first constraint line is lower than the velocity trend of the predicted line by a first set percentage, and the velocity trend of the second constraint line is higher than the velocity trend of the predicted line by a second set percentage.

[0007] Set similar velocity data that are lower than the first constraint line velocity trend or higher than the second constraint line velocity trend to zero, and remove similar velocity data that are located between the first constraint line and the second constraint line and whose similarity to the predicted line velocity trend is less than a third set percentage, to obtain cleaned data.

[0008] The cleaned data is then subjected to K-means clustering to obtain multiple near-surface cluster centers, multiple mid-level cluster centers, and multiple deep cluster centers.

[0009] By performing least-squares linear fitting on multiple near-surface cluster centers, multiple near-surface time-velocity pairs are obtained.

[0010] Calculate the percentage of the average velocity of multiple deep cluster centers to the velocity value of the corresponding point on the prediction line, and correct the deep cluster centers based on the percentage to obtain multiple deep time-velocity pairs.

[0011] Optionally, after calculating the percentage of the average velocity of multiple deep cluster centers to the velocity value of the corresponding point on the prediction line, and correcting the deep cluster centers based on the percentage to obtain the deep time-velocity pair, the method further includes:

[0012] The full-area three-dimensional velocity field is obtained based on multiple near-surface time-velocity pairs, multiple deep time-velocity pairs, and multiple mid-layer cluster centers. The full-area three-dimensional velocity field is then used to perform dynamic correction on the CMP gather.

[0013] Optionally, the first set percentage is 10%;

[0014] The second set percentage is 30%.

[0015] Optionally, the third set percentage is 10%.

[0016] Optionally, the cleaned data is subjected to K-means clustering to obtain multiple near-surface cluster centers, multiple mid-level cluster centers, and multiple deep cluster centers, including:

[0017] If there are multiple cluster center points in a local area, the center points of the multiple clusters are merged according to the principle of nearest time.

[0018] In a second aspect, the present invention provides an electronic device, the electronic device comprising:

[0019] At least one processor; and,

[0020] A memory communicatively connected to the at least one processor; wherein,

[0021] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the unsupervised automatic velocity spectrum picking method based on K-means clustering as described in any one of claims 1-5.

[0022] Thirdly, the present invention provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the unsupervised automatic velocity spectrum picking method based on K-means clustering as described in any one of claims 1-5.

[0023] Fourthly, the present invention provides an unsupervised automatic velocity spectrum picking device based on K-means clustering, comprising:

[0024] The prediction line generation module is used to pick up multiple time-velocity pairs across the entire area and generate a prediction line using the multiple time-velocity pairs as control points.

[0025] The constraint line generation module is used to generate a first constraint line and a second constraint line based on the predicted line, wherein the velocity trend of the first constraint line is lower than the velocity trend of the predicted line by a first set percentage, and the velocity trend of the second constraint line is higher than the velocity trend of the predicted line by a second set percentage.

[0026] The data cleaning module is used to set similar velocity data that is lower than the first constraint linear velocity trend or higher than the second constraint linear velocity trend to zero, and to remove similar velocity data that is located between the first constraint line and the second constraint line and whose similarity to the predicted linear velocity trend is less than a third set percentage, so as to obtain cleaned data.

[0027] The center point calculation module is used to perform K-means clustering on the cleaned data to obtain multiple near-surface cluster center points, multiple mid-level cluster center points, and multiple deep cluster center points;

[0028] The velocity spectrum picking module is used to perform least-squares linear fitting on multiple near-surface cluster center points to obtain multiple near-surface time-velocity pairs.

[0029] Calculate the percentage of the average velocity of multiple deep cluster centers to the velocity value of the corresponding point on the prediction line, and correct the deep cluster centers based on the percentage to obtain multiple deep time-velocity pairs.

[0030] Optionally, the device further includes:

[0031] The CMP gather correction module is used to obtain a full-area three-dimensional velocity field based on multiple near-surface time-velocity pairs, multiple deep time-velocity pairs, and multiple mid-layer cluster centers, and to perform dynamic correction on the CMP gather using the full-area three-dimensional velocity field.

[0032] Optionally, the first set percentage is 10%;

[0033] The second set percentage is 30%;

[0034] The third set percentage is 10%.

