A method and system for thin interbedded salt cavern background injection recovery calibration

Abstract

The application discloses a kind of thin interbedded salt cavity background injection recovery calibration method and system, mainly related to underground salt cavity identification technical field.It includes: reconstructing measured seismic data volume as measured background and calculating local data quality field;Constructing parameterized salt cavity hypothesis library with known boundary true value, generating differential response composed of reflection response and scattering response and calibrated by background amplitude;Avoid the location of existing anomaly body by injecting differential response into background, to obtain recovery volume with the same detector as actual detection, and obtain error by comparing with true value;Calibration model is fitted with the resolution characteristics that recovery volume can obtain;Calculate the resolution characteristics of actual candidate anomaly body, correct boundary according to model prediction deviation, and output volume interval and calibration shell.The beneficial effects of the application are that it realizes the quantitative correction of salt cavity boundary and volume and the quantitative output of uncertainty.

Description

Technical Field

[0001] This invention relates to the field of underground salt cavity identification technology, specifically a method and system for background injection recovery calibration of thin interlayered salt cavities. Background Technology

[0002] Thin-layered salt rock deposits are often composed of frequent interbedded layers of salt rock, mudstone, and anhydrite mudstone. After long-term water-soluble mining, the salt cavities exhibit complex morphologies such as irregular tops and bottoms, near-vertical sidewalls, multiple cavities connected in series, and internal residual interlayers. Abandoned or renovated salt cavities can be used for compressed air energy storage, natural gas storage, or hydrogen storage. Before engineering utilization, it is essential to identify the location, boundaries, connectivity, and volume of the cavities, and explain the reliability of the interpretation results.

[0003] Existing salt cavity identification methods mostly start from attributes such as root mean square amplitude, variance, or dissimilarity, and obtain anomalies through single threshold or multi-attribute fusion. Downhole sonar is limited by the number of wells and the accessibility of detection. Although 3D seismic has the advantage of regional coverage, the cavity response is affected by thin-layer tuning, diffraction scattering, limited frequency band, offset illumination and acquisition footprint. The outer boundary of seismic anomalies is usually not equal to the actual salt cavity boundary.

[0004] However, existing methods struggle to answer the question of what scale and response strength of salt cavities can be reliably detected under specific data quality conditions, and cannot distinguish between detector-inherent expansion, contraction, or fragmentation and the true geological boundary. Existing threshold perturbations only reflect the sensitivity of algorithm parameters and cannot directly measure the systematic error relative to the true boundary, nor can they generate the inner and outer calibration shells and volume ranges of candidate bodies. There is also a lack of closed-loop calibration methods that use measured seismic subbodies as the injection background, the known true values ​​of parameterized salt cavities as the calibration benchmark, and ensure that the injection target and the actual data use the same detector.

[0005] Therefore, there is an urgent need for a thin interlayered salt cavity background injection recovery calibration method and system to solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for the background injection recovery calibration of thin interlayered salt cavity, which realizes quantitative correction of salt cavity boundary and volume and quantitative output of uncertainty.

[0007] To achieve the above objectives, the present invention employs the following technical solution: On one hand, the present invention provides a method for restoring calibration of thin interlayered salt cavity background injection, comprising the following steps: Step S1: Reconstruct the measured 3D seismic data volume as the measured background, and limit the target time window, and calculate the local data quality field within the target time window; Step S2: Construct a parameterized salt cavity hypothesis library with known boundary truth values, generate a differential response consisting of reflection and scattering responses based on the geometric and contrast parameters of each salt cavity hypothesis, and calibrate the differential response according to the amplitude level of the measured background in Step S1. Step S3: Inject the calibrated differential response obtained in step S2 into the measured background of step S1 to avoid the location of existing anomalies. Use the same detector as the actual detection to recover the injected data to obtain the recovered volume. Compare the recovered volume with the known boundary true value of the hypothesis library to obtain the detection mark, boundary deviation and volume deviation. Step S4: Using the resolution features obtained from the recovery body in step S3 as independent variables and the detection markers, boundary deviations, and volume deviations in step S3 as dependent variables, a calibration model is fitted. Step S5: Use the same detector as in step S3 to detect the actual candidate anomaly and calculate the resolution features. When the resolution features fall within the applicable domain of the calibration model obtained in step S4, predict the boundary deviation and volume deviation according to the calibration model, perform corrosion or expansion correction on the candidate body boundary according to the predicted deviation sign, and output the volume range and inner and outer calibration shells based on the recovery error statistics.

