Three-dimensional data quality evaluation method based on conjoint analysis of attributes of adjacent shot gathers

Abstract

The invention belongs to the technical field of geophysical exploration, relates to a quality evaluation method, and particularly discloses a three-dimensional data quality evaluation method based on conjoint analysis of attributes of adjacent shot gathers, which comprises the following steps of: 1, extracting attribute characteristics of three-dimensional shot gather data; 2, grouping the attribute basic data; step 3, identifying an abnormal value; 4, reconstructing an abnormal value and a missing value; and 5, forming a visual data quality evaluation map. According to the method, the attribute features of the data are divided into a plurality of adjacent areas according to the specified distance range through the clustering algorithm, the accuracy of the three-dimensional data and the processing efficiency of the three-dimensional data are improved through the methods of anomaly recognition, data reconstruction and the like, and the method is suitable for accurately, efficiently and comprehensively evaluating the quality of mass three-dimensional data.

Description

Technical Field

[0001] This invention belongs to the field of geophysical exploration technology and relates to quality evaluation methods, specifically a three-dimensional data quality evaluation method based on joint analysis of neighboring shot gather attributes. Background Technology

[0002] The Ordos Basin has fully entered the era of 3D seismic nodal acquisition, playing a vital role in the submission of oil and gas field reserves and the development of oil and gas fields. With the large-scale application of high-density, high-coverage, and nodal acquisition technologies, the amount of seismic data has grown exponentially, with each shot receiving over 10,000 channels, and large-scale projects acquiring over 100,000 shots.

[0003] Currently, commercial software such as Krono and RTQC real-time monitoring software are commonly used to evaluate the quality of large-scale 3D data. The method typically involves comparing the data with standard shot data, using the quality of the standard shot as a threshold for shot gather quality. Specifically, after seismic data acquisition, it is directly compared with standard shot data to classify the shot gather quality. However, this evaluation method has some shortcomings:

[0004] (1) It is usually difficult to select a standard shot, and the quality of the selection directly affects the evaluation results, resulting in low accuracy.

[0005] (2) The evaluation effect is poor in areas with low signal-to-noise ratio. For example, two shots with the same excitation factors have similar energy amplitude levels, but due to environmental interference, some channels have excessive amplitude, resulting in abnormal total energy after summing the RMS values ​​of all channels, which is not a true reflection of the actual situation.

[0006] (3) The evaluation did not take into account the impact of local environmental interference or geological and topographical factors caused by changes in spatial location on the quality of data, making it difficult to fully and comprehensively grasp the quality of three-dimensional data from multiple perspectives.

[0007] The above problems result in a lack of comprehensive and effective quality assessment of seismic data, leading to low efficiency and accuracy in the assessment. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention aims to provide a method for evaluating the quality of 3D data based on joint analysis of neighboring shot gather attributes, so as to achieve an accurate, efficient, and comprehensive evaluation of the quality of raw data from massive 3D datasets.

[0009] To achieve the above objectives, the technical solution adopted by the present invention is as follows: including the following steps:

[0010] S1, Extract the attribute features of the 3D shot gather data: Input the 3D seismic data into the data processing software system, extract the attribute features of the shot gather data, and output the basic attribute data of each trace of the seismic data in text format.

[0011] S2, grouping the basic attribute data: using a clustering algorithm to divide the basic attribute data obtained in S1 into multiple neighboring regions according to a specified distance range, and obtaining several groups of neighboring region cluster attribute values;

[0012] S3, Identify outliers: The local outlier factor algorithm is used to detect outliers in each group of neighboring shot gather attribute values ​​and identify outlier attribute values ​​in each group of neighboring shot gather attribute values.

[0013] S4, Reconstructing outliers and missing values: Machine learning algorithms are used to process the normal attribute values ​​in each group to obtain curves with multiple data statistics. By resampling the curves, the reconstructed values ​​of the outlier and missing value positions are obtained, thus completing the reconstruction of the outlier attribute values ​​and missing data in each group obtained in S3.

