A feature fusion crack detection method based on seismic texture analysis

Abstract

The application provides a feature fusion crack detection method based on seismic texture analysis. The method first acquires seismic data of a target area and processes the seismic data; then performs seismic texture analysis based on cepstrum transformation, effectively and accurately characterizes weak response of the seismic data, and enhances seismic response features of cracks; on this basis, gradient structure tensor algorithm and ant colony algorithm are used to extract crack features of different scales respectively, and feature proportion fusion analysis method is used to fuse the extracted crack features of different scales; finally, reliable detection of cracks is realized by comprehensively considering geological information of the target area. The application first introduces seismic texture analysis into crack detection, improves the detection capability of micro fractures and cracks, and realizes fine description of cracks by combining feature fusion analysis, thereby providing a reliable basis for reservoir prediction and evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of oil and gas seismic exploration, in particular, to a feature fusion fracture detection method based on seismic texture analysis. BACKGROUND

[0002] Fracture is a discontinuous surface formed in the rock by tectonic deformation or physical diagenesis without obvious displacement, which is an important reservoir space and migration channel of low permeability oil and gas reservoir. Reliable detection of fractures is of great significance to oil and gas exploration and development. However, due to the multi-scale, anisotropy and weak response of fractures under deep burial conditions, the fine characterization of fractures has great challenges, and the related technical research has been a hot and difficult point in the field of oil and gas seismic exploration.

[0003] Through continuous exploration and research of predecessors, many seismic fracture detection methods have been developed, including coherence analysis, curvature analysis, ant tracking, etc. However, due to the fracture origin and scale problems, the relationship between fracture distribution and seismic response is extremely complex, the use of these methods has certain applicable conditions, there are successful cases, and there are also complete failure cases. Under certain conditions, the existing methods can accurately detect large-scale fractures, but the reliability of detecting small-scale fractures is poor, and even some small fractures with weak seismic response cannot be detected. In addition, with the continuous advancement of oil and gas exploration and the increasing complexity of exploration objects, the multi-scale, anisotropy and weak response characteristics of fractures under deep burial conditions are more and more obvious, making the reliable detection of fractures more and more difficult. SUMMARY

[0004] To solve or partially solve the problems existing in the prior art, the present application provides a feature fusion fracture detection method based on seismic texture analysis, which first introduces seismic texture analysis into fracture detection, enhances the seismic response characteristics of fractures, improves the detection ability of micro-fractures and fractures, and realizes the fine characterization of multi-scale fractures, especially micro-fractures, by combining feature fusion analysis, providing a reliable basis for reservoir prediction and evaluation.

[0005] To achieve the above purpose, the technical scheme adopted by the present application is a feature fusion fracture detection method based on seismic texture analysis, which can include:

[0006] Obtaining seismic data of a target area and processing it;

[0007] Performing seismic texture analysis based on cepstrum transformation on the processed seismic data to effectively and accurately represent the weak response of seismic data, enhance the seismic response characteristics of fractures, and improve the detection ability of micro-fractures and fractures;

[0008] On the basis of seismic texture analysis, gradient structure tensor algorithm and ant colony algorithm are used to extract different scale crack features, wherein the gradient structure tensor algorithm extracts larger scale crack features and the ant colony algorithm extracts micro crack features;

[0009] By using the feature proportion fusion analysis method, the extracted multi-scale crack features are fused according to the proportion of the dominant part to obtain a crack detection fusion feature data body.

[0010] According to the obtained fusion feature data body, the multi-scale cracks, especially the micro cracks, are finely depicted by comprehensively considering the geological information of the target region, so that the spatial distribution of the underground cracks is obtained.

[0011] Preferably, the processing of the obtained seismic data comprises:

[0012] The seismic data is subjected to amplitude compensation and structural smoothing, etc., so as to reduce or eliminate the random noise interference in the seismic data, improve the signal-to-noise ratio, and enhance the continuity of the seismic reflection events.

[0013] Preferably, the seismic texture analysis based on the cepstrum transform is performed on the seismic data, so as to effectively and accurately represent the weak response of the seismic data and improve the detection capability of the micro fractures and cracks, specifically comprising:

[0014] On the basis of the frequency spectrum analysis, the convolution of the seismic signal in the time domain, i.e. the convolution in the frequency spectrum, is changed to addition in the cepstrum, and then changed to subtraction in the decomposition process, so as to separate and extract the weak signal reflecting the crack information in the seismic data, effectively and accurately represent the weak response of the seismic data, enhance the seismic response characteristics of the cracks, and improve the detection capability of the micro fractures and cracks.

