A micro-motion exploration data processing method, system, medium, equipment and product for automatic partitioning of a two-dimensional dense array

Abstract

This invention proposes a method, system, medium, equipment, and product for automatic partitioning of micro-motion exploration data in a two-dimensional dense array, relating to the field of micro-motion exploration technology. The method includes: acquiring the spatial coordinates and waveform data of stations in the two-dimensional array, and performing batch preprocessing on the waveform data; performing sequential k-nearest neighbor rolling partitioning of stations based on their spatial coordinates; calculating the dispersion energy spectrum of stations in each partition based on the batch preprocessed waveform data, and using an automatic correction picking method to pick the dispersion energy spectrum; and establishing a modular workflow for data processing, including data import, data preprocessing, station partitioning, and dispersion calculation. This invention can significantly improve the data processing efficiency and effectiveness of micro-motion exploration in two-dimensional dense arrays.

Description

Technical Field

[0001] This invention relates to the field of micro-motion exploration technology, and in particular to a method, system, medium, equipment, and product for micro-motion exploration data processing based on automatic zoning of a two-dimensional dense array. Background Technology

[0002] Microseismic exploration is a seismic exploration method that utilizes array observations of weak vibration signals in the background noise of the environment to extract surface wave dispersion information, invert the shear wave velocity structure of the subsurface medium, and perform imaging. Surface waves of different frequencies have different penetration depths, and microseismic exploration utilizes this dispersion characteristic of surface waves. To obtain surface wave signals of different frequencies, it is necessary to adjust the size of the observation array, the observation duration, and the dominant frequency of the seismic pickups. Currently, in practical observations, microseismic exploration mainly achieves different detection depth requirements by setting different array observation radii. For two-dimensional planar array observation methods, dense arrays of seismic pickups (nodal seismographs) are often deployed to achieve high resolution and high productivity through high-density deployment. High-density deployment inevitably leads to cumbersome station combinations and zonal data. Currently, this observation method still faces the following challenges:

[0003] (1) At present, the two-dimensional micro-motion exploration processing technology mainly relies on the application of one-dimensional algorithms. There is no complete data processing workflow framework for two-dimensional data, which makes it impossible to achieve seamless connection between various processing modules and algorithms and visualized quality control of the process.

[0004] (2) At present, the two-dimensional dense array zoning technology mainly relies on manual selection and artificial combination of arrays for dispersion calculation, which is time-consuming and labor-intensive, prone to combination errors, and the coverage of the arrays after artificial combination is low. There is a lack of effective automatic zoning algorithms for the two-dimensional dense array observation method.

[0005] (3) Mainstream processing technologies cannot realize batch preprocessing calculation of massive array waveform data, and the extraction of dispersion energy spectrum from the partitioned array combination data still cannot be automatically calculated;

[0006] (4) Two-dimensional dispersion data with topography and apparent shear wave velocity data are difficult to generate three-dimensional volume maps, which cannot substantially guide production. Summary of the Invention

[0007] The purpose of this invention is to address the problems of low coverage and inability to automatically extract dispersion energy spectra in current two-dimensional dense array zoning technologies. This invention proposes a micro-motion exploration data processing method for automatic zoning of two-dimensional dense arrays, comprising the following steps:

[0008] S1. Obtain the spatial coordinates and waveform data of the stations in the two-dimensional array, and perform batch preprocessing on the waveform data of the stations.

[0009] S2. Based on the spatial location coordinates of the stations, perform sequential k-nearest neighbor rolling partitioning on the stations;

[0010] S3. Based on the pre-processed station waveform data, calculate the dispersion energy spectrum of the stations in each zone, and use an automatic correction picking method to pick up the dispersion energy spectrum.

[0011] S4. Establish a modular workflow for data processing, including data import, data preprocessing, station partitioning, and frequency calculation, and perform modular processing on steps S1-S3.

[0012] Furthermore, the station is partitioned using the automatic partitioning method of sequential k-nearest neighbor rolling partitioning, as detailed below:

[0013] The set of N stations in a two-dimensional array represented by coordinates is: ,Will As the first reference station, the first partition is constructed using the following formula:

[0014] ,

[0015] , ,

[0016] in, For the first partition to be constructed, Indicates station The set of k nearest neighbor stations, Representing stations Hetai Station Euclidean distance, Indicates partition The number of intermediate stations is k+1;

[0017] The remaining partitions are constructed as follows:

[0018] Scan stations in sequence The first station scanned that does not belong to the partition constructed in the previous step and has not been used as a reference station is taken as the new reference station. Use the following formula to construct partitions :

[0019] ,

[0020] in, Indicates station The set of k nearest neighbor stations;

[0021] Repeat the scanning of stations until all stations are partitioned.

