Complex-structure low-signal-to-noise-ratio seismic data interpretation method

Abstract

The invention discloses a complex-structure low-signal-to-noise-ratio seismic data interpretation method, and relates to the technical field of seismic data processing and interpretation, a low-signal-to-noise-ratio region is cut off, an image is repaired by using a Criminisi algorithm in the image processing field, the Criminisi algorithm can reserve local texture features of a to-be-repaired region, and the local texture features of the to-be-repaired region can be restored. And the repaired seismic data are ensured not to be distorted. And then, the overall stability of the seismic data is maintained in combination with a gradient diffusion algorithm. In the process, low-frequency components of seismic data are reserved, so that useful signals are reserved as much as possible. The method can enhance the continuity of the event, improve the signal-to-noise ratio of the seismic data and improve the efficiency and quality of interpreting the horizon interpretation of an interpretation worker under the condition of ensuring the credibility of the seismic data. According to the method, the processed local low-signal-to-noise-ratio seismic data are cut off through an image processing method, the processed local low-signal-to-noise-ratio seismic data are reconstructed and repaired, then useful signals are superposed, and the signal-to-noise ratio of the seismic data is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of seismic data processing and interpretation, and more particularly to a method for interpreting seismic data with a complex structure and a low signal-to-noise ratio. Background Art

[0002] In exploration and development, seismic horizon interpretation is the basis of oil and gas exploration, and its reliability directly affects the rationality of subsequent reservoir prediction and oil and gas reservoir evaluation results. When the signal-to-noise ratio of seismic data is high, automatic tracking technology is generally used for tracking, which can greatly save the time of interpreters. In the case of low signal-to-noise ratio caused by complex structure, developed faults, and drastic lateral changes, the automatic tracking algorithm is difficult to meet the requirements and can only rely on manual interpretation. Manual interpretation is a tedious and time-consuming task, which often depends on the experience and knowledge of the interpreter. In areas with low signal-to-noise ratio, the continuity of the phase axis is poor, and it is generally tracked according to the equal thickness principle of horizon interpretation. However, the horizon information obtained by such tracking has low credibility and cannot provide help for subsequent interpretation work. Complex surface environments and areas with relatively developed structures will cause low signal-to-noise ratios in seismic data. For example, when collecting seismic data in mountainous or desert environments, near complex structural belts such as fault development zones, thrust nappe zones, and basin-foreland zones, the seismic events will be cluttered and the signal-to-noise ratio of the seismic data will be low. Traditional methods to improve the signal-to-noise ratio are generally carried out during the processing of seismic data, such as pre-stack data reconstruction, denoising, big data synchronization, and velocity modeling under constraints. These treatments have a certain effect on improving the signal-to-noise ratio of seismic data, but it is found in the interpretation work that in areas with relatively complex structures, there are still common situations of cluttered events and low signal-to-noise ratios, which makes it impossible to track seismic events.

[0003] Low signal-to-noise ratio data are common in actual seismic data. Seismic data collected in the field are often affected by construction conditions or complex structures, which will cause the local signal-to-noise ratio of the collected seismic data to be low, which is not conducive to a series of work such as seismic data processing and interpretation. Therefore, in order to improve the signal-to-noise ratio of seismic data, it is necessary to process the missing parts to improve the signal-to-noise ratio of seismic data.

[0004] For missing seismic signals, it is generally considered to repair or reconstruct the seismic signals. Common seismic data reconstruction algorithms can be roughly divided into three categories: the first is the prediction filter reconstruction method, such as co-phase axis attribute interpolation, power spectrum difference tilt coherence interpolation, and gather weighted tilt stacking interpolation algorithms; the second is to repair the data by relying on operators such as dip time difference correction and azimuth time difference correction to complete the reconstruction of the wave field extension, such as dip time difference correction method, shot spacing domain extension method, etc.; the third is to reconstruct and repair the data in the transform domain, such as Fourier domain reconstruction method, Radon domain reconstruction, etc. These reconstruction and repair processes are effective in improving the signal-to-noise ratio of seismic data, but there will still be local low signal-to-noise ratio areas in the interpretation process, which will cause great difficulties for the interpretation work of the interpreters.