[0014] S5. A visual data quality evaluation map is generated. The attribute values ​​of all traces in each shot set of all adjacent shot sets after reconstruction are summed to obtain the new attribute values ​​of all sample points. The data quality evaluation map of the whole area is generated through visualization.

[0015] As a limitation of the present invention: the three-dimensional seismic data in S1 is statically corrected shot gather data.

[0016] As a limitation of the present invention, the attribute characteristics of the shot collection data in S1 include energy, main frequency, and signal-to-noise ratio.

[0017] As a limitation of the present invention: the process of extracting basic attribute data using a data processing software system in S1 is as follows:

[0018] ① Use the seismic trace attribute extraction module of the data processing software system to calculate the attribute value of each trace of the seismic data;

[0019] ② Use the seismic gather attribute output module of the data processing software system to output the attribute values ​​of each trace of the seismic data, that is, to obtain the basic attribute data of each trace of the seismic data in text format.

[0020] As a limitation of the present invention: the clustering algorithm in S2 is the mean clustering algorithm, and the specific steps are as follows:

[0021] ① Randomly select n center points, that is, divide the basic attribute data into n groups according to the specified distance range. The value of n is selected according to the regional terrain, computer configuration, and the size of the data to be processed.

[0022] ② Calculate the distance from each sample to each of the n center points;

[0023] ③ Assign the sample points to the clusters closest to the center point;

[0024] ④ Calculate the mean of the sample points in each cluster as the new center point;

[0025] ⑤ Iterate until the results are the same or the termination condition is met, and complete the data grouping.

[0026] As a limitation of the present invention: the local outlier factor algorithm in S3 is based on the density estimation strategy to identify outliers. The outlier obtained by the density-based outlier diagnosis step is the abnormal attribute value in the shot gather attribute value of each group of neighboring regions.

[0027] As a limitation of the present invention, the formula used in the local outlier factor algorithm includes:

[0028] Local neighborhood density calculation:

[0029]

[0030] Outlier factor (i.e., relative density) calculation:

[0031]

[0032] Where N(x,k) is the set of k-nearest neighbors that do not contain x, |N(x,k)| is the size of the set, y is a set of nearest neighbors, and distance(x,y) is the reachability distance.

[0033] As a limitation of the present invention: the machine learning algorithm in S4 adopts the prediction method in the K-nearest neighbor algorithm.

[0034] As a limitation of the present invention, the prediction principle of the K-nearest neighbor algorithm is as follows:

[0035] ① Calculate distance: For a new data point to be predicted, the algorithm calculates the distance between it and every data point in the training set;

[0036] ② Select the nearest neighbors: Based on the calculated distance, select the K nearest data points to the data point to be predicted as its neighbors. Here, K is a pre-set parameter that determines the number of neighbors that affect the prediction result;

[0037] ③ Prediction results: Once the K nearest neighbors are selected, the algorithm will take the average of the target variables of these neighbors, and this average will be used as the prediction result for the data points to be predicted.

[0038] By adopting the above technical solution, the beneficial effects achieved by the present invention compared with the prior art are as follows:

[0039] (1) This invention extracts text-format attribute basic data from the bottom layer data of three-dimensional seismic data. Before the data quality evaluation, the data is grouped, anomaly identification is performed, and anomaly values ​​are reconstructed. This can restore the true situation of data quality in the whole region and improve the accuracy of the evaluation effect by more than 40%, and more objectively and realistically reflect the changes in data quality.

[0040] (2) Considering the impact of spatial location changes on the quality of data, the present invention divides the basic attribute data into multiple neighboring regions according to a specified distance range, and performs data analysis and processing on this basis, which is conducive to fully understanding the true situation of the quality of the three-dimensional data in the whole area.

[0041] (3) The group processing method of the present invention is to perform rolling processing of each group of data in steps, which speeds up the algorithm process, improves the data processing efficiency of the computer, solves hardware limitations, lays a technical foundation for the algorithm to process massive data, and the number of groups can be determined according to the amount of data to be processed and the computer configuration.