[0035] The beneficial effects of this invention are as follows: It provides an unsupervised automatic velocity spectrum picking method based on K-means clustering, comprising: manually picking multiple time-velocity pairs and generating a prediction line; generating a first constraint line and a second constraint line based on the prediction line; cleaning the data and performing K-means clustering on the cleaned data to obtain multiple near-surface cluster centers, multiple mid-level cluster centers, and multiple deep cluster centers; performing least-squares linear fitting on the multiple near-surface cluster centers to obtain multiple near-surface time-velocity pairs; calculating the percentage of the average velocity of the multiple deep cluster centers to the velocity value of the corresponding point on the prediction line, and correcting the deep cluster centers based on this percentage to obtain multiple deep time-velocity pairs. This method can effectively reduce the density of velocity control points and the manual picking time, and has picking accuracy comparable to manual picking, thereby reducing the cost of data processing.

[0036] Other features and advantages of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0037] The above and other objects, features and advantages of the present invention will become more apparent from the more detailed description of exemplary embodiments of the invention in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments of the invention.

[0038] Figure 1 A flowchart of an unsupervised automatic velocity spectrum picking method based on K-means clustering according to Embodiment 1 of the present invention is shown.

[0039] Figure 2 A flowchart of an unsupervised automatic velocity spectrum picking method based on K-means clustering according to Embodiment 2 of the present invention is shown.

[0040] Figure 3a A schematic diagram of a typical similar velocity spectrum of a work area according to Embodiment 2 of the present invention is shown.

[0041] Figure 3b A schematic diagram of the clustering results of directly processing similar velocity spectra using K-means clustering according to Embodiment 2 of the present invention is shown.

[0042] Figure 4a A schematic diagram showing the location of control points within the work area according to Embodiment 2 of the present invention is provided.

[0043] Figure 4b A time-velocity pair intersection diagram of the control point being picked up according to Embodiment 2 of the present invention is shown.

[0044] Figure 5aA schematic diagram of similar velocity spectra after adding constraint lines to the CMP point according to Embodiment 2 of the present invention is shown.

[0045] Figure 5b A schematic diagram of similar velocity spectra after CMP point cleaning according to Embodiment 2 of the present invention is shown.

[0046] Figure 5c A schematic diagram of the clustering results according to Embodiment 2 of the present invention is shown.

[0047] Figure 6a A schematic diagram of the near-surface time-velocity pair pickup principle according to Embodiment 2 of the present invention is shown.

[0048] Figure 6b A schematic diagram of the deep time-velocity pair picking principle according to Embodiment 2 of the present invention is shown.

[0049] Figure 6c A schematic diagram of the time-velocity pair for the final automatic pickup according to Embodiment 2 of the present invention is shown.

[0050] Figure 7a A schematic diagram of a CMP gather according to Embodiment 2 of the present invention is shown.

[0051] Figure 7b A schematic diagram of the superimposed velocity spectrum automatically picked up after velocity similarity cleaning according to Embodiment 2 of the present invention is shown.

[0052] Figure 7c A schematic diagram of the dynamically corrected CMP gather according to Embodiment 2 of the present invention is shown. Detailed Implementation

[0053] Preferred embodiments of the invention will now be described in more detail. While preferred embodiments of the invention are described below, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0054] Example 1:

[0055] See Figure 1 This disclosure provides an unsupervised automatic velocity spectrum picking method based on K-means clustering, comprising the following steps:

[0056] S1: Pick multiple time-velocity pairs across the entire area and generate a prediction line using these multiple time-velocity pairs as control points.

[0057] S2: Generate a first constraint line and a second constraint line based on the predicted line. The velocity trend of the first constraint line is lower than the velocity trend of the predicted line by a first set percentage, and the velocity trend of the second constraint line is higher than the velocity trend of the predicted line by a second set percentage.

[0058] In this embodiment, the first set percentage is 10%;

[0059] The second percentage is set at 30%.

[0060] In practice, the first and second set percentages can be tried and adjusted based on the actual data to eliminate as much redundant data as possible.

[0061] S3: Set similar velocity data that are lower than the first constraint linear velocity trend or higher than the second constraint linear velocity trend to zero, and remove similar velocity data that are located between the first constraint line and the second constraint line and whose similarity to the predicted linear velocity trend is less than a third set percentage, to obtain the cleaned data.

[0062] In this embodiment, the third percentage is set to 10%.

[0063] S4: Perform K-means clustering on the cleaned data to obtain multiple near-surface cluster centers, multiple mid-level cluster centers, and multiple deep cluster centers;

[0064] In this step, if there are multiple cluster center points in a local area, the center points of multiple clusters are merged according to the principle of the closest time.