[0005] In recent decades, due to the great similarity between image data and seismic data in processing, the seismic exploration field has introduced image processing into the processing and interpretation of seismic data, and has achieved good application results. For example, image edge enhancement technology can highlight the boundaries of attribute bodies in the interpretation process, median filtering technology can filter inter-well seismic data in different gather domains, and image noise suppression technology can suppress noise in seismic data and improve the signal-to-noise ratio. Summary of the invention

[0006] In order to overcome the defects existing in the above-mentioned prior art, the present invention discloses a method for interpreting seismic data with low signal-to-noise ratio of complex structures. The purpose of the present invention is to solve the problem that the horizon is difficult to track due to the low signal-to-noise ratio in the process of horizon interpretation. The present invention cuts off the low signal-to-noise ratio area and uses the Criminisi algorithm in the field of image processing to repair the image. The Criminisi algorithm can retain the local texture features of the area to be repaired to ensure that the repaired seismic data is not distorted. Subsequently, the overall stability of the seismic data is maintained in combination with the gradient diffusion algorithm. In this process, the low-frequency components of the seismic data are also retained to retain useful signals as much as possible. This method can enhance the continuity of the event axis, improve the signal-to-noise ratio of the seismic data, and improve the efficiency and quality of the interpretation of the horizon by the interpreter while ensuring the credibility of the seismic data. The present invention cuts off the processed local low signal-to-noise ratio seismic data by an image processing method, reconstructs and repairs it, and then superimposes the useful signal to improve the signal-to-noise ratio of the seismic data, so that the continuity of the seismic event axis becomes stronger and it is easy to track the interpretation horizon.

[0007] In order to achieve the above objectives, the technical solution adopted by the present invention is: A method for interpreting complex structure low signal-to-noise ratio seismic data, comprising the following steps: S1, low-pass filtering and structural guidance filtering are performed on the original seismic data S(t) in sequence to obtain low-frequency data B(t); In the present invention, before processing the data, the low-frequency components of the seismic data are extracted to retain the useful signals as much as possible.

[0008] Preferably, in step S1, the original seismic data S(t) is input, and the original seismic data S(t) is subjected to low-pass filtering to obtain low-frequency seismic data A(t), specifically: Where S(t) is the input original seismic signal; h(t) is the filtering factor of the low-pass filter; A(t) is the low-frequency seismic data after low-pass filtering.

[0009] In the above steps, the original seismic data S(t) is input and low-pass filtered to obtain low-frequency seismic data A(t). The purpose of low-pass filtering the original seismic data is to obtain low-frequency data and retain useful signals as much as possible.

[0010] Preferably, in step S1, the low-frequency seismic data A(t) is subjected to structure-guided filtering to obtain low-frequency data B(t), specifically: Where A(t) is the low-frequency seismic data after low-pass filtering, and B(t) is the low-frequency data.

[0011] In the above steps, the low-frequency seismic data A(t) after low-pass filtering is subjected to structural guidance filtering. The purpose of structural guidance filtering is to maintain continuity along the event axis, that is, to ensure the lateral continuity of the useful signal.

[0012] S2, cutting out the low signal-to-noise ratio area of ​​the original seismic data S(t), and repairing the cut seismic data using the criminisi algorithm and the gradient diffusion algorithm in turn, to obtain the repaired seismic data E(t); In the present invention, the criminisi algorithm is combined with the gradient diffusion algorithm to repair seismic data, enhance the continuity of seismic event axes, and improve the signal-to-noise ratio. Among them, the Criminisi algorithm can retain the local texture features of the area to be repaired, ensuring that the repaired seismic data is not distorted. The gradient diffusion algorithm can maintain the overall stability of the seismic data.