[0015] Preferably, the seismic texture analysis is a new method for studying the seismic response characteristics and detection technology of the underground fluid by establishing the consistency between the generation mechanism of the seismic record and the sound record and by referring to the sound signal processing method; wherein the seismic texture is a parameter introduced by referring to the concept of the voiceprint, and its definition is the wave pattern on the seismic data that can identify the change of the geological body property, which is the comprehensive representation of the seismic wave dynamic characteristics of the specific geological body.

[0016] Preferably, the key of the seismic texture analysis of the seismic data is not the extraction of the signal features, but the determination of the identification of the specific geological target body, and the cepstrum analysis has a good effect on separating and extracting the weak seismic signal.

[0017] Preferably, the cepstrum analysis is a homomorphic transform, a nonlinear signal processing technique widely used in speech and image processing, which is the basis of homomorphic system theory, and its characteristics are to separate and extract weak signals by logarithmic transformation of the spectrum function of time series, enhance the crack seismic response characteristics, and improve the ability to characterize small fractures and cracks, and the basic expression is:

[0018]

[0019] wherein is the original signal, is the Z transform or Fourier transform.

[0020] Preferably, the cepstrum calculation process needs to pre-process and add a time window function to the signal, and the Fourier transform or Z transform of the seismic signal is converted to the frequency domain, and then converted to the cepstrum domain.

[0021] Preferably, the gradient structure tensor algorithm is used to extract larger scale crack features, specifically:

[0022] Based on the seismic texture analysis based on the cepstrum transform of the seismic data, the gradient structure tensor algorithm for three-dimensional data is used to calculate the gradient vector at each point in the three-dimensional seismic data volume, and then the gradient structure tensor algorithm at the position is constructed using the calculated gradient vector; the gradient structure tensor at the position is constructed using the calculated directional derivative gradient vector, and then the crack features contained in the seismic data are extracted.

[0023] Preferably, the gradient structure tensor algorithm is an image processing method applied in seismic exploration, which calculates the local directional gradient to obtain the local gradient structure tensor, so as to reflect the texture characteristics of the earthquake, extract the stratigraphic dip angle and stratigraphic azimuth angle attributes from the seismic data, and then extract the crack features.

[0024] Preferably, the ant colony algorithm is used to extract small scale crack features, including:

[0025] Based on the seismic texture analysis based on the cepstrum transform of the seismic data, the crack edge is enhanced by calculating the seismic variance volume, and the spatial discontinuity of the seismic data volume is highlighted; after the edge enhancement processing of the seismic data volume, the ant colony algorithm is used to calculate the variance volume, and the crack features are extracted by optimizing the ant colony parameters suitable for the target area.

[0026] Preferably, the ant colony parameters that need to be optimized include seed points, foraging route offset degree, ant search step length, illegal step length, legal step length, and search termination threshold value.

[0027] Preferably, the ant colony algorithm is a bio-inspired swarm intelligence optimization algorithm, which follows the principle similar to the ants in their nest and food source, using pheromone that can attract ants to convey information to find the shortest path, to find the trace of the crack, and highlight the discontinuity of the seismic data to provide the basis for depicting the micro crack.

[0028] Preferably, the fusion feature data body obtained by fusing the extracted multi-scale crack features according to the proportion of the dominant part using the feature proportion fusion analysis method fuses the different scale crack features extracted using the gradient structure tensor algorithm and the ant colony algorithm. When crack detection is performed, because the fusion feature contains the feature information of the multi-scale crack, some uncertainty caused by using a single feature for crack detection can be reduced, the multi-scale crack detection is realized, and the reliability of the detection result is improved.

[0029] The beneficial effects of the present application are: compared with the prior art, the present application first introduces seismic texture analysis into crack seismic detection, realizes effective and accurate representation of weak response of seismic data, enhances the seismic response characteristics of the crack, improves the detection capability of the micro fracture and the crack, realizes fine depiction of the multi-scale crack, especially the micro crack, by combining the feature fusion analysis, is simple and easy to implement, overcomes the defects existing in the prior art, and provides a reliable basis for oil and gas reservoir prediction and evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. In the drawings:

[0031] Figure 1 A flowchart of a feature fusion crack detection method based on seismic texture analysis provided by the embodiments of the present application is shown in the figure.

[0032] Figure 2 A crack detection result graph obtained by directly extracting multi-scale crack features from seismic data and fusing them provided by the embodiments of the present application is shown in the figure.

[0033] Figure 3 A crack detection result graph obtained by extracting multi-scale crack features after using seismic texture analysis and fusing them provided by the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0034] In order to enable the persons skilled in the art to better understand the technical solutions in the present application, the technical solutions of the embodiments of the present application will be described clearly and completely below with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments of the present application.