[0008] Preferably, in step S1, the local data quality field is composed of signal-to-noise ratio, lateral continuity, effective bandwidth proxy, and effective channel coverage after robust quantile normalization.

[0009] Preferably, the method for generating and calibrating the differential response in step S2 is as follows: Differential response: ; in, The standard deviation is equal to Gaussian transverse smoothing operator, Configured in terms of track spacing and horizontal resolution of the work area; Main frequency equals The Reckon wave, Determined based on the effective bandwidth of measured seismic data; For the salt cavity hypothesis The effective contrast is determined based on the difference in acoustic impedance between the salt cavity hypothesis and the surrounding rock. For vertical derivative; For the scattering response, its weights Controlled by interlayer structural parameters and satisfying ; To iterate through the hypothesis pool for each salt cavity hypothesis; Calibration coefficient: ; in, The target signal-to-noise ratio is to be injected and is a configurable parameter; The root mean square amplitude of the background measured at the injection location; It is the minimum background scale threshold and is a configurable parameter used to suppress distortion calibration in the low amplitude region; It is a small positive number, used to ensure numerical stability; Injected data: ; in, For the measured background data at the injection location, This is the data that the restorer will use after injection.

[0010] Preferably, in step S3, the boundary deviation and volume deviation are characterized as follows: ; ; in, This is the equivalent radius deviation. , These are the equivalent diameters of the restored volume and the known boundary truth values, respectively. For logarithmic volume deviation, , These are the volumes of the restored volume and the known boundary truth values, respectively. It is a small positive number; the detection indicator is a binary result of whether the restorer has reached the detection threshold.

[0011] Preferably, the fitting method for the calibration model in step S4 is as follows: Distinguish feature vectors: ; in, The local mass at the location of the candidate object; For local signal-to-noise ratio; The aspect ratio for the detection of the restored body; , The equivalent diameter and height of the restored body were determined. , , The sampling interval between the main survey line, connecting lines, and depth direction; The amplitude intensity of the candidate object relative to the background is determined by the ratio of the candidate object amplitude to the root mean square amplitude of the background; all of the above components can be directly obtained from the detected object and the mass field; Calibration model: ; in, This is a sample matrix with the distinguishable feature vectors as rows. To detect the dependent variable matrix composed of the mark, boundary deviation, and volume deviation, This is a regularization parameter and it is a configurable parameter. It is the identity matrix. This is the regression coefficient matrix; Predicted detection probability Predicting boundary deviation With volume deviation It is obtained by linear combination of the corresponding regression coefficient vector and the distinguishing feature vector.

[0012] Preferably, the calibration model is applied in step S5 as follows: Boundary correction: ; in, The candidate obtained from the detection; , These are the boundary deviation and volume deviation predicted by the calibration model, respectively; This is a morphological operator, with the subscript value being the scale parameter. A positive subscript performs dilation, and a negative subscript performs erosion. Inner and outer calibration shells: ; in, For erosion calculation, For expansion operations, The envelope distance is determined based on the larger of the boundary deviation residual quantile and the boundary distance quantile of the detected cases. Central volume value: ; Lower limit of volume: ; in, To detect volume, , These are the 90th and 10th percentiles of the volume deviation residuals, respectively.

[0013] Preferably, in step S5, the applicable domain of the calibration model is determined by the parameter range of the parameterized salt cavity hypothesis library; The calibration support classifies candidates into three levels: resolvable, critically resolvable, and indistinguishable, based on the preset resolvable threshold and critical resolvable threshold. For candidate objects whose resolvable features exceed the applicable domain or whose calibration support is below the indistinguishable threshold, no closed-correction entity is generated; only a conservative volume range and outer envelope starting from zero are output.

[0014] Preferably, in steps S3 and S5, the detectors that are the same as those used in the actual detection employ fixed attribute normalization upper and lower limits and detection thresholds, and apply the same detection and three-dimensional connected body labeling operations to the injected restorers and the actual candidate anomalous bodies.