[0013] Preferably, the S2 step comprises the following steps: S21, perform data evaluation on the original seismic data S(t), cut off the low signal-to-noise ratio area, and obtain the seismic data C(t) after cutting off; S22, using the criminisi algorithm to repair the seismic data C(t), restore its local texture features, and obtain the preliminarily repaired seismic data D(t); Preferably, the criminisi algorithm in step S22 to repair seismic data includes the following steps: A. Confirm the area to be filled in the seismic data C(t) to be repaired, and divide the area to be filled into several target blocks to be repaired; In step A, the area to be filled is represented as , the boundary of the filled area is represented by , the known area is expressed as , p represents the point with higher priority in the area to be repaired, Indicates the patch target block; B. calculating the priorities of a plurality of target blocks to be repaired, and using the priorities to determine the order in which the target blocks to be repaired will be repaired; In step B, the priority is: , where C(p) represents the confidence, i.e., the proportion of the known part in the patched target block; D(p) represents the data item, which is the boundary The unit normal vector at p The vector of the iso-illuminance line at this point The product of C. According to the order in which the target blocks to be repaired are to be repaired as determined in step B, the best matching blocks are sequentially searched in the known area, and the corresponding best matching blocks are sequentially filled into the corresponding positions of the target blocks to be repaired for repairing; In the step C, the best matching block is searched by using the square difference between the gray value of the patch target area and the gray value of the known area. The smaller the square difference, the higher the priority.

[0014] D. Update the filled data, reset the target block to be repaired and the confidence, and repeat the above AC steps to complete the repair of the entire seismic data and obtain the preliminary repaired seismic data D(t).

[0015] S23. Repair the seismic data D(t) using a gradient diffusion algorithm to maintain the stability of the overall structural characteristics of the seismic data, and obtain the repaired seismic data E(t).

[0016] Preferably, in step S23, the gradient diffusion filtering is: ; In the formula, t is the diffusion time, div is the divergence operator, is the gradient operator, g(•) is the diffusion function, , k is the gradient filter threshold.

[0017] Preferably, in step S23, the repair model is: Where t is the iteration time, which is equivalent to the diffusion time. is the coefficient of change of the diffusion filter direction, and N is the number of iterations.

[0018] Preferably, in step S23, the repair model is used to iteratively repair the seismic data D(t) to be repaired until the absolute value of the residual before and after the iteration of the area to be repaired is less than 0, the iteration is terminated, and the repaired seismic data E(t) is output.

[0019] S3, superimposing the low-frequency data B(t) and the restored seismic data E(t) and outputting the superimposed data F(t); In the above steps, the low-frequency data B(t) and the repaired seismic data E(t) are superimposed to obtain the superimposed data F(t), which is specifically: .

[0020] S4. Perform amplitude equalization processing on the stacked data F(t), output the amplitude equalized seismic data S(t) and perform complex structural interpretation on the seismic data S(t).

[0021] In the above steps, the superimposed seismic data F(t) is amplitude equalized, the repaired seismic data S(t) is output, and the structural interpretation is completed on the repaired seismic data.

[0022] Beneficial effects of the present invention: The complex structure low signal-to-noise ratio seismic data interpretation method provided by the present invention is different from the common method of improving the signal-to-noise ratio. The core idea of ​​the present invention is to use the criminisi algorithm combined with the gradient diffusion algorithm to repair the seismic data, enhance the continuity of the seismic phase axis, and improve the signal-to-noise ratio. In this process, the low-frequency components of the seismic data are extracted to maintain the low-frequency authenticity of the seismic data, so as to achieve the purpose of preserving the fidelity of the seismic data, effectively reduce the loss of useful signals in the seismic data, and ensure the credibility of the repaired seismic data.