[0035] The embodiment of the present specification provides a feature fusion crack detection method based on seismic texture analysis. Referring to Figure 1 As shown in some embodiments, the feature fusion crack detection method based on seismic texture analysis can include the following steps:

[0036] Step 101, obtaining target area seismic data and processing;

[0037] Step 102, performing seismic texture analysis based on cepstrum transformation on the processed seismic data to effectively and accurately characterize the weak response of the seismic data, enhance the seismic response characteristics of the cracks, and improve the detection ability of the micro fractures and cracks;

[0038] Step 103, on the basis of the seismic texture analysis, gradient structure tensor algorithm and ant colony algorithm are used to extract different scale crack features, wherein the gradient structure tensor algorithm extracts larger scale crack features and the ant colony algorithm extracts micro crack features;

[0039] Step 104, using feature proportion fusion analysis method, the extracted multi-scale crack features are fused according to the proportion of the dominant part to obtain crack detection fusion feature data volume;

[0040] Step 105, according to the obtained fusion feature data volume, the multi-scale cracks, especially the micro cracks, are finely described by comprehensively considering the geological information of the target area, and the spatial distribution of the underground cracks is obtained.

[0041] In one example, the seismic data is processed, including:

[0042] The seismic data is processed for amplitude compensation and structural smoothing, etc., which aims to reduce or eliminate random noise interference in the seismic data, improve the signal-to-noise ratio, and enhance the continuity of the seismic reflection events.

[0043] In one example, the seismic data is processed for seismic texture analysis based on cepstrum transformation to effectively and accurately characterize the weak response of the seismic data and improve the detection ability of the micro fractures and cracks, specifically:

[0044] Based on the spectrum analysis, the convolution of the seismic signal in the time domain, that is, the convolution in the frequency spectrum, is changed to addition in the cepstrum, and then changed to subtraction in the decomposition process, so as to separate and extract the weak signal reflecting the crack information in the seismic data, realize more effective and accurate characterization of the weak response of the seismic data, enhance the seismic response characteristics of the cracks, and improve the detection ability of the micro fractures and cracks.

[0045] In one example, the seismic data is subjected to seismic texture analysis, and the key is not the extraction of signal features, but the determination of the identification of specific geological target bodies, and the seismic texture analysis based on the cepstrum transform has a good effect on the separation and extraction of weak seismic signals.

[0046] In one example, the gradient structure tensor algorithm is used to extract large-scale crack features, and specifically:

[0047] Based on the seismic texture analysis, the gradient structure tensor algorithm for three-dimensional data first calculates the gradient vector at each point in the three-dimensional seismic data body, and then uses the calculated gradient vector to construct the gradient structure tensor algorithm at the position; the calculated directional derivative gradient vector is used to construct the gradient structure tensor at the position, and then the crack features contained in the seismic data are extracted.

[0048] In one example, the ant colony algorithm is used to extract micro-scale crack features, including:

[0049] Based on the seismic texture analysis based on the cepstrum transform, the crack edge is enhanced by calculating the seismic variance body, and the spatial discontinuity of the seismic data body is highlighted; after the edge enhancement processing of the seismic data body, the ant colony algorithm is used to calculate the variance body, and the crack features are extracted by optimizing the ant colony parameters suitable for the target area.

[0050] In one example, the ant colony parameters that need to be optimized include seed points, foraging route offset degree, ant search step length, illegal step length, legal step length, and search termination threshold value.

[0051] In one example, the feature ratio fusion analysis method is used to fuse the extracted multi-scale crack features according to the proportion of the dominant part, and a new feature fusion data body can be obtained. The new data body obtained by fusing the different scale crack features extracted by the gradient structure tensor algorithm and the ant colony algorithm can reduce the uncertainty caused by using one feature for crack detection, and improve the reliability of the detection result.

[0052] In order to make the effect of the crack detection method provided by the embodiments of the present application more clear, the crack detection of a target area is taken as an example, and the effect diagram of the prior art and the embodiments of the present application is described.

[0053] Firstly, seismic data of the target area is acquired and amplitude compensation and structure smoothing are performed; then, seismic texture analysis is performed on the processed seismic data to effectively and accurately characterize weak response of the seismic data, enhance seismic response characteristics of the fractures, and improve detection capability of micro fractures and cracks; then, gradient structure tensor algorithm is used to extract large-scale fracture characteristics, and ant colony algorithm is used to extract micro-scale fracture characteristics; then, feature proportion fusion analysis method is used to fuse the extracted multi-scale fracture characteristics to obtain fracture detection fusion feature data volume; finally, according to the obtained fusion feature data volume, multi-scale fractures, especially micro fractures, are reliably detected according to the obtained fusion feature data volume and comprehensive geological information of the target area.

