First arrival identification method based on spatial attribute change

Abstract

The invention provides a first arrival identification method based on spatial attribute change. The first arrival identification method based on spatial attribute change comprises the following steps: step 1, inputting a seismic trace set of which first arrival needs to be picked up; step 2, inter-track time difference elimination based on a small smooth reference surface is carried out; step 3, reconstructing the gather by using the spatial correlation of the gather by using a PCA method, extracting a wavelet template on the PCA reconstructed gather, and calculating the spatial attribute of each point of the gather by using a correlation coefficient method; and 4, identifying the first-motion wave according to the spatial attribute change. According to the first arrival recognition method based on the spatial attribute change, the problem that first arrival wave recognition is inaccurate under the complex earth surface condition is solved, statistical correlation in the first arrival signal space is fully utilized, element-mode seismic wavelets of seismic exploration are combined, and stable first arrival wave recognition under the complex earth surface condition is achieved.

Description

Technical Field

[0001] This invention relates to the field of geophysical exploration technology, and in particular to a method for identifying first arrivals based on spatial attribute changes. Background Technology

[0002] Seismic exploration of oil and gas has entered a stage of exploring complex surfaces, structures, reservoirs, and deep targets, and the "two-wide-one-high" seismic data acquisition has become an industry consensus. The foreland / piedmont zones of oil and gas basins in western my country have become strategic replacement areas for my country's oil and gas resources, achieving good exploration results. However, this does not mean that we have completely solved the core technical problems of seismic exploration in "dual-complex" areas. The fundamental problem in foreland seismic data imaging processing lies in how to establish a velocity model that meets the requirements of migration imaging based on rapidly varying inter-track time differences and low signal-to-noise ratio data. First-arrival identification and travel time detection play a crucial role in the estimation and modeling of surface and shallow medium velocities. For foreland areas with complex near-surface conditions, conventional first-arrival picking methods are insufficient, thus necessitating the development of first-arrival picking methods under complex surface conditions.

[0003] First-arrival picking is an essential step in seismic data processing. The first arrival is the first signal generated by the source recorded in the seismic trace ensemble. Typically, the first arrival is associated with the direct wave at near offsets and the refracted or rotating wave at far offsets. The picked first arrival is often used in tomography to estimate near-surface velocity structure and for static correction. The accuracy of tomography and the effectiveness of static correction both depend on the accuracy of the first arrival picking. For "two-wide-one-high" seismic data acquisition, the massive amount of seismic data generates a significant workload for first arrival picking. Therefore, with the increasing number of channels in seismic acquisition systems, automatic or semi-automatic picking methods are essential to reduce seismic data processing time.

[0004] Over the past few decades, numerous methods have been developed to improve the effectiveness and efficiency of first-arrival picking, taking into account various seismic physics and properties. In recent years, many machine learning methods have also been applied to automatic first-arrival picking, including artificial neural networks (Mai et al., 2014), fuzzy clustering (Chen, 2017), and support vector machines (Duan and Zhang, 2019). Based on CNNs (Hollander et al., 2018), a variety of new networks have been applied to first-arrival picking, such as SegNet (Badrinarayanan et al., 2017; Wu et al., 2019), PhaseNet (Zhuand Beroza, 2018), and U-Net (Hu et al., 2019; Ma et al., 2020). However, machine learning methods are highly dependent on labels; label selection and network training computations are both very time-consuming during the training process. Furthermore, overfitting and lack of diversity in the training data also reduce the stability of these machine learning methods, making them perform well on training data but poorly on new data.

[0005] Although these methods have been successfully applied and proven to be effective, they are almost all very sensitive to noise. If the signal-to-noise ratio of the gather is low, incorrect first arrival picking results will be produced.