[0015] On the other hand, the present invention also provides a thin interlayered salt cavity background injection recovery calibration system for implementing the method described above, comprising: The data reading and quality assessment module is used to reconstruct the measured seismic data volume and calculate the local data quality field; The hypothesis library and differential response module are used to construct a parameterized salt cavity hypothesis library with known boundary truth values ​​and generate a differential response consisting of reflection and scattering responses calibrated according to the measured background amplitude. The injection recovery and error evaluation module is used to inject the differential response into the measured background to avoid the location of the existing anomalous body, and to recover the recovered body with the same detector as the actual detection. The recovered body is then compared with the known boundary true value to obtain the detection mark, boundary deviation and volume deviation. The calibration model module is used to fit a calibration model with the resolvable features available in the restorer as independent variables; The candidate body correction and output module is used to calculate the distinguishing features of the actual candidate anomalies, correct the boundary according to the predicted deviation sign when they fall into the applicable domain of the calibration model, and output the volume range and inner and outer calibration shells based on the recovery error statistics.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention uses the reconstructed measured seismic data volume as the measured background and calculates the local data quality field, preserving the measured noise and spatial nonstationarity in the calibration benchmark, so that the injection and recovery are carried out under the conditions of real data, avoiding calibration distortion caused by the disconnect between synthetic data and measured data.

[0017] 2. This invention constructs a parameterized salt cavity hypothesis library with known boundary truth values, generates a differential response consisting of reflection and scattering responses and calibrated according to the background amplitude, so that the injector matches the measured background in terms of amplitude level and frequency band characteristics, thereby improving the approximation of the calibration benchmark to the complex response of thin interlayered salt cavities.

[0018] 3. This invention avoids the location of existing anomalous bodies by injecting differential response into the measured background, recovers the recovered body using the same detector as the actual detection, and compares it with the known boundary true value. This eliminates the systematic interference of detector expansion, contraction, or fragmentation on the calibration results, and directly measures the systematic error relative to the true boundary.

[0019] 4. This invention fits a calibration model with the resolvable features of the restored object as independent variables, and incorporates available quantities such as data quality, detection scale, and amplitude intensity of the candidate object relative to the background into a unified regression framework, thus solving the problem that forward prediction is impossible because the true value of the actual candidate object is not available.

[0020] 5. This invention corrects the corrosion or expansion of candidate body boundaries based on the predicted deviation sign, and outputs the volume range and inner and outer calibration shells based on the recovery error statistics, thereby achieving quantitative correction and uncertainty quantification output of salt cavity boundaries and volumes. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the system structure of the present invention; Figure 3 This is a schematic diagram of the local data quality field of the measured earthquake background of the present invention; Figure 4 This is a schematic diagram of the parameterized known boundary salt cavity hypothesis library of the present invention; Figure 5 This is a schematic diagram illustrating an example of synthetic differential response injection and recovery using the same algorithm of the present invention; Figure 6 This is a schematic diagram illustrating the relationship between detectability and boundary deviation calibration in this invention; Figure 7 This is a schematic diagram of the actual SEG-Y candidate anomaly plane recognition results of the present invention; Figure 8 This is a schematic diagram of the boundary correction candidate and the outer envelope seismic profile of the present invention; Figure 9 This is a three-dimensional schematic diagram of the calibration salt chamber and inner and outer calibration shells of the present invention; Figure 10 This is a schematic diagram summarizing the injection recovery calibration results of the zhuanli.sgy invention. Detailed Implementation

[0022] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined in this application.

[0023] In this invention, terms such as "upper," "lower," "left," "right," "front," "back," "vertical," "horizontal," "side," and "bottom" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used only to facilitate the description of the structural relationships of the various components or elements of this invention and do not specifically refer to any component or element in this invention. They should not be construed as limiting the invention.

[0024] Example 1: like Figure 1 As shown, this embodiment provides a method for restoring calibration of thin interlayered salt cavity background injection, including the following steps: Step S1: Reconstruct the measured 3D seismic data volume as the measured background, and limit the target time window, and calculate the local data quality field within the target time window; This embodiment uses a SEG-Y data file as input. The data volume contains 817 time samples, 139 main survey lines, and 87 connecting lines, totaling 12093 channels. The sampling interval is 1.0 ms. The acquired metadata includes a receiving channel spacing of 20m, a line spacing of 30m, and an firing distance of 10m. The regular voxel plane spacing calculated from the coordinates of the common center point of the channel head is 5m×5m. The salt cavity interpretation time window is limited to 400ms to 700ms. The time depth conversion example uses an average velocity of 4500m / s, corresponding to a vertical voxel thickness of 2.25m. Specifically, the system reads 3D seismic data that has undergone amplitude fidelity processing, static correction, velocity analysis, stacking, and migration imaging. It reconstructs a regular grid based on binary roll headers and trace headers, records missing traces with masks, and calculates the local data quality field within the target time window. This quality field is composed of signal-to-noise ratio, lateral continuity, effective bandwidth proxy, and effective trace coverage after robust quantile normalization. The local data quality field is calculated as follows: at each planar location... Construct mass vector Their components represent the local signal-to-noise ratio. Horizontal continuity Effective bandwidth proxy volume and effective road coverage Each component is robustly normalized to its quantiles within the target time window: ; in, , , and All parameters are robustly normalized to quantiles within the target time window. The mass field participates in both the hierarchical selection of subsequent injection locations and serves as an input feature for the calibration model, such as... Figure 3 As shown, the mass field varies with spatial location, reflecting the non-stationarity of noise, collected footprints, missing traces, and local imaging quality in the measured background.