[0023] The complex structure low signal-to-noise ratio seismic data interpretation method provided by the present invention is mainly to solve the problem that the horizon is difficult to track due to the low signal-to-noise ratio during the horizon interpretation process. The image is repaired by cutting out the low signal-to-noise ratio area using the Criminisi algorithm in the field of image processing. The Criminisi algorithm can retain the local texture features of the area to be repaired, ensuring that the repaired seismic data is not distorted. Subsequently, the overall stability of the seismic data is maintained in combination with the gradient diffusion algorithm. In this process, the low-frequency components of the seismic data are also retained to retain useful signals as much as possible. This method can enhance the continuity of the event axis, improve the signal-to-noise ratio of the seismic data, and improve the efficiency and quality of the horizon interpretation of the interpreter while ensuring the credibility of the seismic data. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The interpretation process of low signal-to-noise ratio seismic data of complex structures in the present invention; Figure 2 This is a schematic diagram of the Criminisi algorithm patching of the present invention. DETAILED DESCRIPTION

[0025] The concept, specific structure and technical effects of the present invention will be clearly and completely described below in conjunction with the embodiments and drawings to fully understand the purpose, characteristics and effects of the present invention.

[0026] Example 1 A method for interpreting low signal-to-noise ratio seismic data in complex structures, such as Figure 1 As shown, the following steps are included: S1, low-pass filtering and structural guidance filtering are performed on the original seismic data S(t) in sequence to obtain low-frequency data B(t); S2, cutting out the low signal-to-noise ratio area of ​​the original seismic data S(t), and repairing the cut seismic data using the criminisi algorithm and the gradient diffusion algorithm in turn, to obtain the repaired seismic data E(t); S3, superimposing the low-frequency data B(t) and the restored seismic data E(t) and outputting the superimposed data F(t); S4. Perform amplitude equalization processing on the stacked data F(t), output the amplitude equalized seismic data S(t) and perform complex structural interpretation on the seismic data S(t).

[0027] In this embodiment, the criminisi algorithm is combined with the gradient diffusion algorithm to repair the seismic data, enhance the continuity of the seismic event axis, and improve the signal-to-noise ratio. In this process, the low-frequency components of the seismic data are extracted to maintain the low-frequency authenticity of the seismic data, achieve the purpose of preserving the fidelity of the seismic data, effectively reduce the loss of useful signals in the seismic data, and ensure the credibility of the repaired seismic data.

[0028] In this embodiment, it is mainly to solve the phenomenon that the horizon is difficult to track due to the low signal-to-noise ratio during the horizon interpretation process. The image is repaired by cutting out the low signal-to-noise ratio area and using the Criminisi algorithm in the field of image processing. The Criminisi algorithm can retain the local texture features of the area to be repaired and ensure that the repaired seismic data is not distorted. Subsequently, the gradient diffusion algorithm is combined to maintain the overall stability of the seismic data. In this process, the low-frequency components of the seismic data are also retained to retain useful signals as much as possible. This method can enhance the continuity of the event axis, improve the signal-to-noise ratio of the seismic data, and improve the efficiency and quality of the horizon interpretation of the interpreter while ensuring the credibility of the seismic data.

[0029] Example 2 This embodiment further describes step S1 based on embodiment 1. In the present invention, before processing the data, the low-frequency components of the seismic data are extracted to retain the useful signals as much as possible.

[0030] In step S1, the original seismic data S(t) is input, and the original seismic data S(t) is low-pass filtered to retain useful signals as much as possible to obtain low-frequency seismic data A(t), specifically: Where S(t) is the input original seismic signal; h(t) is the filtering factor of the low-pass filter; A(t) is the low-frequency seismic data after low-pass filtering.

[0031] In the step S1, the low-frequency seismic data A(t) is subjected to structure-guided filtering to maintain continuity along the event axis, that is, to ensure the lateral continuity of the useful signal, and obtain low-frequency data B(t), specifically: Where A(t) is the low-frequency seismic data after low-pass filtering, and B(t) is the low-frequency data.

[0032] Example 3 This embodiment further elaborates on step S2 based on embodiment 2. In the present invention, the criminisi algorithm is combined with the gradient diffusion algorithm to repair seismic data, enhance the continuity of seismic event axes, and improve the signal-to-noise ratio. Among them, the Criminisi algorithm can retain the local texture features of the area to be repaired, ensuring that the repaired seismic data is not distorted. The gradient diffusion algorithm can maintain the overall stability of the seismic data.