[0042] In summary, this invention, through big data algorithms such as grouping, anomaly identification, and reconstruction, suppresses the effects of anomalies caused by spatial location and environmental interference, thereby improving the accuracy of 3D data and its processing efficiency. It is suitable for accurately, efficiently, and comprehensively evaluating the quality of massive amounts of raw 3D data. Attached Figure Description

[0043] Figure 1 This is a flowchart of the method steps in Embodiment 1 of the present invention;

[0044] Figure 2 This is a schematic diagram of the operation interface for calculating the main parameters of energy properties in Embodiment 1 of the present invention;

[0045] Figure 3 This is a schematic diagram of the operation interface for the main parameters of energy attribute output in Embodiment 1 of the present invention;

[0046] Figure 4 This is a flowchart of the mean clustering algorithm in Embodiment 1 of the present invention;

[0047] Figure 5 This is a schematic diagram of the local outlier factor algorithm steps in Embodiment 1 of the present invention;

[0048] Figure 6 This is a comparison chart of the reconstructed RMS values ​​of the single-gun energy attribute (each channel) in Embodiment 1 of the present invention before and after reconstruction;

[0049] Figure 7 This is a comparison chart of the total RMS value (sum of all traces) of the single-gun energy attribute before and after reconstruction in Embodiment 1 of the present invention;

[0050] Figure 8 This is a diagram illustrating the grouping effect in Embodiment 2 of the present invention;

[0051] Figure 9 This is an outlier identification diagram from Embodiment 2 of the present invention;

[0052] Figure 10 This is a data quality evaluation diagram of statically corrected shot gather data before energy attribute processing in Embodiment 2 of the present invention;

[0053] Figure 11 This is a data quality evaluation diagram of statically corrected shot gather data after energy attribute processing in Embodiment 2 of the present invention;

[0054] Figure 12 This is an evaluation diagram of the conventional evaluation method (RTQC graded evaluation of real-time monitoring software) in Embodiment 2 of the present invention. Detailed Implementation

[0055] The preferred embodiments of the present invention will now be described with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustrative and explanatory purposes only and do not constitute a limitation thereof.

[0056] Example 1

[0057] This embodiment discloses a three-dimensional data quality evaluation method based on joint analysis of neighboring shot gather attributes, such as... Figure 1 As shown, the specific steps include:

[0058] S1, Extracting attribute features from 3D shot gather data: Input 3D seismic data into the data processing software system, extract the attribute features of the shot gather data, and output the basic attribute data of each trace of the seismic data in text format.

[0059] In this embodiment, the input data is statically corrected shot gather data. Using the GeoEast processing software system from BGP Inc., the attribute features of the shot gather data are extracted. Shot gather attributes include energy, dominant frequency, signal-to-noise ratio, etc. The extraction methods for each attribute are the same in this step. The specific attribute extraction steps are as follows:

[0060] ① The TakeTraceAttri module of the GeoEast processing system is used to calculate the attribute value of each trace of the seismic data, such as... Figure 2 As shown, the main parameters are:

[0061] Analysis Time Window: Select "rectangle" for the time window type;

[0062] Time window size: Enter 0 for rectangle. 2500 represents the length of the rectangular time window for analyzing 0-2500ms.

[0063] Attribute content: Operations in window. Entering True means analysis, False means no analysis. You can choose to enter attributes such as Energy, Power, RMS, MFrequency, etc., according to your needs.

[0064] ② Use the TakeGatherAttri module of the GeoEast processing system to output the attribute values ​​of each trace of the seismic data, such as... Figure 3 As shown, the main parameters are:

[0065] Output file type: Select text file format for the output option;

[0066] Gather type: Select "user defined" for the gather type;

[0067] First keyword: Select file number (Field file number);

[0068] Second keyword: Select the trace number;

[0069] Attribute type: Select "time window" for attribute type;

[0070] Time window: Select window1 RMS (select the corresponding attribute type according to the attributes analyzed in step ①);

[0071] Time window operation: Select Sum.

[0072] Through the above two steps, attribute values ​​are calculated and output successively, and finally the basic data of each trace of seismic data in text format is obtained, which serves as the basis for subsequent processing.