[0025] Step S2: Construct a parameterized salt cavity hypothesis library with known boundary truth values, generate a differential response consisting of reflection and scattering responses based on the geometric and contrast parameters of each salt cavity hypothesis, and calibrate the differential response according to the amplitude level of the measured background in Step S1. A parameterized salt cavity hypothesis library was established based on horizontal diameter, height, aspect ratio, residual interlayer ratio, and response intensity factor. The injection hypothesis library in this embodiment covers 18 injection cases with diameters ranging from 50m to 120m, heights from 35m to 80m, residual interlayer ratios from 0 to 0.40, and response intensity factors from 1.5 to 5.0. Figure 4 As shown, each hypothesis contains a three-dimensional truth mask, and the measured background sub-body is selected at the position of different local data quality scores; The differential response is generated as follows: Let the first... The truth mask for the parameterized salt cavity hypothesis is: Effective contrast is achieved by the truth mask and the proportion of residual mezzanine layers. ; Divide along the time sampling direction and sum with the dominant frequency. Zero-phase Reichsson wavelet Convolution, applying a lateral Gaussian blur to the convolution result. And superimposed a deterministic scattering response controlled by the proportion of residual interlayers. The differential response is obtained. It consists of reflection response and scattering response, where the reflection response... Reflection characteristics and scattering response of the top and bottom interfaces of the salt cavity Corresponding to diffraction scattering within thin interlayers: ; in, The standard deviation is equal to Gaussian transverse smoothing operator, The configuration is based on track spacing and horizontal resolution of the work area. In this embodiment, we take... Double track spacing; Main frequency equals The Reckon wave, The effective bandwidth of the measured seismic data is determined in this embodiment. Hz; For the salt cavity hypothesis The effective contrast is determined based on the difference in acoustic impedance between the salt cavity hypothesis and the surrounding rock. For vertical derivative; For the scattering response, its weights Controlled by interlayer structural parameters and satisfying In this embodiment, ; Subscript Iterate through each salt cavity hypothesis in the hypothesis library; The peak value of the differential response is calibrated using the root mean square amplitude of the background sub-body measured at the injection location, with the scaling factor being: ; in, To inject the target signal-to-noise ratio, which is a configurable parameter, the value in this embodiment ranges from 1.5 to 5.0; The root mean square amplitude of the background measured at the injection location; The minimum background scale threshold is a configurable parameter; in this embodiment, it is set to... Used to suppress distortion calibration in the low amplitude region; For small positive numbers, take This is used to ensure numerical stability; The calibrated differential response is superimposed onto the corresponding measured background sub-body to obtain the injected data: ; in, For the measured background data at the injection location, The injection background is taken directly from the measured data to provide data for the restorer after injection. Therefore, the non-stationary noise, acquisition footprint, residual static correction error and local imaging differences of the sub-body are retained.