[0022] Furthermore, the central stations for each zone are determined using the following method:

[0023] ,

[0024] ,

[0025]

[0026] in, This represents the sum of Euclidean distances between all stations in the partition, where m represents the number of stations in the partition. Indicates station Hetai Station Euclidean distance, This represents the coordinates of the k-th station. This represents the coordinates of the l-th station. The index represents the coordinates of the central station. The coordinates of the central stations in each partition are as follows: .

[0027] Furthermore, based on the size of the waveform data from the stations in each partition, a data size threshold is set. If the data volume of all stations in a partition exceeds the threshold, a multi-threaded parallel processing mode is adopted; otherwise, a serial sequential processing mode is adopted.

[0028] Furthermore, the automatic correction pickup method specifically represents the pickup dispersion energy spectrum as follows:

[0029] Given dispersion The velocity vector of each value , Frequency vector of values and energy matrix , then collect the points and the speed range of input First, correct the frequency according to the following formula:

[0030]

[0031] in, Represents frequency Corrected index;

[0032] Constructing a range on the velocity axis And denote the set of indices as ,like If the speed is zero, then maintain the original speed value; otherwise, let:

[0033]

[0034] in, Indicates speed Corrected index, Indicates frequency as The speed is The energy of the point, Indicates the speed range of the input;

[0035] Output pick point The corrected pickup point is .

[0036] Furthermore, the nearest-neighbor inverse distance weighted interpolation algorithm is used to calculate the dispersive energy spectrum, specifically as follows:

[0037] sampling points Organized into a three-dimensional Trees, among which, It is the spatial position of the i-th velocity point. It is a location The velocity value at point N is the total number of velocity points.

[0038] Given query point Nearest neighbor search proceeds down the tree from top to bottom, simultaneously obtaining a nearest neighbor with a size not exceeding [a certain value]. ordered set ;in , Indicates the first The spatial location of each velocity point, during backtracking, if the distance from the query point to the splitting hyperplane is less than the current set... The farthest Continue into the other subtree, otherwise prune;

[0039] based on Extract the k nearest samples to each node from the tree; if the actual number of returned samples is less than k, directly calculate the arithmetic mean. When the number of samples reaches k, the first-order inverse distance weights are used:

[0040]

[0041]

[0042] in, Indicates at the query point The interpolation result at point m represents the actual number of samples returned that are less than k. Indicates the first The velocity values ​​of the nearest neighbor samples, This represents the interpolation result after weighted averaging of velocity values. Indicates the first The weight of each nearest neighbor, Indicates the first The Euclidean distance from the nearest neighbor sample to the query point A constant value introduced to avoid division by zero when nodes coincide with samples.

[0043] This invention also proposes a micro-motion exploration data processing system for automatic partitioning of a two-dimensional dense array, comprising:

[0044] The data acquisition module is used to acquire the spatial coordinates of the stations in the two-dimensional array, the station waveform data, and to perform batch preprocessing on the station waveform data.

[0045] The station partitioning module is used to sequentially partition stations based on their spatial location coordinates. Nearest neighbor rolling partition;

[0046] The dispersive energy spectrum calculation module is used to calculate the dispersive energy spectrum of stations in each zone based on the pre-processed station waveform data in batches, and to pick up the dispersive energy spectrum using an automatic correction picking method.

[0047] The workflow building module is used to build a modular workflow for data processing, including data import, data preprocessing, station partitioning, and frequency calculation, and to perform modular processing on steps S1-S3.

[0048] The present invention also proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method for micro-motion exploration data processing of automatic partitioning of a two-dimensional dense array.

[0049] The present invention also proposes an electronic device, including a processor and a memory, wherein the processor and the memory are interconnected, wherein the memory is used to store a computer program, the computer program including computer-readable instructions, and the processor is configured to invoke the computer-readable instructions to execute the above-described method for micro-motion exploration data processing of automatic partitioning of a two-dimensional dense array.

[0050] The present invention also proposes a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the above-described micro-motion exploration data processing method for automatic partitioning of a two-dimensional dense array.