[0073] S2, grouping the basic attribute data: using a clustering algorithm to divide the basic attribute data obtained in S1 into multiple neighboring regions according to a specified distance range, and obtaining several groups of neighboring region cluster attribute values.

[0074] Purpose of grouping: The amount of attribute data in 3D seismic data is very large (reaching tens to hundreds of GB), and computer memory is limited. In order to improve the efficiency of big data processing, a strategy of grouping and cyclic processing is adopted to overcome hardware limitations. The number of groups depends on the computer's configuration.

[0075] Grouping Method: Most existing data grouping methods simply divide regions according to the x and y coordinates, resulting in uneven distribution of sample sizes in each group. Processing data in groups with insufficient samples leads to inaccurate results. This embodiment employs a clustering algorithm from industrial big data analysis to divide large-scale 3D shot gather attribute data into multiple neighboring regions according to a specified distance range. This adapts to changes in data quality due to spatial and topographical variations, while simultaneously accelerating the algorithm's process and laying a technical foundation for handling massive amounts of data. This embodiment uses the mean clustering algorithm, a fast clustering algorithm that flexibly calculates the distance between sample points, resulting in a relatively uniform number of sample points in each group, which is beneficial for improving the stability of anomaly identification. The mean clustering algorithm divides m sample points in the sample space into n clusters. The existing mean clustering algorithm process is as follows... Figure 4 As shown, the specific steps are as follows:

[0076] ① Randomly select n center points (in this embodiment, it means dividing the basic attribute data into n groups, and the value of n is selected according to the computer configuration and the amount of data to be processed);

[0077] ② Calculate the distance from each sample to each of the n center points;

[0078] ③ Assign the sample points to the clusters closest to the center point;

[0079] ④ Calculate the mean of the sample points in each cluster as the new center point;

[0080] ⑤ Iterate until the results are the same or the termination condition is met.

[0081] S3, outlier identification: The local outlier factor algorithm is used to detect outlier data in several groups of neighboring area shot gather attribute values ​​and identify outlier attribute values ​​in each group of neighboring area shot gather attribute values.

[0082] The Local Outlier Factor (LOF) algorithm can accurately identify outliers in the shot gather properties of neighboring regions. Specifically, it obtains the outlier score of a sample by estimating the separation between each sample and its local neighborhood. If a sample has low local density, its LOF score will be large, and it may be considered an outlier. In this embodiment, the LOF anomaly detection algorithm is based on a density estimation strategy to identify outliers. The algorithm steps are as follows: Figure 3 As shown, the formulas used include:

[0083] Local neighborhood density calculation:

[0084]

[0085] Outlier factor (i.e., relative density) calculation:

[0086]

[0087] Where N(x,k) is the set of k-nearest neighbors that do not contain x, |N(x,k)| is the size of the set, y is a set of nearest neighbors, and distance(x,y) is the reachability distance.

[0088] Then, calculations are performed according to the LOF anomaly detection algorithm implementation steps. Specifically, for the grouped data, the anomaly attribute identification of the entire work area is completed through multi-region iterative processing based on the divided regions.

[0089] S4, Reconstructing outliers and missing values: Using machine learning algorithms, combined with techniques such as linear regression and curve fitting, a curve with multiple statistical data is obtained. By resampling the curve, the reconstructed values ​​of the outlier and missing value locations are obtained, thus completing the reconstruction of the outlier attribute values ​​and channel data, i.e. missing values, obtained in S3.

[0090] In the data obtained from actual testing, in addition to some outliers caused by factors such as the collection environment and region, there are also some missing data points, i.e., missing values. In order to maintain the consistency of the total number of analysis sample points (i.e., the total number of traces) for each shot set, outliers and missing values ​​cannot be directly discarded, as this would affect the accuracy of the data. Therefore, this step requires reconstructing outliers and missing values, using the KNN (K-Nearest Neighbors) algorithm. This is a basic machine learning algorithm, and the prediction method in the KNN algorithm is used to solve regression problems. Its principle is as follows:

[0091] ① Calculate distance: For a new data point to be predicted, the algorithm calculates the distance between it and every data point in the training set.