[0026] Step S3: Inject the calibrated differential response obtained in step S2 into the measured background of step S1 to avoid the location of existing anomalies. Use the same detector as the actual detection to recover the injected data to obtain the recovered volume. Compare the recovered volume with the known boundary true value of the hypothesis library to obtain the detection mark, boundary deviation and volume deviation. The calibrated differential response is injected into the measured background to avoid the location of existing anomalous bodies. The injected data is then reconstructed using the same detector as the actual detection data to obtain the recovered volume. Figure 5 As shown, an example of synthetic differential response injection and recovery using the same algorithm is presented. In this embodiment, the base detector employs a detection method that fuses root mean square amplitude and dissimilarity, using the same program, attribute definition, robust normalization boundary, and detection parameters for both the measured and injected data. The base detector can also be replaced with coherence, structure tensor, machine learning segmentation, or manual interactive segmentation, as long as the injected and actual data use the same algorithm. From the post-injection detection results, the recovered connected components that intersect with the neighborhood of the ground truth mask are selected. The recovered components are compared with the known boundary ground truth. The main evaluation metrics for the recovered components and the ground truth components include the Dess coefficient. Intersection and Union 95% boundary distance Equivalent radius deviation and logarithmic volume deviation : Dess coefficient: ; Intersection and Union: ; 95% boundary distance: ; in, and These are the restored volume and the known truth volume, respectively. and Each represents its boundary. This represents the minimum Euclidean distance between boundary points. This indicates that the 95th percentile is used to suppress a small number of outlier boundary points; Boundary deviation is expressed as equivalent radius deviation. Characterization, volume deviation is represented by logarithmic volume deviation. Characterization: Equivalent radius deviation: ; Logarithmic volume deviation: ; in, and These are the equivalent diameters of the restored volume and the known boundary truth values, respectively. and These are the volumes of the restored volume and the known boundary truth value, respectively. A small positive number indicates whether the detection threshold has been reached (binary result). Figure 6 As shown, the relationship between detectability and boundary deviation calibration is illustrated. As shown in Table 1, among the 18 injection cases in this embodiment, 9 reached the preset detection threshold, accounting for 50.0%, with an average Dyss coefficient of 0.1999 and a maximum Dyss coefficient of 0.6300. This result only represents the program operation results under the current data, hypothesis library, basic detector, and threshold settings, and should not be taken as a general statistical conclusion or as the accuracy rate of on-site salt cavity identification.

[0027] Table 1. Display of SEG-Y Data File Injection-Recovery Calibration Implementation Parameters and Results

[0028] Step S4: Using the resolution features obtained from the recovery body in step S3 as independent variables and the detection markers, boundary deviations, and volume deviations in step S3 as dependent variables, a calibration model is fitted. Using the recognizable features of the recovered body as independent variables, the recognizable feature vector Including local quality Local signal-to-noise ratio Detecting aspect ratio Detection of equivalent diameter With height The ratio relative to the sampling interval, and the amplitude intensity of the candidate relative to the background. Each of the above components can be directly obtained from the detector and the mass field, without relying on the true value, thus solving the problem that forward prediction is impossible because the true value of the actual candidate is unavailable. ; in, , , In this embodiment, the sampling interval for the main survey line, connecting lines, and depth direction is... m, m; The ratio of the candidate body amplitude to the root mean square amplitude of the background is used to determine the above components, which can all be directly obtained from the detected body and the mass field. The calibration model uses ridge regression with intercept to fit the detection markers, equivalent radius bias, and log volume bias, respectively. ; in, This is a sample matrix with the distinguishable feature vectors as rows. To detect the dependent variable matrix composed of the mark, boundary deviation, and volume deviation, This is a regularization parameter and it is configurable. In this embodiment, we take... Used to suppress overfitting in small sample sizes. It is the identity matrix. This is the regression coefficient matrix; The regression output of the detection marker is mapped to 0 to 1 using a constraint function and serves as the calibration support. This value is not a statistical confidence probability. ; in, To predict the detection probability, i.e., to calibrate the support; is the transpose of the regression coefficient matrix, and x is the resolving eigenvector; This indicates that the linear regression output is truncated to the [0,1] interval, predicting boundary bias. With volume deviation It is obtained by linear combination of the corresponding regression coefficient vector and the distinguishing feature vector.