[0065] In this embodiment, the first four cluster centers are designated as near-surface cluster centers, the last two cluster centers are designated as deep cluster centers, and the remaining cluster centers are designated as mid-level cluster centers. In practice, the number of cluster centers can be selected based on the relationship between the prediction line and the cluster centers.

[0066] S5: Perform least-squares linear fitting on multiple near-surface cluster centers to obtain multiple near-surface time-velocity pairs.

[0067] S6: Calculate the percentage of the average velocity of multiple deep cluster centers to the velocity value of the corresponding point on the prediction line, and correct the deep cluster centers based on the percentage to obtain multiple deep time-velocity pairs.

[0068] S7: Obtain the three-dimensional velocity field of the entire region based on multiple near-surface time-velocity pairs, multiple deep time-velocity pairs, and multiple mid-level cluster centers, and use the three-dimensional velocity field of the entire region to perform dynamic correction on the CMP gather.

[0069] Specifically, this method can effectively reduce the density of speed control points and manual picking time, and has picking accuracy comparable to manual picking, thereby reducing the cost of data processing.

[0070] Example 2:

[0071] like Figure 2 As shown, this embodiment provides an unsupervised automatic velocity spectrum picking method based on K-means clustering, including the following steps:

[0072] Step 1: Manually pick some time-velocity pairs across the entire area and use them as control points for generating the prediction line;

[0073] Step 2: Generate constraint lines at different scales based on the accuracy of the predicted lines;

[0074] Step 3: Clean the data, that is, set the similar velocity data outside the constraint line on the velocity spectrum to zero, and remove the data with a similarity of less than 0.1 inside the constraint line;

[0075] Step 4: Perform K-means clustering on the cleaned data, and merge the centroids of multiple clusters in some areas according to the recentest time principle;

[0076] Step 5: Perform least squares linear fitting based on the first four cluster centers to generate near-surface time-velocity pairs. Calculate the percentage of the average of the last two cluster centers relative to the predicted line, and use this percentage to calculate the deep time-velocity pairs.

[0077] Step 6: Use this method to obtain the three-dimensional velocity field of the entire area, and perform dynamic correction on the CMP gather to verify the velocity field picking accuracy.

[0078] Furthermore, this embodiment uses this method to process experimental data from an oilfield in western China, with a full coverage area of ​​approximately 150 km². 2 . Figure 3a The typical similar velocity spectrum of the work area is shown. From the velocity spectrum, we can see that the high correlation coefficient is concentrated in the time range of 200ms to 3000ms, and the correlation coefficient value is less than 0.1 in the vicinity. Figure 3b The figure shows the clustering results of directly processing similar velocity spectra using K-means clustering. The red dots circled in green represent the center points of the clusters. From this figure, we can see that the time-velocity pairs were not automatically picked in a regular manner. In other words, the similarity of the original data must be cleaned before it is input into the K-means clustering process.

[0079] To better eliminate the similarity of velocities, some constraints need to be introduced. First, some time-velocity pairs are manually picked and used as control points for generating the prediction line. CMPs far from the control points will use the average time-velocity pairs as the prediction line. Figure 4a The locations of control points within the work area are shown. Figure 4bThe time-velocity cross plot of the picked control points is shown. There are a total of 41 control points in the entire 3D work area. Figure 5a The diagram shows the similar velocity spectrum after adding constraint lines to CMP points. The red line at the top is the velocity prediction line for CMP points; although the velocity trend is almost correct, it does not pass through the actual peak point we would normally pick manually. The green line indicates that its velocity trend is 10% lower than the predicted trend, the cyan line indicates that its velocity trend is 10% higher than the predicted trend, the white line indicates 20% higher, and the yellow line indicates 30% higher. Velocity similarity data that are lower than the green constraint line trend or higher than the yellow constraint line trend are set to zero. We also set a threshold to remove data with a velocity similarity below 0.1. Figure 5b The similar velocity spectrum after CMP point cleaning is shown, with the retained velocity similarity data used as input for K-means clustering. Figure 5c The clustering results are shown, with the cluster centroids corresponding to the picked time-velocity pairs. In some regions, multiple cluster centroids are present, and these are merged based on the nearest time principle.