[0092] ② Select the nearest neighbors: Based on the calculated distance, select the K nearest data points to the data point to be predicted as its neighbors. Here, K is a pre-set parameter that determines the number of neighbors that affect the prediction result.

[0093] ③ Prediction results: Once the K nearest neighbors are selected, the algorithm will take the average of the target variables of these neighbors, and this average will be used as the prediction result for the data points to be predicted.

[0094] By leveraging the similarity of neighboring data, algorithms such as linear regression and curve fitting are used to obtain curves with multi-data statistics. By resampling the curves, reconstructed values ​​for the locations of outliers and missing values ​​are obtained.

[0095] like Figure 6The figure shows the changes in the statistical curves of the RMS values ​​of typical single-shot energy attributes (each shot in the shot set) before and after outlier reconstruction in a set of actual data. The red dots represent outliers, and the blue dots represent the data values ​​after reconstruction of outliers and missing values. After reconstruction, outliers are significantly eliminated, and energy attributes are significantly improved.

[0096] S5, Create a visual data quality evaluation map: Sum the attribute values ​​of all traces in each shot set of all neighboring shot sets after reconstruction to obtain the new attribute values ​​of all sample points, such as... Figure 7 As shown, it displays Figure 6 The image shows a comparison of the total RMS value (sum of all traces) of a typical single-shot energy attribute before and after reconstruction. The red bars represent the sum of RMS values ​​before processing, and the blue bars represent the sum of RMS values ​​after reconstruction. A significant change in values ​​can be seen before and after reconstruction. The reconstructed data is then visualized to create a data quality evaluation map for the entire region, thus completing the quality evaluation of the data across the entire region.

[0097] This embodiment provides a 3D data quality evaluation method based on joint analysis of neighboring shot gather attributes. By extracting text-formatted attribute base data from the underlying 3D seismic data, the method performs grouping, anomaly identification, and reconstruction on the data before quality evaluation, restoring the true quality of the data across the entire region. In one actual processing, a total of 29,815 shots were evaluated, with 12,522 shots showing outliers and requiring reconstruction, representing 42% of the total. This improved the accuracy of the evaluation by more than 40%. The data grouping is based on the influence of spatial location changes, dividing the data into regions within a specified distance range to form multiple neighboring shot gather regions. The grouping and rolling processing method has high computational efficiency, ultimately providing a comprehensive evaluation of the massive amount of raw 3D data.

[0098] Example 2

[0099] Based on Example 1, this example further illustrates the evaluation of energy attributes in a three-dimensional data quality evaluation method based on joint analysis of neighboring shot gather attributes. The data in this example comes from the Gengwan 3D project in the Ordos Basin in 2023.

[0100] S1. Using the GeoEast processing system, extract the energy attribute features (root mean square amplitude of energy) of the static correction shot gather data. Refer to the main parameter operation methods in the energy attribute calculation and energy attribute output in Example 1 to extract the basic data of the shot gather energy attributes in text format.

[0101] S2, grouping the energy attribute data of the gun gathering, such as Figure 8As shown, the shot point coordinates are grouped using the mean clustering algorithm, dividing them into 18 neighboring regions of shot gather attributes according to a specified distance range, resulting in several groups of neighboring region shot gather attribute values. This grouping method is suitable for the irregular shape of the boundary of this type of work area, has high accuracy, and the number of samples in each group is evenly distributed, which is beneficial to improving the accuracy of subsequent anomaly identification algorithms.

[0102] S3, Identify outliers. Calculate the local outlier factor algorithm for each of the shot gather energy attributes in the adjacent areas to identify outlier values ​​in each group of adjacent area shot gather energy attribute values. For example... Figure 9 As shown, this is the outlier identification effect among 16 groups of adjacent area shot gather energy attribute values. Red dots in the scatter plot represent outliers, and blue dots represent normal values. The outlier identification effect is good.

[0103] S4 reconstructs the identified outliers and missing values. Utilizing the similarity of neighboring data, algorithms such as linear regression and curve fitting are used to obtain curves with multi-data statistical properties. By resampling the curves, the reconstructed values ​​of the outlier and missing value locations are obtained.