[0051] The beneficial effects of the technical solution provided by this invention are:

[0052] The order of the invention The automatic partitioning method using nearest-neighbor rolling partitions skips the previous partition in each rolling partition, maintaining geographical proximity while ensuring overall coverage continuity. After loading the array data for the specified partition, batch dispersive energy spectrum calculations are performed. The overall task is broken down into fine-grained steps according to partitions and queues are generated for rational resource allocation. This invention proposes an automatic correction picking method for dispersive energy spectrum picking, searching for the minimum dispersive energy value only within a local velocity window to achieve correction. This invention first performs batch preprocessing of waveform data, then performs automatic array partitioning, and uses the partitioned array data for batch spectrum calculation, automatic correction and picking of dispersive data, and three-dimensional nearest-neighbor inverse distance weighted interpolation mapping. The entire process significantly improves the data processing efficiency and effectiveness of two-dimensional dense array micro-motion exploration. Attached Figure Description

[0053] Figure 1 This is a flowchart of a micro-motion exploration data processing method for automatic partitioning of a two-dimensional dense array according to an embodiment of the present invention;

[0054] Figure 2 This is a schematic diagram of partial overlap of station zones according to an embodiment of the present invention. Figure 2 Distribution type 1 in (a) is a diagonal extension with local overlap. Figure 2 Distribution type 2 in (b) is characterized by a dense central area with local overlap. Figure 2 Distribution type 3 in (c) is characterized by a local overlap between dense on the left and sparse on the right. Figure 2 Distribution type 4 in (d) is a parallel arrangement with local overlap;

[0055] Figure 3 This is a block diagram of a modular workflow for data processing according to an embodiment of the present invention.

[0056] Figure 4 This is a two-dimensional dense array observation image loaded in an embodiment of the present invention;

[0057] Figure 5 This is a comparison chart of dispersive pickup methods. The left chart shows the traditional pickup method, and the right chart shows the automatic correction pickup method of this invention.

[0058] Figure 6 This is a micro-motion exploration three-dimensional dispersion map obtained by the method of this embodiment of the invention;

[0059] Figure 7 This is a three-dimensional image of micro-motion exploration obtained by the method of this embodiment of the invention;

[0060] Figure 8 This is a three-dimensional horizontal slice image of micro-motion exploration obtained by the method of this embodiment of the invention;

[0061] Figure 9 This is a micro-motion exploration three-dimensional vertical slice image obtained by the method of this embodiment of the invention;

[0062] Figure 10 This is a block diagram of an electronic device according to an exemplary embodiment of the present invention. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0064] The flowchart of the micro-motion exploration data processing method for automatic partitioning of a two-dimensional dense array according to an embodiment of the present invention is as follows: Figure 1 Specifically, it includes the following steps:

[0065] S1. Obtain the position coordinates and waveform data of the stations in the two-dimensional array, and perform batch preprocessing on the waveform data.

[0066] Specifically, the collected raw station waveform data needs to undergo rapid batch preprocessing using methods such as downsampling, mean removal, trend removal, moving absolute average normalization, spectral whitening, and Gaussian filtering. Before batch preprocessing, the amount of raw waveform data needs to be assessed, and a threshold is set according to requirements. Data larger than the threshold is considered a large dataset, while data smaller than the threshold is considered a small dataset. Small datasets are processed sequentially, while large datasets are processed in a multi-threaded parallel manner. In parallel processing, the pipeline follows the logic of thread pool allocation → parallel channel processor → single-channel data processor → single-channel data processing pipeline; in serial processing, the pipeline follows the logic of sequential channel processing → serial channel processor → single-channel data processing pipeline.

[0067] S2. Based on the spatial coordinates of the stations, perform sequential analysis of the stations. Nearest neighbor rolling partition.

[0068] In a preferred embodiment of the present invention, an automatic partitioning method based on sequential k-nearest neighbor rolling partitioning is used to partition the stations, with each partition containing k+1 stations, as detailed below:

[0069] Two-dimensional array represented by coordinates N The stations are grouped together as follows ,Will As the first reference station, the first partition is constructed using the following formula:

[0070] ,

[0071] , ,

[0072] in, For the first partition to be constructed, Indicates station of The nearest neighbor stations are collected. Representing stations Hetai Station Euclidean distance, Indicates partition The number of central stations is +1.

[0073] The Euclidean distance between stations is calculated using the following formula:

[0074] ,

[0075] in, Indicates station Hetai Station The distances between them are used to construct an N×N Euclidean distance matrix, which facilitates automatic partitioning of the stations.

[0076] The remaining partitions are constructed as follows:

[0077] Scan stations in sequence The first station scanned that does not belong to the partition constructed in the previous step and has not been used as a reference station is taken as the new reference station. Use the following formula to construct partitions :

[0078] ,

[0079] in, Indicates station The set of k nearest neighbor stations.

[0080] Repeat the scanning of stations until all stations are partitioned.

[0081] order Nearest neighbor automatic zoning requires a comprehensive balance between the number of two-dimensional dense arrays and the number of zoning zones in the target area. That is, the number of zoning zones cannot be too large, while the number of sub-arrays after zoning is too small, which cannot meet the detection depth requirements; nor can the number of zoning zones be too small, while the number of sub-arrays after zoning is too large, which cannot meet the spatial coverage and lateral resolution requirements.