[0054] Figure 2 and Figure 3 are fracture detection results of different methods, wherein Figure 2 is a fracture detection result plane obtained by directly extracting multi-scale fracture characteristics from seismic data and performing proportion fusion, and Figure 3 is a fracture detection result plane obtained by performing processing on seismic data by using seismic texture analysis, then extracting multi-scale fracture characteristics and performing proportion fusion. It can be seen from Figure 3 , Figure 3 , the depiction result of micro fractures in Figure 2 is more detailed than that in Figure 2 , and the overall large fracture part is consistent, the main fracture direction is northwest-southeast direction, the reliability of fracture detection is enhanced by comprehensive multi-feature fusion analysis, and the depiction effect of micro fractures is more obvious after using seismic texture analysis. By comparing and analyzing Figure 2 and Figure 3 , it is shown that compared with the conventional method, the feature fusion fracture detection method based on seismic texture analysis has better depiction effect on fine cracks and can more comprehensively detect fine cracks.

[0055] Those skilled in the art should understand that the above embodiments are only used to exemplarily illustrate the beneficial effects of the present application, and are not exhaustive. Any modification, equivalent replacement, improvement, etc. made without departing from the principles and spirits of the present specification should not be excluded from the protection scope of the present application.

Claims

1. A feature fusion crack detection method based on seismic texture analysis, characterized in that, The method comprises the following steps: acquiring and processing seismic data of a target area, and retaining as much as possible weak signals in the seismic data reflecting small-scale crack information; performing seismic striation analysis on the processed seismic data based on cepstrum transformation, separating and extracting weak signals in the seismic data reflecting small-scale crack information, effectively and accurately representing the weak signals, enhancing the seismic response characteristics of cracks, and improving the detection capability of small-scale cracks; on the basis of the weak signals reflecting small-scale crack information obtained through the seismic striation analysis, extracting crack features of different scales by using gradient structure tensor algorithm and ant colony algorithm, first, the gradient structure tensor algorithm is used to extract stratum dip angle and stratum azimuth angle information, then, a seismic variance volume is calculated to enhance crack edges, and then the ant colony algorithm is used to calculate the variance volume and extract crack features, wherein the gradient structure tensor algorithm extracts large-scale crack features, and the ant colony algorithm extracts small-scale crack features; using a feature proportion fusion analysis method, fusing the extracted multi-scale crack features according to the proportions of dominant parts to obtain crack detection fusion feature data volume; according to the obtained fusion feature data volume, comprehensively considering geological information of the target area to accurately detect multi-scale cracks, and forming a crack seismic detection method integrating four elements of “seismic weak signal separation, crack feature extraction, crack feature fusion, and geologically constrained crack detection”. processing the acquired seismic data comprises amplitude compensation and structural smoothing processing, which aims to reduce or eliminate random noise interference in the seismic data, improve the signal-to-noise ratio, and enhance the continuity of seismic reflection events.

2. The method according to claim 1, wherein the method is characterized by: performing seismic striation analysis on the seismic data based on cepstrum transformation, specifically, on the basis of spectral analysis, the convolution of the seismic signal in the time domain, that is, the convolution in the frequency spectrum, is changed to addition in the cepstrum, and then changed to subtraction in the decomposition process, thereby separating and extracting weak signals in the seismic data reflecting crack information, effectively and accurately representing the weak signals, enhancing the seismic response characteristics of cracks, and improving the detection capability of small-scale cracks, wherein the seismic striation is a parameter introduced by referring to the concept of voiceprint, and is defined as a change ripple on the seismic data that can identify the characteristics of a geological body.

3. The method according to claim 1, wherein the method is characterized by: extracting large-scale crack features by using the gradient structure tensor algorithm, specifically, on the basis of separating and extracting weak signals in the seismic data reflecting crack information through the seismic striation analysis, the gradient structure tensor algorithm for three-dimensional data is first used to calculate the gradient vector at each point in the three-dimensional seismic data volume, then the gradient vector is used to construct the gradient structure tensor algorithm at the position, the gradient vector of the calculated directional derivative is used to construct the gradient structure tensor at the position, and then stratum dip angle and stratum azimuth angle crack feature information contained in the weak signals of the seismic data are extracted.

4. The feature fusion fracture detection method based on seismic texture analysis according to claim 1, characterized in that: ​

Images