[0104] S5, forming a visual data quality evaluation map: sum the attribute values ​​of all traces in each shot set of all adjacent shot sets after reconstruction to obtain the new attribute values ​​of all sample points, and form a data quality evaluation map of the whole area through visualization.

[0105] like Figure 10 The image shown is a visualization of the energy attribute data of the statically corrected shot gather data from the Gengwan 3D project in the Ordos Basin in this embodiment, after reconstruction of outliers and missing values. Figure 11 This is a visual data quality evaluation chart generated directly without anomaly data reconstruction. The values ​​corresponding to each colored dot represent the range of energy attribute characteristics (root mean square amplitude, RMS) values. A comparison shows that... Figure 11 The presence of numerous outliers affects the evaluation of results and results in low accuracy. Figure 10 As can be seen from the data, after processing using the method based on joint analysis of neighboring shot gather attributes, Figure 11 The abnormal energy attribute values ​​represented by the anomalous points (larger values ​​in reddish-brown and smaller values ​​in dark blue) can be recovered. Furthermore, a higher RMS value indicates sufficient target layer reflection energy and better data quality in single-shot data, while a lower RMS value indicates weak target layer reflection energy and poorer data quality. Therefore, in Figure 10It can also be seen that yellow and orange (RMS values ​​125-135) are more common in the central and western parts of the region, indicating that the target layer reflection energy of single-shot data is sufficient and the data quality is better. Blue (RMS values ​​110-115) is more common in the central and eastern parts, indicating that the target layer reflection energy of single-shot data is weaker and the quality is poorer. This is related to the reduced amount of explosive charge per shot on both sides of the valley terrain during construction and the significant interference from roads and villages. In summary, the evaluation in this embodiment is comprehensive and effective, and truly reflects the three-dimensional data quality of the Gengwan area in the Ordos Basin.

[0106] To better illustrate the evaluation effect of this embodiment, a conventional standard shot comparison evaluation method was used to evaluate the seismic 3D data in this embodiment. The results are as follows: Figure 12 The image shown is a graded evaluation map obtained using the real-time quality control (RTQC) software. It compares the data from each shot with a standard shot, then grades the data into two levels: Level 1 and Level 2. It can be seen that this method produces poorly consistent results on a planar map, making it difficult to demonstrate variations in the quality of 3D seismic data, especially under different geological conditions, topography, and regional environments. Figure 10 and Figure 11 It is known that the actual energy attribute characteristics (root mean square amplitude of energy, RMS) distribution in this region have regional characteristics, which cannot be reflected in the results of conventional evaluation methods. At the same time, it is difficult to unify the standards for manually selecting standard shots, and human factors cause large errors in the standard shots. Furthermore, outliers in various attribute values ​​in the data of each shot can lead to deviations in the evaluation results, resulting in low reliability of the results of conventional evaluation methods.

[0107] This embodiment provides a three-dimensional data quality evaluation method based on joint analysis of neighboring shot gather attributes. This method evaluates the seismic data quality of the Gengwan three-dimensional project in the Ordos Basin. By extracting text-formatted attribute data from the underlying data, the method performs grouping, anomaly identification, and reconstruction on the data before quality evaluation. This restores the true quality of the data across the entire region and improves accuracy. Furthermore, the grouping is based on the influence of spatial location changes, dividing the region into multiple neighboring shot gather areas according to a specified distance range. The grouping and rolling processing method has high computational efficiency, ultimately resulting in a comprehensive evaluation of the original seismic data of the Gengwan area in the Ordos Basin.