[0082] By constructing the Euclidean distance matrix of all array stations and setting the number of stations in each partition, a sequential scanning method is used to select stations that do not belong to the previous partition and have not been used as reference stations as reference points for the new partition, and based on... The nearest neighbor principle is used to select other stations to form partitions. This method skips the reference station of the previous partition each time it is selected, so that adjacent partitions maintain local overlap, which not only ensures geographical proximity, but also achieves continuous coverage of the entire array.

[0083] The central station for each zone is determined in the following way:

[0084] ,

[0085] ,

[0086]

[0087] in, This represents the sum of the Euclidean distances of all stations in the partition. m Indicates the number of stations in the partition. Indicates station Hetai Station Euclidean distance, This represents the coordinates of the k-th station. Indicates the first The coordinates of each station The index represents the coordinates of the central station. The coordinates of the central stations in each partition are as follows: .

[0088] At this point, each partition contains exactly k nearest stations. Since the reference station skips the previous partition each time, partial overlap occurs between partitions, maintaining both geographical proximity and ensuring continuous overall coverage. (See the schematic diagram of partial overlap in this embodiment of the invention.) Figure 2 (a)-(d) in:

[0089] Gray triangles represent observatories, while red ones are marked with... For the reference stations selected in this round, each black rectangle represents a reference station and its... A partition is composed of its nearest neighbors.

[0090] Figure 2 (a) The central station extends roughly along the diagonal, with windows from... Smoothly advance to The two partitions overlap only in a small part in the middle, reflecting the characteristic of local close proximity;

[0091] Figure 2 (b) The center is more densely packed, so start with Scroll to the top (in chunks) and bottom right This forms two or three overlapping areas at the center point, acting as a bridge. The upper and lower right clusters are connected into one through the central dense area.

[0092] Figure 2 (c) It exhibits a characteristic of being denser on the left and sparser on the right. Here, it is used as... The resulting left partition is for obtaining -1 nearest neighbor is stretched into a horizontal strip for coverage. Then, on the sparse right side... A vertical rectangle is formed, and the two only intersect at the narrow part of the transition zone, indicating that the window will adaptively lengthen along the main density direction while maintaining the minimum necessary overlap;

[0093] Figure 2 (d) It is almost divided into two columns, left and right. First, start with the top of the left column. Once you have the partitions for the entire left column, the next reference point... Another block is formed in the right column, and the two overlap only near the midline. At this point, the partitioning result is two parallel stripes, which avoids forcibly mixing distant clusters while maintaining global continuous coverage.

[0094] S3. Based on the pre-processed station waveform data, calculate the dispersion energy spectrum of the stations in each partition, and use an automatic correction picking method to pick up the dispersion energy spectrum.

[0095] After completing the automatic partitioning of the two-dimensional dense array, the dispersive energy spectrum is configured. First, the number of CPU cores needs to be configured according to the computing system resources, that is, the number of cores for parallel computing; second, the array data after the specified partitioning is loaded, and the batch dispersive energy spectrum is calculated using spatial autocorrelation methods or other methods.

[0096] The specific execution process of batch computing proposed in this method is as follows: The system begins by initializing the executor and reading parameters, then breaks down the overall task into fine-grained parts according to "partitioning + data processing method" and generates queues. Subsequently, the intelligent thread pool configures resources based on the number of physical cores, reserved system resources, and concurrency limits. The scheduling center uniformly manages running threads and waiting queues. Tasks are retrieved sequentially and delivered to the computing processors. The completed computing results are collected and persistently stored in the data management center, and a unique identifier consisting of partition, station, method, and timestamp is generated for each result. Meta-information is saved along with the results. The system performs status checks, triggers subsequent batches, and performs resource reclamation upon completion during task rotation.

[0097] In terms of thread and scheduling processes, the system first performs CPU resource assessment and concurrency calculation to establish the initial state of the thread pool. The dynamic load balancer continuously checks, and if the conditions are met, it retrieves the next task from the queue, creates a computing thread and allocates independent processing space, while establishing progress, completion, and error signal connections, registering the thread in the active list and starting the computation.

[0098] During the computation phase, tasks are intelligently routed to the appropriate paths, with real-time progress updates. After a task is completed, the status is reported, results are collected, thread resources are reclaimed, completion status and global progress are updated, and the next round of scheduling is triggered after persistent storage.