[0006] Chinese patent application CN202111422056 discloses a first-arrival picking method based on strong noise and weak signal detection. For mountainous data in my country with complex surface conditions, characterized by poor signal-to-noise ratios and drastic inter-trace time differences, traditional picking methods often fail to yield satisfactory results. This invention proposes a first-arrival picking process specifically for mountainous data. The process begins with preprocessing, including inter-trace time difference abrupt change suppression based on a smooth floating reference surface, amplitude compensation, and first-arrival filtering using an anisotropic Gaussian filter. Next, high-dimensional, multi-attribute extraction is performed. Then, a clustering algorithm is used to determine the approximate range for first-arrival picking. Finally, multi-attribute Markov decision-making is applied to pick the first arrival, yielding the final picking result. Compared to traditional methods, this invention achieves automated and intelligent picking, specifically targeting seismic data from mountainous areas in my country with low signal-to-noise ratios and drastic inter-trace time differences. However, this method extracts the first-arrival structure based on three attributes: energy ratio, kurtosis, and edge intensity. Among these, the energy ratio and kurtosis attributes are single-channel attributes and do not utilize the spatial correlation of the first arrival signal. The edge intensity attribute only uses information from two adjacent channels. Furthermore, this method is an image-based approach, requiring the seismic image to be converted to grayscale, which in turn affects the extraction accuracy of edge attributes. Therefore, none of these three attributes fully utilize the spatial statistical correlation of the first arrival signal. In addition, these three attributes are general attributes for signal extraction across various fields and do not effectively integrate the unique meta-modes in exploration seismicity—the seismic wavelet—therefore, there is room for improvement.

[0007] The paper "Research on First Arrival Picking Method for Low SNR Microseismic Signals Based on Adaptive Morlet Wavelet" (Progress in Geophysics, 2019, 34(2):561-567) addresses the problem of low signal-to-noise ratio (SNR) data affecting first arrival picking in microseismic events. Traditional methods often suffer from unsatisfactory picking accuracy and stability. To overcome the inability to effectively pick first arrivals under low SNR conditions, this paper designs an adaptive Morlet wavelet basis with high sensitivity to target components. By using this wavelet basis to perform wavelet decomposition on microseismic records, and utilizing the characteristic correlation of the effective components of the three-component data in the wavelet domain, principal component analysis is performed on the three-component wavelet coefficients to extract principal component features. Finally, weighted reconstruction of the principal components at each level is performed to achieve first arrival picking for low SNR microseismic signals. In model experiments with different SNRs and in practical data applications, the algorithm demonstrates excellent noise resistance. Even under extremely low SNR conditions, it can still accurately indicate the first arrival of the effective components. Both model experiments and actual data processing results have verified the effectiveness and practicality of this method in first arrival picking of microseismic data with extremely low signal-to-noise ratio. This method has high theoretical and application value in microseismic monitoring and related fields.

[0008] The paper "Detection of First Arrival Signals in Low Signal-to-Noise Ratio Microseismic Data by Combining CEEMDAN and Principal Component Analysis" (Petroleum Geophysical Exploration, 2019, Vol. 54, Issue (1): 45-53) effectively detects first arrival features in low signal-to-noise ratio (SNR) microseismic data by combining adaptive noise-complete empirical mode decomposition (CEEMDAN) with principal component analysis (PCA). After CEEMDAN processing of low SNR microseismic data, PCA is performed on the intrinsic mode functions (IMFs) of each order, and then the principal components of each IMF are weighted and reconstructed. At the same time, minor components are suppressed and removed, so that the first arrival information with strong consistency in the three-component signal is preserved. Multiple sets of test signals with different SNRs are designed to test the feasibility of the method, and finally applied to the three-component measured signals. The results show that the method can still effectively identify and detect the first arrival of microseismic signals under extremely low SNR conditions.

[0009] Although the two articles above apply the PCA method to low signal-to-noise ratio (SNR) first-arrival picking, the former first uses wavelet decomposition to extract features from the first-arrival signal, and then uses PCA to reconstruct the extracted features (i.e., wavelet decomposition coefficients) to extract the principal features; the latter first uses the CEEMDAN algorithm to perform empirical mode decomposition on the low SNR data, and then uses PCA to extract the principal components and suppress and remove the minor components of each order of intrinsic mode functions (IMFs). Both articles first use other methods to extract features from the first-arrival signal for feature generation, and then use PCA to select features from the generated multiple features to extract the main feature components of the first arrival and generate the final features used. This is the conventional use of the PCA method. Applying it to multiple feature selection after feature generation, extracting the principal features from multiple different features, the final result is highly dependent on the feature generation in the first step. Only when the features of the first arrival are truly generated can PCA be used to capture the principal features in the generated features and achieve the expected results.

[0010] The existing technologies described above are quite different from the present invention and have failed to solve the technical problem we want to solve. Therefore, we have invented a new method for initial arrival identification based on changes in spatial attributes. Summary of the Invention

[0011] The purpose of this invention is to provide a first-arrival identification method based on spatial attribute changes that achieves stable identification of first-arrival waves under complex surface conditions.