[0029] Step S5: Use the same detector as in step S3 to detect the actual candidate anomaly and calculate the resolution features. When the resolution features fall within the applicable domain of the calibration model obtained in step S4, predict the boundary deviation and volume deviation according to the calibration model, perform corrosion or expansion correction on the candidate body boundary according to the predicted deviation sign, and output the volume range and inner and outer calibration shells based on the recovery error statistics. The same detector as in step S3 was used to detect the actual candidate anomalies. In this embodiment, a total of 11 actual candidate anomalies were identified, such as... Figure 7 As shown, the actual SEG-Y candidate anomaly plane identification results are displayed. For each candidate, the resolution features are calculated. When the resolution features fall within the applicable domain of the calibration model, the boundary deviation and volume deviation are predicted based on the calibration model. Corrosion or expansion correction is applied to the candidate body boundary based on the sign of the predicted boundary deviation: a positive prediction deviation results in corrosion, while a negative deviation results in expansion. The correction is scaled on the physical distance and is uniformly denoted as: ; in, The candidate obtained from the detection; , These are the boundary deviation and volume deviation predicted by the calibration model, respectively; This is a morphological operator, where the subscript value is the scale parameter; a positive subscript performs dilation, and a negative subscript performs erosion. The subscript values ​​in the formula are... ,Right now A positive value indicates erosion of the candidate body, while a negative value indicates expansion; morphological operation distances are expressed in meters and converted according to voxel dimensions in the directions of time, principal survey line, and connecting line, such as... Figure 8 As shown, the boundary correction candidate volume and the outer envelope seismic profile are displayed; The envelope distance is determined by the larger of the absolute value of the residual quantile of the equivalent radius deviation and the 90th quantile of the 95th percentile of the boundary distance of detected cases. Construct an inner calibration shell and an outer calibration shell: ; in, For erosion calculation, For expansion operations, The envelope distance is determined based on the larger of the boundary deviation residual quantile and the boundary distance quantile of the detected cases, such as... Figure 9 The figure shows the three-dimensional results of the calibration salt chamber and the inner and outer calibration shells; The volume center value and volume interval are determined by the predicted logarithmic volume deviation and its residual quantiles. The volume center value is: ; The lower and upper limits of the volume are: ; in, To detect volume, , These are the 90th and 10th quantiles of the volume deviation residuals, respectively. In this embodiment, candidates with a calibration support of not less than 0.55 are marked as resolvable, candidates with a support of 0.30 to 0.55 are marked as critically resolvable, and candidates with a support of less than 0.30 are marked as indistinguishable. All three limits are configurable parameters and can be recalibrated based on well, sonar, extended seismic, or other verification data in the work area. As shown in Table 2, the original volume of candidate 1 is 299,194 m³, the calibration center volume is approximately 239,744 m³, the volume range is 185,169 m³ to 323,394 m³, and the calibration support is 0.463, which is critically resolvable. The calibration support of candidates 3, 8, 9, and 10 is below 0.30, and the program does not provide closed-loop correction volumes for them, but only outputs conservative volume ranges and envelopes starting from 0. All volumes are converted based on an average velocity of 4500 m / s as an example and cannot replace the field velocity model; candidate anomalies cannot replace drilling, logging, sonar, or production dynamics verification; Figure 10 As shown, this provides an overview of the SEG-Y data file injection recovery calibration results.

[0030] Table 2. Calibration results of actual candidate anomalies in SEG-Y data files.

[0031] like Figure 2 As shown, this embodiment provides a thin interlayered salt cavity background injection recovery calibration system, including the following modules: The data reading and quality assessment module is used to reconstruct the measured seismic data volume and calculate the local data quality field. This module reads 3D seismic data that has undergone amplitude fidelity processing, static correction, velocity analysis, stacking, and migration imaging. It reconstructs a regular grid based on binary roll headers and trace headers, records missing traces with masks, and calculates a quality vector within the target time window, which is composed of signal-to-noise ratio, lateral continuity, effective bandwidth proxy, and effective trace coverage after robust quantile normalization.

[0032] The hypothesis library and differential response module are used to construct a parameterized salt cavity hypothesis library with known boundary truth values. It generates a differential response consisting of reflection and scattering responses and calibrated according to the measured background amplitude. This module establishes the hypothesis library according to the horizontal diameter, height, aspect ratio, residual interlayer ratio and response intensity factor. It forms an effective contrast by a three-dimensional truth mask, subtracts along the time sampling direction and convolves with the zero-phase Ricker wavelet, applies transverse Gaussian blur and superimposes the scattering response to obtain the differential response, and superimposes it onto the measured background subbody according to the calibration coefficients.

[0033] The injection recovery and error evaluation module is used to inject the differential response into the measured background at a location that avoids existing anomalous bodies. The recovered body is obtained using the same detector as the actual detection, and the recovered body is compared with the known boundary true value to obtain the detection mark, boundary deviation, and volume deviation. This module adopts a detection method that combines root mean square amplitude and dissimilarity. The same program, attribute definition, robust normalization boundary, and detection parameters are used for the injected data and the actual data. The module also calculates the Dess coefficient, cross-union ratio, 95% boundary distance, equivalent radius deviation, and logarithmic volume deviation.

[0034] The calibration model module is used to fit a calibration model with the resolvable features available in the recovery body as independent variables. This module uses the resolvable feature vector as independent variables and employs ridge regression with intercept to fit the detection mark, equivalent radius bias and log volume bias respectively, and outputs calibration support, predicted boundary bias and predicted volume bias.