[0099] To address the tedious manual acquisition of dispersive energy spectra, this invention proposes an automatic correction acquisition method. The automatic correction acquisition method for acquiring dispersive energy spectra is specifically described as follows:

[0100] Given dispersion The velocity vector of each value , Frequency vector of values and energy matrix , then collect the points and the speed range of input First, correct the frequency according to the following formula:

[0101]

[0102] in, Represents frequency Corrected index;

[0103] Constructing a range on the velocity axis And denote the set of indices as ,like If the speed is zero, then maintain the original speed value; otherwise, let:

[0104]

[0105] in, Indicates speed Corrected index, Indicates frequency as speed is The energy of the point, Indicates the speed range of the input;

[0106] Output pick point Corrected pickup point This algorithm achieves correction by searching for the minimum dispersion energy only within a local velocity window. It's worth noting that some dispersion calculation methods focus on maxima, and the dispersion energy spectrum has already undergone dispersion energy processing before picking, such as taking the reciprocal. Therefore, selecting the minimum point here is reasonable.

[0107] To address the computational resource consumption of 3D dispersion data, the time-consuming Kriging interpolation method needs to be abandoned. This invention uses a nearest-neighbor inverse distance weighted interpolation algorithm. The specific calculation approach is as follows:

[0108] First of all N A sampling point with a scalar value , Organized into a three-dimensional Trees. Among them, It is the spatial position of the i-th velocity point. Is a scalar field at position The known observed value at that location, that is, the velocity value. N This represents the total number of velocity points. At depth d, a split axis is taken, and multiplied by the median of the coordinates on that axis to form a hyperplane. This divides the point set in two and recursively continues until the subset is empty. The resulting tree height is approximately logarithmic, allowing subsequent nearest neighbor queries to access only a small number of nodes.

[0109] Given query point The nearest neighbor search proceeds from the top of the tree down to the leaves, while maintaining an ordered set of size no more than k. ;in , Indicates the first The spatial location of each velocity point. During backtracking, if the distance from the query point to the splitting hyperplane is less than the current set... The farthest If the condition is met, proceed to the other subtree; otherwise, prune the branch. This ensures that the final result contains the k samples closest to q.

[0110] Apply local interpolation to each node of the 3D regular mesh: first, by using... Extract the k nearest samples to each node from the tree; if the actual number of returned samples is less than k, directly calculate the arithmetic mean. When the sample reaches When the distance is 1, then the first-order inverse distance weight is used:

[0111]

[0112]

[0113] in, Indicates at the query point The interpolation result at that point, m This indicates the actual number of samples returned that are less than k. Indicates the first The scalar value of the nearest neighbor sample, that is, the velocity value. This represents the interpolation result after weighted averaging of velocity values. Indicates the first The weights are calculated using the nearest neighbors, specifically the first-order inverse distance weights. Indicates the first The Euclidean distance from the nearest neighbor sample to the query point A constant value introduced to avoid division by zero when nodes coincide with samples.

[0114] This method is essentially to Tree acceleration Combining nearest neighbor search with first-order IDW interpolation can provide an empirical approximation of Kriging interpolation in the absence of a covariance model, if we assume that the variogram is linear or only consider distance weights. This improves computational efficiency.

[0115] In the process of interpolation calculation for 3D volume data visualization, nearest neighbor points The value can be viewed as a discrete substitute for the core support radius, and the estimated value is:

[0116]

[0117] Indicates at the query point / grid node The interpolation result at that point, Indicates the current position to be interpolated. This represents the Euclidean distance from the i-th neighboring sample to the query point. Interpolation is equivalent to applying an Euclidean algorithm to the original scatter plot. For kernel function ( d (representing Euclidean distance), the supporting region is the distance of the k-th nearest neighbor. The convolution is determined by k. As k increases, As the frequency increases, the interpolation behavior gradually shifts from high-pass to low-pass. From a Fourier transform perspective, this is equivalent to changing the cutoff frequency of the signal transfer function. Depend on Shift to lower frequencies:

[0118]

[0119] in, This represents the frequency response amplitude of the smoothing operator. Indicates the first Spatial frequencies of the nearest neighbors.

[0120] Therefore, in three-dimensional interpolation of volumetric maps, small Displays sharp textures and isolated peaks, large This softens the lines and smooths out noise. It's necessary to select nearest neighbor points appropriately based on the characteristics of the dispersion data and the interpolation effect. Value range.

[0121] S4. Based on the above steps, create a modular data processing workflow that includes data import, preprocessing, partitioning, and frequency calculation. A block diagram of this invention's modular data processing workflow is shown below. Figure 3The system configures parameters in three modules: data preprocessing, array partitioning, and frequency dispersion calculation, enabling batch computation. This is a data processing workflow that involves organizing, processing, and performing overall quality control of data. The modular workflow reduces intermediate data storage and transmission, minimizes network bandwidth consumption, and improves processing efficiency through parallel computing. The relatively independent nature of each module effectively enhances reusability and maintainability.