[0012] The objective of this invention can be achieved through the following technical measures: a method for identifying the first arrival based on spatial attribute changes, comprising:

[0013] Step 1: Input the seismic gathers from which the first arrival needs to be picked;

[0014] Step 2: Perform inter-track time difference elimination based on a small smoothing reference surface;

[0015] Step 3: Use the PCA method to reconstruct the gather using the spatial correlation of the gather, extract the wavelet template from the PCA reconstructed gather, and use the correlation coefficient method to calculate the spatial attributes of each point in the gather;

[0016] Step 4: Identify the first arrival wave based on changes in spatial properties.

[0017] The objective of this invention can also be achieved through the following technical measures: In step 1, the wavelet convolution reflection coefficient is used as the linear first arrival data after eliminating inter-trace time difference. By default, the elevations of each trace are equal. Nonlinear inter-trace time difference is randomly applied to the data. When a certain surface velocity is given, the elevation is calculated based on the nonlinear inter-trace time difference. At the same time, noise is applied to simulate the seismic gathers that need to be picked up under the complex surface with known surface elevation. This data is used as the input seismic gather.

[0018] In step 1, given a surface velocity of 2000 m / s, the elevation is calculated based on the nonlinear inter-channel time difference, while simultaneously applying noise to achieve a signal-to-noise ratio of -3 dB. The signal-to-noise ratio calculation formula is as follows: s represents the signal and n represents the noise. This is used to simulate the seismic gathers that need to be picked up under the complex surface with known surface elevation. This data is used as the input seismic gather.

[0019] In step 2, the surface elevation is statistically analyzed and smoothed to generate a smoothed surface elevation. Then, the data is corrected from the real surface to the smoothed elevation surface using a formula, so that the initial elevation becomes linear and the spatial correlation between the data is enhanced.

[0020] In step 2, the formula for correcting the data from the true ground surface to a smooth elevation surface is:

[0021]

[0022] Among them, elev smooth : Smoothed elevation surface; elev real : Actual ground elevation; v smoot h Smooth speed, Δt This is the time difference correction amount. Based on the formula above, it is calculated using the actual elevation and smoothed elevation of each seismic trace. Thus, each seismic trace has a time difference correction amount, which is applied to the original data to correct for time difference; that is, each seismic trace is adjusted along the time axis. Δt By translating the length, data can be obtained on a smooth elevation surface, eliminating the arrival time jump between tracks.

[0023] In step 3, spatial attributes are extracted, a wavelet template is selected, and then the wavelet template is matched to obtain attributes that characterize the similarity between the wavelet template and the wavelet template.

[0024] In step 3, firstly, the principal component reconstruction data is selected using PCA on the output results of step 2, and the wavelet extracted from the PCA-reconstructed gather is used as the template. Then, for each seismic trace, the position of the first arrival wave is found on the PCA-reconstructed gather, and windows are opened above and below the time axis as wavelet templates. The correlation coefficient between the wavelet template of each trace and the corresponding window of the output data in step 2 is calculated using the correlation method.

[0025] In step 3, the formula for calculating the correlation coefficient is:

[0026]

[0027] Where w represents the length of the time window. a Represents the wavelet template. ai This represents the amplitude value at each sampling point of the wavelet template; b This represents the signal with the same window length as the wavelet template on the corresponding seismic trace in the original data. bi The amplitude value represents the signal at each sampling point; the correlation coefficient calculation formula characterizes the similarity between the windowed data on the corresponding seismic trace and the wavelet template of that trace; the greater the similarity, the more likely it is to be a first arrival; this formula characterizes the similarity between two time series of equal length, ranging from 0 to 1, with the closer to 1 indicating a greater similarity between the two time series; the larger the value of this formula, the more similar the windowed time series on the original data is to the wavelet template on the reconstructed trace, that is, the greater the probability that the center point of the time series on the original data is a first arrival.

[0028] In step 3, a time window is opened near the first arrival wave of the PCA reconstructed gather as the wavelet template for that gather. The time window is opened on the first gather of the output data in step 2 and the gather is traversed to calculate the spatial attributes corresponding to the first gather. A time window is opened near the first arrival wave of the PCA reconstructed gather as the wavelet template for that gather. The time window is opened on the second gather of the output data in step 2 and the gather is traversed to calculate the spatial attributes corresponding to the second gather. The spatial attributes extracted from the output data in step 2 are calculated one gather at a time.