[0035] The candidate body correction and output module is used to calculate the resolving features of the actual candidate anomalous bodies. When they fall into the applicable domain of the calibration model, the boundary is corrected according to the sign of the predicted deviation. Based on the recovery error statistics, the volume range and inner and outer calibration shells are output. This module performs morphological operations according to the sign of the predicted boundary deviation, constructs the inner and outer calibration shells, and outputs the volume center value and volume range.

[0036] The system in this embodiment is written in MATLAB R2025b, with the main entry points being run_segy_data.m and run_ceshi_sgy.m, and the core workflow being injectionRecoveryPipeline.m. The program does not rely on third-party SEG-Y toolboxes and can read IEEE 32-bit floating-point data in format code 5 of SEG-Y Rev.1. The program also provides a graphical entry point in launch_app.m, a self-contained example in run_demo.m, and automatic testing in tests.run_tests. If the Image ProcessingToolbox exists, an acceleration function can be called; otherwise, the built-in breadth-first search is automatically used to complete the 3D connected component labeling.

[0037] Example 2: Based on the above embodiments, this embodiment also provides the following alternative implementation methods: (1) The parameterized hypothesis library can use different discrete combinations within the range of horizontal diameter, height, major-minor axis ratio, residual interlayer ratio and response intensity factor already implemented in the program; after changing the combination, all injection recovery cases should be rerun.

[0038] (2) The basic detector can be replaced on the premise that the background before injection, the daughter body after injection and the actual data share the same program, normalization upper and lower limits and threshold parameters; the calibration coefficient of the original detector shall not be used after replacement.

[0039] (3) The regularization coefficient λ, detection threshold, indistinguishable threshold and residual quantile of the ridge regression with intercept can be configured, but the model should be refitted and the configuration recorded each time it is changed.

[0040] (4) Input can be SEG-Y data or MAT three-dimensional arrays that are already supported by the program; missing traces are preserved by the effective trace mask and are not directly equivalent to effective observations with zero values.

[0041] (5) Vertical coordinates can be kept in two-way time, or converted to depth when providing average velocity, layer velocity or velocity model; the velocity used for volume conversion must be recorded along with the results.

[0042] (6) Results can be output in the MAT, CSV, OBJ and PNG formats that the program has implemented. Among them, OBJ only outputs the correction entity or calibration shell generated by the program, and does not indicate that the geological authenticity has been externally verified.

[0043] The above is a detailed description of the preferred embodiments of the present invention, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for recovering and calibrating the background of a thin interlayered salt cavity through injection, characterized in that, Includes the following steps: Step S1: Reconstruct the measured 3D seismic data volume as the measured background, and limit the target time window, and calculate the local data quality field within the target time window; Step S2: Construct a parameterized salt cavity hypothesis library with known boundary truth values, generate a differential response consisting of reflection and scattering responses based on the geometric and contrast parameters of each salt cavity hypothesis, and calibrate the differential response according to the amplitude level of the measured background in Step S1. Step S3: Inject the calibrated differential response obtained in step S2 into the measured background of step S1 to avoid the location of existing anomalies. Use the same detector as the actual detection to recover the injected data to obtain the recovered volume. Compare the recovered volume with the known boundary true value of the hypothesis library to obtain the detection mark, boundary deviation and volume deviation. Step S4: Using the resolution features obtained from the recovery body in step S3 as independent variables and the detection markers, boundary deviations, and volume deviations in step S3 as dependent variables, a calibration model is fitted. Step S5: Use the same detector as in step S3 to detect the actual candidate anomaly and calculate the resolution features. When the resolution features fall within the applicable domain of the calibration model obtained in step S4, predict the boundary deviation and volume deviation according to the calibration model, perform corrosion or expansion correction on the candidate body boundary according to the predicted deviation sign, and output the volume range and inner and outer calibration shells based on the recovery error statistics.

2. The method according to claim 1, characterized in that, In step S1, the local data quality field is composed of signal-to-noise ratio, lateral continuity, effective bandwidth proxy, and effective channel coverage after robust quantile normalization.