[0122] This invention significantly improves the data processing efficiency and effectiveness of two-dimensional dense array micro-motion exploration by creating a data processing workflow, batch preprocessing of waveform data, automatic partitioning of the array, batch calculation of spectrum of partitioned array data, automatic correction and picking of dispersion data, and three-dimensional nearest neighbor point inverse distance weighted interpolation mapping.

[0123] Taking microseismic exploration data measured in the field in a certain area as an example, a total of 192 single-component nodal seismographs were deployed in this area, with reference to... Figure 4 , Figure 4 This is a two-dimensional dense array observation map loaded in an embodiment of the present invention. The observation system is an irregular rectangle, with an average spacing of approximately 15m between the seismic instruments at each node in the two-dimensional dense array. The spacing varies at some locations due to obstacles. The node instruments have a main frequency of 5Hz, a sampling rate of 4ms, and an acquisition time of 90 minutes. Data from 192 micromotion instruments were actually collected, with a total data size of approximately 1GB.

[0124] Based on the background noise characteristics of the target area, a combination of signal processing modules, including mean removal, trend removal, sliding absolute smoothing normalization, spectral whitening, and filtering, is selected to perform batch waveform preprocessing before automatic array partitioning. This eliminates the need to manually repeat the same processing steps for all 192 stations. Calculations show that the total preprocessing time for all waveform data from the 192 stations is approximately 8 seconds, demonstrating excellent computational efficiency.

[0125] The preprocessed data stream was partitioned using the sequential k-nearest neighbor partitioning method, with each partition containing 21 stations, resulting in a total of 51 sub-partitions. A selected sub-partition after partitioning is shown below. Figure 4 As shown, the red triangle represents the currently selected partition, and the gray triangle represents the unselected partition. Similarly, the automatic partitioning algorithm achieves high coverage and high efficiency in dividing the workspace into arrays, eliminating the need for manual, mechanical combination of 192 data points. Calculating a partition takes approximately 3 seconds. If a partition is not satisfactory, the number of partitions can be adjusted, and the calculation can be recalculated. Table 1 lists the sequential partitioning... The statistics table of the first 4 partitions in the nearest neighbor partition array is automatically divided, and each sub-partition is given the corresponding 21 station numbers and the number of the central station, so as to facilitate the next step of dispersive energy spectrum calculation.

[0126] Table 1

[0127]

[0128] For all partitions, a dispersion calculation method was specified. This invention selected the spatial autocorrelation method, with a time window of 30 seconds, a minimum frequency of 1 Hz, a maximum frequency of 20 Hz, a minimum phase velocity of 50 m / s, and a maximum phase velocity of 4500 m / s, for batch dispersion energy spectrum calculation. The dispersion energy spectrum calculation for all 51 sub-partitions was completed in 39 seconds, demonstrating very high computational efficiency. This saves a significant amount of manual, repetitive calculations for individual sub-partitions.

[0129] After batch calculations of the dispersion energy spectrum, the phase velocity data can be acquired using an automatic correction acquisition method. (See the comparison chart of dispersion acquisition methods for reference.) Figure 5 The left image shows the traditional pickup method, and the right image shows the automatic correction pickup method of this invention. Figure 5 It is evident that using automatic correction partitioning can more quickly and accurately identify the inflection point morphological characteristics of the dispersive energy spectrum and perform frequency point picking that more closely approximates the true morphology. In contrast, traditional manual picking methods are prone to false anomalies, such as... Figure 5 The inflection point marked by the red arrow is not the true shape of the dispersion curve. Therefore, using an automatic correction picking method can avoid repeated rework due to the inability to accurately pick the dispersion inflection point.

[0130] This invention sets the number of neighboring points to 16, the horizontal grid resolution to 30m, and the vertical grid resolution to 30m when performing 3D inverse distance weighted interpolation of neighboring points. Furthermore, elevation information is added to the coordinate file; that is, 3D dispersion data with terrain is used for interpolation, which better meets actual production requirements. (In 3D vertical slicing...) Figure 8 The terrain undulations can be seen in the two-dimensional velocity profile.

[0131] In summary, the specific calculation time consumed by this invention is shown in Table 2. Compared with the conventional processing flow (traditional method) that does not perform automatic partitioning, batch preprocessing, and automatic correction picking, the method of this invention performs excellently in workflow creation, waveform preprocessing, automatic partitioning, dispersion energy spectrum calculation, dispersion picking, and three-dimensional interpolation for two-dimensional dense array data processing. The time required for each step does not exceed 1 minute. Among them, waveform preprocessing and dispersion energy spectrum calculation use data from 51 partitioned stations as comparison indicators, while automatic partitioning uses data from 192 stations for partitioning as comparison, and dispersion picking only compares the array of a single partition. It can be seen that the method of this invention has a qualitative improvement in computational efficiency, thereby greatly shortening the project processing cycle and saving a lot of time and trial and error costs.