[0080] exist Figure 6a In the diagram, the red dots circled in green represent the automatically picked time-velocity pairs after combination. Generally, due to the influence of first arrival waves and refraction, the signal-to-noise ratio of near-surface similarity is low. We perform linear least-squares fitting based on the first four center points (marked by white arrows) to obtain the predicted near-surface time-velocity pairs (green dots circled in red, marked by red arrows). Since the velocity similarity values ​​are below a threshold, similarities below 2700 ms do not cluster. Figure 6b In the middle, we use the average of the last two cluster centers (marked by white arrows) to create a new point (marked by red arrow) to calculate the predicted velocity percentage. The green dots on the yellow dashed line are the time-velocity pairs calculated as percentages, which can be seen as the velocity trend below 2700ms. Figure 6c The final autopic time-velocity pair is shown, and the difference is obvious compared to the predicted line.

[0081] Figure 7a A selected CMP gather is shown. Figure 7b The stacked velocity spectrum automatically picked up after velocity similarity cleaning is shown in Figure 7c, and the CMP gather after dynamic correction (NMO) is shown in Figure 7c. The results show that the automatically picked velocity spectrum can effectively flatten the CMP gather. The method in this embodiment can reduce the time for manual picking, thereby reducing the cost of data processing.

[0082] Example 3:

[0083] This disclosure also provides an electronic device, which includes:

[0084] At least one processor; and,

[0085] A memory that is communicatively connected to at least one processor; wherein,

[0086] The memory stores instructions that can be executed by at least one processor, which enables the at least one processor to perform the unsupervised automatic velocity spectrum picking method based on K-means clustering in Embodiment 1 or 2.

[0087] An electronic device according to embodiments of the present disclosure includes a memory and a processor. The memory is used to store non-transitory computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0088] The processor may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In one embodiment of this disclosure, the processor is used to execute computer-readable instructions stored in the memory.

[0089] Those skilled in the art will understand that, in order to solve the technical problem of how to achieve a good user experience, this embodiment may also include well-known structures such as communication buses and interfaces, and these well-known structures should also be included within the protection scope of this disclosure.

[0090] For a detailed description of this embodiment, please refer to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.

[0091] Example 4:

[0092] This disclosure provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the unsupervised velocity spectrum auto-picking method based on K-means clustering in Embodiment 1 or 2.

[0093] A computer-readable storage medium according to embodiments of the present disclosure stores non-transitory computer-readable instructions. When these non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the methods described in the foregoing embodiments of the present disclosure are performed.

[0094] The aforementioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or portable hard drive), media with built-in rewritable non-volatile memory (e.g., memory card), and media with built-in ROM (e.g., ROM cartridge).

[0095] Example 5:

[0096] Referring to the figure, this disclosure provides an unsupervised velocity spectrum automatic picking device based on K-means clustering, comprising:

[0097] The prediction line generation module is used to pick up multiple time-velocity pairs across the entire area and generate prediction lines using multiple time-velocity pairs as control points.

[0098] The constraint line generation module is used to generate a first constraint line and a second constraint line based on the predicted line. The velocity trend of the first constraint line is lower than the velocity trend of the predicted line by a first set percentage, and the velocity trend of the second constraint line is higher than the velocity trend of the predicted line by a second set percentage.

[0099] The data cleaning module is used to set similar velocity data that are lower than the first constraint linear velocity trend or higher than the second constraint linear velocity trend to zero, and to remove similar velocity data that are located between the first constraint line and the second constraint line and whose similarity to the predicted linear velocity trend is less than a third set percentage, so as to obtain cleaned data.

[0100] The center point calculation module is used to perform K-means clustering on the cleaned data to obtain multiple near-surface cluster centers, multiple mid-level cluster centers, and multiple deep cluster centers;

[0101] The velocity spectrum picking module is used to perform least-squares linear fitting on multiple near-surface cluster center points to obtain multiple near-surface time-velocity pairs.

[0102] Calculate the percentage of the average velocity of multiple deep cluster centers to the velocity value of the corresponding point on the prediction line, and then correct the deep cluster centers based on the percentage to obtain multiple deep time-velocity pairs.

[0103] Optionally, the automatic pickup device further includes:

[0104] The CMP gather correction module is used to obtain the full-area three-dimensional velocity field based on multiple near-surface time-velocity pairs, multiple deep time-velocity pairs, and multiple mid-layer cluster centers, and to perform dynamic correction on the CMP gather using the full-area three-dimensional velocity field.

[0105] Optionally, the first set percentage is 10%;

[0106] The second percentage is set at 30%;

[0107] The third percentage is set at 10%.