[0033] The S2 step includes the following steps: S21, perform data evaluation on the original seismic data S(t), cut off the low signal-to-noise ratio area, and obtain the seismic data C(t) after cutting off; S22, using the criminisi algorithm to repair the seismic data C(t), restore its local texture features, and obtain the preliminarily repaired seismic data D(t); The criminisi algorithm of step S22 to repair seismic data includes the following steps: A. Confirm the area to be filled in the seismic data C(t) to be repaired, and divide the area to be filled into several target blocks to be repaired; like Figure 2 As shown, in step A, the area to be filled is represented as , the boundary of the filled area is represented by , the known area is represented by , p represents the point with higher priority in the area to be repaired, Indicates the patch target block; B. calculating the priorities of a plurality of target blocks to be repaired, and using the priorities to determine the order in which the target blocks to be repaired will be repaired; In step B, the priority is: , where C(p) represents the confidence, i.e., the proportion of the known part in the patched target block; D(p) represents the data item, which is the boundary The unit normal vector at p The vector of the iso-illuminance line at this point The product of C. According to the order in which the target blocks to be repaired are to be repaired as determined in step B, the best matching blocks are sequentially searched in the known area, and the corresponding best matching blocks are sequentially filled into the corresponding positions of the target blocks to be repaired for repairing; In the step C, the best matching block is searched by using the square difference between the gray value of the patch target area and the gray value of the known area. The smaller the square difference, the higher the priority.

[0034] D. Update the filled data, reset the target block to be repaired and the confidence, and repeat the above AC steps to complete the repair of the entire seismic data and obtain the preliminary repaired seismic data D(t).

[0035] S23. Repair the seismic data D(t) using a gradient diffusion algorithm to maintain the stability of the overall structural characteristics of the seismic data, and obtain the repaired seismic data E(t).

[0036] In step S23, the gradient diffusion filtering is: ; In the formula, t is the diffusion time, div is the divergence operator, is the gradient operator, g(•) is the diffusion function, , k is the gradient filter threshold.

[0037] In the step S23, the repair model is: Where t is the iteration time, which is equivalent to the diffusion time. is the coefficient of change of the diffusion filter direction, and N is the number of iterations.

[0038] In step S23, the repair model is used to iteratively repair the seismic data D(t) to be repaired until the absolute value of the residual before and after the iteration of the repaired area is less than 0, the iteration is terminated, and the repaired seismic data E(t) is output.

[0039] Example 4 This embodiment further elaborates on step S3 and step S4 based on embodiment 3. In step S3, the low-frequency data B(t) and the repaired seismic data E(t) are superimposed to obtain the superimposed data F(t), which is specifically: In step S4, amplitude equalization is performed on the superimposed seismic data F(t), and the repaired seismic data S(t) is output, and structural interpretation is completed on the repaired seismic data.

[0040] The complex structure low signal-to-noise ratio seismic data interpretation method provided by the present invention can be optimized in two aspects. First, the Criminisi algorithm used in the technology can be replaced by an algorithm with similar effects. For example, a mathematical modeling method is combined with the Criminisi algorithm to achieve local image repair, maintain texture features and improve the signal-to-noise ratio. Compressed sensing or unsupervised neural network algorithms can also be selected to repair the image. Second, the structure-guided filtering used in the technology is to ensure the lateral continuity of the useful signal, which can also be replaced by a similar filter.

[0041] The above is a specific description of the implementation mode of the present invention, but the present invention is not limited to the described embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention, and these equivalents or substitutions are all included in the scope defined by the claims of the present invention.