[0132] Table 2

[0133]

[0134] The method of this embodiment of the invention provides a reference for micro-motion exploration three-dimensional dispersion map. Figure 6 This figure compiles the apparent shear wave velocity curves calculated after all partitioning, serving as preparatory data for 3D volume data interpolation. In the figure, blue markers represent low-velocity values, and red markers represent high-velocity values. The 3D dispersion data exhibits shallow low-velocity (blue markers) and deep high-velocity (red markers). The quality and velocity distribution characteristics of the data can be roughly determined from the 3D dispersion data points.

[0135] Reference image of micro-motion exploration obtained by the method of this embodiment of the invention Figure 7 In the three-dimensional volume, the shallow part along the Z-direction is a low-velocity region, corresponding to the Quaternary overburden, while the deep part is a high-velocity region, corresponding to the limestone strata. The three-dimensional volume data more intuitively shows the stratigraphic spatial division of the exploration target area, as well as the overall spatial location of anomalous areas such as karst bands.

[0136] Reference image of micro-motion exploration three-dimensional horizontal slice obtained by the method of this embodiment of the invention Figure 8 Along the XY plane, in a horizontal slice at a depth of -60m in the Z direction, the velocity in the profile shows localized low-velocity blue areas and striped blue velocity distribution areas. These low-velocity patterns can provide important references for determining the spatial location of limestone karstification and karst development.

[0137] Reference image of micro-motion exploration three-dimensional vertical slice obtained by the method of this embodiment of the invention Figure 9 Vertical slices of the three-dimensional volume map along the YZ plane, unlike horizontal slices, reflect the velocity changes in the vertical direction. From the blue low-velocity layer on the undulating surface at an elevation of 14m to the relatively high-velocity strata below an elevation of -25m, vertical slices can intuitively depict the vertical variation characteristics of the strata.

[0138] The three-dimensional volumetric map and various slice maps calculated by this invention can directly guide the deployment of boreholes in field exploration, improving the basis and accuracy of borehole deployment.

[0139] This invention also proposes a micro-motion exploration data processing system for automatic partitioning of a two-dimensional dense array, comprising:

[0140] The data acquisition module is used to acquire the spatial coordinates of the stations in the two-dimensional array, the station waveform data, and to perform batch preprocessing on the station waveform data.

[0141] The station partitioning module is used to sequentially partition stations based on their spatial location coordinates. Nearest neighbor rolling partition;

[0142] The dispersive energy spectrum calculation module is used to calculate the dispersive energy spectrum of stations in each zone based on the pre-processed station waveform data in batches, and to pick up the dispersive energy spectrum using an automatic correction picking method.

[0143] The workflow building module is used to build a modular workflow for data processing, including data import, data preprocessing, station partitioning, and frequency calculation, and to perform modular processing on steps S1-S3.

[0144] In one exemplary embodiment, a computer-readable storage medium is included, which stores a computer program that, when executed by a processor, implements the above-described micro-motion exploration data processing method for automatic partitioning of a two-dimensional dense array.

[0145] Please see Figure 10 In one exemplary embodiment, the device further includes an electronic device including at least one processor, at least one memory, and at least one communication bus.

[0146] The memory stores a computer program, which includes computer-readable instructions. The processor calls the computer-readable instructions stored in the memory through the communication bus to execute the micro-motion exploration data processing method for automatic partitioning of the two-dimensional dense array.

[0147] In one exemplary embodiment, a computer program product is proposed, including a computer program / instruction that, when executed by a processor, implements the steps of the micro-motion exploration data processing method for automatic partitioning of a two-dimensional dense array described above.

[0148] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for processing microtremor exploration data by two-dimensional dense array automatic partitioning, characterized by, Includes the following steps: S1. Obtain the spatial coordinates and waveform data of the stations in the two-dimensional array, and perform batch preprocessing on the waveform data of the stations. S2. Based on the spatial location coordinates of the stations, perform sequential k-nearest neighbor rolling partitioning on the stations; S3. Based on the pre-processed station waveform data, calculate the dispersion energy spectrum of the stations in each zone, and use an automatic correction picking method to pick up the dispersion energy spectrum. S4. Establish a modular workflow for data processing, including data import, data preprocessing, station partitioning, and frequency calculation, and perform modular processing on steps S1-S3. The stations are partitioned using the automatic partitioning method of sequential k-nearest neighbor rolling partitioning, as detailed below: The set of N stations in a two-dimensional array represented by coordinates is: ,Will As the first reference station, the first partition is constructed using the following formula: ， ， ， in, For the first partition to be constructed, Indicates station The set of k nearest neighbor stations, Representing stations Hetai Station Euclidean distance, Indicates partition The number of intermediate stations is k+1; The remaining partitions are constructed as follows: sequentially scanning stations the first scanned station not belonging to the partition built in the previous step and not being a reference station as a new reference station building the partition using the following formula : ， wherein, represents a set of k nearest stations to the station ; Repeat the scanning of stations until all stations are partitioned.