Claims

1. A method for evaluating the quality of three-dimensional data based on joint analysis of neighboring shot gather attributes, characterized in that, Includes the following steps: S1, Extract the attribute features of the 3D shot gather data: Input the 3D seismic data into the data processing software system, extract the attribute features of the shot gather data, and output the basic attribute data of each trace of the seismic data in text format. S2, grouping the basic attribute data: using a clustering algorithm to divide the basic attribute data obtained in S1 into multiple neighboring regions according to a specified distance range, and obtaining several groups of neighboring region cluster attribute values; S3, Identify outliers: The local outlier factor algorithm is used to detect outliers in each group of neighboring shot gather attribute values ​​and identify outlier attribute values ​​in each group of neighboring shot gather attribute values. S4, Reconstructing outliers and missing values: Machine learning algorithms are used to process the normal attribute values ​​in each group to obtain curves with multiple data statistics. By resampling the curves, the reconstructed values ​​of the outlier and missing value positions are obtained, thus completing the reconstruction of the outlier attribute values ​​and missing data in each group obtained in S3. S5. A visual data quality evaluation map is generated. The attribute values ​​of all traces in each shot set of all adjacent shot sets after reconstruction are summed to obtain the new attribute values ​​of all sample points. The data quality evaluation map of the whole area is generated through visualization.

2. The method for evaluating the quality of three-dimensional data based on joint analysis of neighboring shot gather attributes as described in claim 1, characterized in that: The three-dimensional seismic data in S1 is statically corrected shot gather data.

3. The method for evaluating the quality of three-dimensional data based on joint analysis of neighboring shot gather attributes as described in claim 2, characterized in that: The attribute characteristics of the shot collection data in S1 include energy, dominant frequency, and signal-to-noise ratio.

4. The method for evaluating the quality of three-dimensional data based on joint analysis of neighboring shot gather attributes as described in claim 1, characterized in that: The process of extracting basic attribute data using a data processing software system in S1 is as follows: ① Use the seismic trace attribute extraction module of the data processing software system to calculate the attribute value of each trace of the seismic data; ② Use the seismic gather attribute output module of the data processing software system to output the attribute values ​​of each trace of the seismic data, that is, to obtain the basic attribute data of each trace of the seismic data in text format.

5. The method for evaluating the quality of three-dimensional data based on joint analysis of neighboring shot gather attributes according to claim 1, characterized in that: The clustering algorithm in S2 is the mean clustering algorithm, and the specific steps are as follows: ① Randomly select n center points, that is, divide the basic attribute data into n groups according to the specified distance range. The value of n is selected according to the regional terrain, computer configuration, and the size of the data to be processed. ② Calculate the distance from each sample to each of the n center points; ③ Assign the sample points to the clusters closest to the center point; ④ Calculate the mean of the sample points in each cluster as the new center point; ⑤ Iterate until the results are the same or the termination condition is met, and complete the data grouping.

6. The method for evaluating the quality of three-dimensional data based on joint analysis of neighboring shot gather attributes according to claim 1, characterized in that: The local outlier factor algorithm in S3 identifies outliers based on a density estimation strategy. The outliers obtained through the density-based outlier diagnosis step are the abnormal attribute values ​​in the shot gather attribute values ​​of each group of neighboring regions.

7. The method for evaluating the quality of three-dimensional data based on joint analysis of neighboring shot gather attributes as described in claim 6, characterized in that: The formula used in the local outlier factor algorithm includes: Local neighborhood density calculation: Outlier factor (i.e., relative density) calculation: Where N(x,k) is the set of k-nearest neighbors that do not contain x, |N(x,k)| is the size of the set, y is a set of nearest neighbors, and distance(x,y) is the reachability distance.

8. The method for evaluating the quality of three-dimensional data based on joint analysis of neighboring shot gather attributes according to claim 1, characterized in that: The machine learning algorithm in S4 uses the prediction method of the K-nearest neighbor algorithm.

9. A method for evaluating the quality of three-dimensional data based on joint analysis of neighboring shot gather attributes as described in claim 8, characterized in that: The prediction principle of the K-nearest neighbor algorithm is as follows: ① Calculate distance: For a new data point to be predicted, the algorithm calculates the distance between it and every data point in the training set; ② Select the nearest neighbors: Based on the calculated distance, select the K data points that are closest to the data point to be predicted as its neighbors. Here, K is a pre-set parameter that determines the number of neighbors that affect the prediction result. ③ Prediction results: Once the K nearest neighbors are selected, the algorithm will take the average of the target variables of these neighbors, and this average will be used as the prediction result for the data points to be predicted.

Images