Claims

1. A method for interpreting complex structure low signal-to-noise ratio seismic data, characterized in that: The following steps are involved: S1, low-pass filtering and structural guidance filtering are performed on the original seismic data S(t) in sequence to obtain low-frequency data B(t); S2, cutting out the low signal-to-noise ratio area of ​​the original seismic data S(t), and repairing the cut seismic data using the criminisi algorithm and the gradient diffusion algorithm in turn, to obtain the repaired seismic data E(t); S3, superimposing the low-frequency data B(t) and the restored seismic data E(t) and outputting the superimposed data F(t); S4. Perform amplitude equalization processing on the stacked data F(t), output the amplitude equalized seismic data S(t) and perform complex structural interpretation on the seismic data S(t).

2. The seismic data interpretation method according to claim 1, characterized in that: In step S1, the original seismic data S(t) is input, and the original seismic data S(t) is subjected to low-pass filtering to obtain low-frequency seismic data A(t), specifically: Where S(t) is the input original seismic signal; h(t) is the filtering factor of the low-pass filter; A(t) is the low-frequency seismic data after low-pass filtering.

3. The seismic data interpretation method according to claim 2, characterized in that: In step S1, the low-frequency seismic data A(t) is subjected to structure-guided filtering to obtain low-frequency data B(t), specifically: Where A(t) is the low-frequency seismic data after low-pass filtering, and B(t) is the low-frequency data.

4. The seismic data interpretation method according to claim 1, characterized in that: The S2 step includes the following steps: S21, perform data evaluation on the original seismic data S(t), cut off the low signal-to-noise ratio area, and obtain the seismic data C(t) after cutting off; S22, using the criminisi algorithm to repair the seismic data C(t), restore its local texture features, and obtain the preliminarily repaired seismic data D(t); S23. Repair the seismic data D(t) using a gradient diffusion algorithm to maintain the stability of the overall structural characteristics of the seismic data, and obtain the repaired seismic data E(t).

5. The seismic data interpretation method according to claim 4, characterized in that: The criminisi algorithm of step S22 for repairing seismic data includes the following steps: A. Confirm the area to be filled in the seismic data C(t) to be repaired, and divide the area to be filled into several target blocks to be repaired; B. calculating the priorities of a plurality of target blocks to be repaired, and using the priorities to determine the order in which the target blocks to be repaired will be repaired; C. According to the order in which the target blocks to be repaired are to be repaired as determined in step B, the best matching blocks are sequentially searched in the known area, and the corresponding best matching blocks are sequentially filled into the corresponding positions of the target blocks to be repaired for repairing; D. Update the filled data, reset the target block to be repaired and the confidence, and repeat the above AC steps to complete the repair of the entire seismic data and obtain the preliminary repaired seismic data D(t).

6. The seismic data interpretation method according to claim 5, characterized in that: In step A, the area to be filled is represented as , the boundary of the filled area is represented by , the known area is represented by , p represents the point with higher priority in the area to be repaired, Indicates the patch target block; In step B, the priority is: , where C(p) represents the confidence, that is, the proportion of the known part in the patch target block; D(p) represents the data item, which is the boundary The unit normal vector at p The iso-illuminance line vector at this point The product of .

7. The seismic data interpretation method according to claim 6, characterized in that: In the step C, the best matching block is searched by using the square difference between the gray value of the patch target area and the gray value of the known area. The smaller the square difference, the higher the priority.

8. The seismic data interpretation method according to claim 4, characterized in that: In step S23, the gradient diffusion filtering is: ； In the formula, t is the diffusion time, div is the divergence operator, is the gradient operator, g(•) is the diffusion function, , k is the gradient filter threshold.

9. The seismic data interpretation method according to claim 8, characterized in that: In the step S23, the repair model is: Where t is the iteration time, which is equivalent to the diffusion time. is the coefficient of change of the diffusion filter direction, and N is the number of iterations.

10. The seismic data interpretation method according to claim 9, characterized in that: In step S23, the repair model is used to iteratively repair the seismic data D(t) to be repaired until the absolute value of the residual before and after the iteration of the repaired area is less than 0, the iteration is terminated, and the repaired seismic data E(t) is output.

Images