2. The method according to claim 1, wherein, The central station for each zone is determined in the following way: ， ， in, This represents the sum of the Euclidean distances of all stations in the partition. m Indicates the number of stations in the partition. Indicates station Hetai Station Euclidean distance, Indicates the first The coordinates of each station Indicates the first l The coordinates of each station The index represents the coordinates of the central station. The coordinates of the central stations in each partition are as follows: .

3. The method for automatic partitioning of micro-motion exploration data in a two-dimensional dense array according to claim 1, characterized in that, Based on the size of the waveform data of the stations in each partition, a data size threshold is set. If the data volume of all stations in a partition exceeds the threshold, a multi-threaded parallel processing mode is adopted. Conversely, a serial sequential processing mode is used.

4. The method of claim 1, wherein, The automatic correction pickup method for picking up dispersion energy spectrum is specifically represented as follows: given a velocity vector of values , a frequency vector of values , an energy matrix , and an input velocity span , first correct the frequencies according to the following formula: wherein indicates the frequency corrected index; Constructing a range on the velocity axis And denote the set of indices as ,like If the speed is zero, then maintain the original speed value; otherwise, let: in, Indicates speed Corrected index, Indicates frequency as The speed is The energy of the point, Indicates the speed range of the input; Output pickup point The corrected pickup point is .

5. The method of claim 1, wherein, The nearest neighbor inverse distance weighted interpolation algorithm is used to calculate the dispersion energy spectrum, specifically as follows: sampling points Organized into a three-dimensional Trees, among which, It is the spatial position of the i-th velocity point. It is a location The velocity value at that location, N It is the total number of speed points; Given query point Nearest neighbor search proceeds down the tree from top to bottom, simultaneously obtaining a nearest neighbor with a size not exceeding [a certain value]. ordered set ;in , Indicates the first The spatial location of each velocity point, during backtracking, if the distance from the query point to the splitting hyperplane is less than the current set... The farthest Continue into the other subtree, otherwise prune; Based on The tree fetches the last k samples; if the actual number of returned samples is less than k, the arithmetic mean is directly calculated When the samples reach k, the first inverse distance weight is used: in, Indicates at the query point The interpolation result at that point, m This indicates the actual number of samples returned that are less than k. Indicates the first The velocity values ​​of the nearest neighbor samples, This represents the interpolation result after weighted averaging of velocity values. Indicates the first The weight of each nearest neighbor, Indicates the first The Euclidean distance from the nearest neighbor sample to the query point A constant value introduced to avoid division by zero when nodes coincide with samples.

6. A microtremor survey data processing system for automatic partitioning of a two-dimensional dense array, characterized by, include: The data acquisition module is used to acquire the spatial coordinates of the stations in the two-dimensional array, the station waveform data, and to perform batch preprocessing on the station waveform data. The station partitioning module is used to sequentially partition stations based on their spatial location coordinates. Nearest neighbor rolling partition; The dispersive energy spectrum calculation module is used to calculate the dispersive energy spectrum of stations in each zone based on the pre-processed station waveform data in batches, and to pick up the dispersive energy spectrum using an automatic correction picking method. The workflow building module is used to build a modular workflow for data processing, including data import, data preprocessing, station partitioning, and frequency calculation, and to modularize steps S1-S3. The stations are partitioned using the automatic partitioning method of sequential k-nearest neighbor rolling partitioning, as detailed below: The set of N stations in a two-dimensional array represented by coordinates is: ,Will As the first reference station, the first partition is constructed using the following formula: ， ， ， in, For the first partition to be constructed, Indicates station The set of k nearest neighbor stations, Representing stations Hetai Station Euclidean distance, Indicates partition The number of intermediate stations is k+1; The remaining partitions are constructed as follows: Scan stations in sequence The first station scanned that does not belong to the partition constructed in the previous step and has not been used as a reference station is taken as the new reference station. Use the following formula to construct partitions : ， wherein, represents a set of k nearest stations to the station ; Repeat the scanning of stations until all stations are partitioned.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it implements the method as described in any one of claims 1-5.

8. An electronic device, characterized in that, The device includes a processor and a memory, the processor being interconnected with the memory, wherein the memory is used to store a computer program, the computer program including computer-readable instructions, and the processor is configured to invoke the computer-readable instructions to perform the method as described in any one of claims 1-5.

9. A computer program product comprising computer programs / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-5.

Images