[0108] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. An unsupervised automatic velocity spectrum picking method based on K-means clustering, characterized in that, Includes the following steps: Multiple time-velocity pairs are picked up across the entire area, and prediction lines are generated using the multiple time-velocity pairs as control points. A first constraint line and a second constraint line are generated based on the predicted line. The velocity trend of the first constraint line is lower than the velocity trend of the predicted line by a first set percentage, and the velocity trend of the second constraint line is higher than the velocity trend of the predicted line by a second set percentage. Set similar velocity data that are lower than the first constraint line velocity trend or higher than the second constraint line velocity trend to zero, and remove similar velocity data that are located between the first constraint line and the second constraint line and whose similarity to the predicted line velocity trend is less than a third set percentage, to obtain cleaned data. The cleaned data is then subjected to K-means clustering to obtain multiple near-surface cluster centers, multiple mid-level cluster centers, and multiple deep cluster centers. By performing least-squares linear fitting on multiple near-surface cluster centers, multiple near-surface time-velocity pairs are obtained. Calculate the percentage of the average velocity of multiple deep cluster centers to the velocity value of the corresponding point on the prediction line, and correct the deep cluster centers based on the percentage to obtain multiple deep time-velocity pairs.

2. The unsupervised automatic velocity spectrum picking method based on K-means clustering according to claim 1, characterized in that, After calculating the percentage of the average velocity of multiple deep cluster centers to the velocity value of the corresponding point on the prediction line, and correcting the deep cluster centers based on the percentage to obtain deep time-velocity pairs, the process further includes: The full-area three-dimensional velocity field is obtained based on multiple near-surface time-velocity pairs, multiple deep time-velocity pairs, and multiple mid-layer cluster centers. The full-area three-dimensional velocity field is then used to perform dynamic correction on the CMP gather.

3. The unsupervised automatic velocity spectrum picking method based on K-means clustering according to claim 1, characterized in that: The first set percentage is 10%; The second set percentage is 30%.

4. The unsupervised automatic velocity spectrum picking method based on K-means clustering according to claim 1, characterized in that: The third set percentage is 10%.

5. The unsupervised automatic velocity spectrum picking method based on K-means clustering according to claim 1, characterized in that, The cleaned data is then subjected to K-means clustering to obtain multiple near-surface cluster centers, multiple mid-level cluster centers, and multiple deep cluster centers, including: If there are multiple cluster center points in a local area, the center points of the multiple clusters are merged according to the principle of nearest time.

6. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the unsupervised automatic velocity spectrum picking method based on K-means clustering as described in any one of claims 1-5.

7. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing a computer to execute the unsupervised automatic velocity spectrum picking method based on K-means clustering as described in any one of claims 1-5.

8. An unsupervised automatic velocity spectrum acquisition device based on K-means clustering, characterized in that, include: The prediction line generation module is used to pick up multiple time-velocity pairs across the entire area and generate a prediction line using the multiple time-velocity pairs as control points. The constraint line generation module is used to generate a first constraint line and a second constraint line based on the predicted line, wherein the velocity trend of the first constraint line is lower than the velocity trend of the predicted line by a first set percentage, and the velocity trend of the second constraint line is higher than the velocity trend of the predicted line by a second set percentage. The data cleaning module is used to set similar velocity data that is lower than the first constraint linear velocity trend or higher than the second constraint linear velocity trend to zero, and to remove similar velocity data that is located between the first constraint line and the second constraint line and whose similarity to the predicted linear velocity trend is less than a third set percentage, so as to obtain cleaned data. The center point calculation module is used to perform K-means clustering on the cleaned data to obtain multiple near-surface cluster center points, multiple mid-level cluster center points, and multiple deep cluster center points; The velocity spectrum picking module is used to perform least-squares linear fitting on multiple near-surface cluster center points to obtain multiple near-surface time-velocity pairs. Calculate the percentage of the average velocity of multiple deep cluster centers to the velocity value of the corresponding point on the prediction line, and correct the deep cluster centers based on the percentage to obtain multiple deep time-velocity pairs.

9. The unsupervised automatic velocity spectrum picking device based on K-means clustering according to claim 7, characterized in that, Also includes: The CMP gather correction module is used to obtain a full-area three-dimensional velocity field based on multiple near-surface time-velocity pairs, multiple deep time-velocity pairs, and multiple mid-layer cluster centers, and to perform dynamic correction on the CMP gather using the full-area three-dimensional velocity field.

10. The unsupervised automatic velocity spectrum picking device based on K-means clustering according to claim 7, characterized in that: The first set percentage is 10%; The second set percentage is 30%; The third set percentage is 10%.

Images