3. The method according to claim 1, characterized in that, The method for generating and calibrating the differential response in step S2 is as follows: Differential response: ; in, The standard deviation is equal to Gaussian transverse smoothing operator, Configured in terms of track spacing and horizontal resolution of the work area; Main frequency equals The Reckon wave, Determined based on the effective bandwidth of measured seismic data; For the salt cavity hypothesis The effective contrast is determined based on the difference in acoustic impedance between the salt cavity hypothesis and the surrounding rock. For vertical derivative; For the scattering response, its weights Controlled by interlayer structural parameters and satisfying ; To iterate through the hypothesis pool for each salt cavity hypothesis; Calibration coefficient: ; in, The target signal-to-noise ratio is to be injected and is a configurable parameter; The root mean square amplitude of the background measured at the injection location; It is the minimum background scale threshold and is a configurable parameter used to suppress distortion calibration in the low amplitude region; It is a small positive number, used to ensure numerical stability; Injected data: ; in, For the measured background data at the injection location, This is the data that the restorer will use after injection.

4. The method according to claim 1, characterized in that, In step S3, the boundary deviation and volume deviation are characterized as follows: ； ； in, This is the equivalent radius deviation. , These are the equivalent diameters of the restored volume and the known boundary truth values, respectively. For logarithmic volume deviation, , These are the volumes of the restored volume and the known boundary truth values, respectively. It is a small positive number; the detection indicator is a binary result of whether the restorer has reached the detection threshold.

5. The method according to claim 1, characterized in that, The fitting method for the calibration model in step S4 is as follows: Distinguish feature vectors: ; in, The local mass at the location of the candidate object; For local signal-to-noise ratio; The aspect ratio for the detection of the restored body; , The equivalent diameter and height of the restored body were determined. , , The sampling interval between the main survey line, connecting lines, and depth direction; The amplitude intensity of the candidate object relative to the background is determined by the ratio of the candidate object amplitude to the root mean square amplitude of the background; all of the above components can be directly obtained from the detected object and the mass field; Calibration model: ; in, This is a sample matrix with the distinguishable feature vectors as rows. To detect the dependent variable matrix composed of the mark, boundary deviation, and volume deviation, This is a regularization parameter and it is a configurable parameter. It is the identity matrix. This is the regression coefficient matrix; Predicted detection probability Predicting boundary deviation With volume deviation It is obtained by linear combination of the corresponding regression coefficient vector and the distinguishing feature vector.

6. The method according to claim 1, characterized in that, The application method of the calibration model in step S5 is as follows: Boundary correction: ; in, The candidate obtained from the detection; , These are the boundary deviation and volume deviation predicted by the calibration model, respectively; This is a morphological operator, with the subscript value being the scale parameter. A positive subscript performs dilation, and a negative subscript performs erosion. Inner and outer calibration shells: ; in, For erosion calculation, For expansion operations, The envelope distance is determined based on the larger of the boundary deviation residual quantile and the boundary distance quantile of the detected cases. Central volume value: ; Lower and upper limits of volume: ; in, To detect volume, , These are the 90th and 10th percentiles of the volume deviation residuals, respectively.

7. The method according to claim 1, characterized in that, In step S5, the applicable domain of the calibration model is determined by the parameter range of the parameterized salt cavity hypothesis library; The calibration support classifies candidates into three levels: resolvable, critically resolvable, and indistinguishable, based on the preset resolvable threshold and critical resolvable threshold. For candidate objects whose resolvable features exceed the applicable domain or whose calibration support is below the indistinguishable threshold, no closed-correction entity is generated; only a conservative volume range and outer envelope starting from zero are output.

8. The method according to claim 1, characterized in that, In steps S3 and S5, the detectors that are the same as those used in the actual detection employ fixed attribute normalization upper and lower limits and detection thresholds, and apply the same detection and three-dimensional connected body labeling operations to the injected restorers and the actual candidate anomalous bodies.

9. A thin interlayered salt cavity background injection recovery calibration system, used to implement the method as described in any one of claims 1-8, characterized in that, include: The data reading and quality assessment module is used to reconstruct the measured seismic data volume and calculate the local data quality field; The hypothesis library and differential response module are used to construct a parameterized salt cavity hypothesis library with known boundary truth values ​​and generate a differential response consisting of reflection and scattering responses calibrated according to the measured background amplitude. The injection recovery and error evaluation module is used to inject the differential response into the measured background to avoid the location of the existing anomalous body, and to recover the recovered body with the same detector as the actual detection. The recovered body is then compared with the known boundary true value to obtain the detection mark, boundary deviation and volume deviation. The calibration model module is used to fit a calibration model with the resolvable features available in the restorer as independent variables; The candidate body correction and output module is used to calculate the distinguishing features of the actual candidate anomalies, correct the boundary according to the predicted deviation sign when they fall into the applicable domain of the calibration model, and output the volume range and inner and outer calibration shells based on the recovery error statistics.

Images