[0029] In step 4, for the output of step 3, find the maximum value point of the spatial attribute in each path, that is, the position of the initial wave.

[0030] The objective of this invention can also be achieved through the following technical measures: a first arrival identification system based on spatial attribute changes, which uses a first arrival identification method based on spatial attribute changes to identify first arrival waves under complex surface conditions.

[0031] The first-arrival identification method based on spatial attribute variations in this invention solves the problem of inaccurate first-arrival identification under complex surface conditions (low signal-to-noise ratio, drastic changes in inter-trace time difference). Based on the understanding of the spatial correlation of first-arrivals, spatial attributes are extracted, and the changes in spatial attributes are utilized to stably identify first-arrivals under complex surface conditions (low signal-to-noise ratio, drastic changes in inter-trace time difference). Compared with conventional first-arrival identification methods, this invention starts from the characteristics of first-arrival data under complex surface conditions, and uses an inter-trace time difference jump suppression method based on a small smooth datum surface for preprocessing to eliminate inter-trace time difference, restoring the first-arrival to (approximately) linear. Then, it fully utilizes the statistical correlation information on the (approximately) linear co-axial space to generate a reconstructed gather and extract wavelet templates. Subsequently, the correlation coefficient is calculated for each trace to obtain spatial attributes. This fully utilizes the statistical correlation on the first-arrival signal space (manifested as an approximately linear co-axial structure), and combines it with the meta-mode seismic wavelet of seismic exploration to achieve stable first-arrival identification under complex surface conditions (low signal-to-noise ratio, drastic changes in inter-trace time difference). Attached Figure Description

[0032] Figure 1 This is a flowchart of a specific embodiment of the first arrival identification method based on spatial attribute changes of the present invention;

[0033] Figure 2 This is a schematic diagram of the first arrival in a specific embodiment 1 of the present invention, where there is an inter-track time difference;

[0034] Figure 3 This is a schematic diagram of the initial arrival after inter-channel time difference jump suppression in a specific embodiment 1 of the present invention;

[0035] Figure 4 This is a schematic diagram of the reconstructed trace set obtained by PCA reconstruction of the output trace set in step 2 in a specific embodiment 1 of the present invention;

[0036] Figure 5 This is a schematic diagram of the spatial attributes extracted from the result of step 2 according to step 3 in a specific embodiment of the present invention;

[0037] Figure 6 This is a schematic diagram of the initial arrival result identified in the attribute space in a specific embodiment 1 of the present invention;

[0038] Figure 7 This is a schematic diagram of the first arrival result identified in a specific embodiment 1 of the present invention, displayed on the data after the inter-channel time difference jump is suppressed;

[0039] Figure 8 This is a schematic diagram of the actual data input in a specific embodiment 2 of the present invention, corresponding to the input earthquake record in step 1;

[0040] Figure 9This is a schematic diagram of the result of inter-channel time difference jump suppression in a specific embodiment 2 of the present invention;

[0041] Figure 10 This is a schematic diagram of a gather generated from actual data by PCA reconstruction in a specific embodiment 2 of the present invention;

[0042] Figure 11 This is a schematic diagram of the spatial attributes extracted in a specific embodiment 2 of the present invention;

[0043] Figure 12 This is a schematic diagram of the initial arrival result identified in the attribute space in a specific embodiment 2 of the present invention;

[0044] Figure 13 This is a schematic diagram of the first arrival result identified in a specific embodiment 2 of the present invention, displayed on the data after the inter-channel time difference jump is suppressed;

[0045] Figure 14 This is a schematic diagram of the actual data input in specific embodiment 3 of the present invention, corresponding to the input earthquake record in step 1;

[0046] Figure 15 This is a schematic diagram of the result of suppressing the inter-channel time difference jump in a specific embodiment 3 of the present invention;

[0047] Figure 16 This is a schematic diagram of the gather generated by PCA reconstruction of actual data in a specific embodiment 3 of the present invention;

[0048] Figure 17 This is a schematic diagram of the spatial attributes extracted in a specific embodiment 3 of the present invention;

[0049] Figure 18 This is a schematic diagram illustrating the identification of initial arrival results within the attribute space in a specific embodiment 3 of the present invention;

[0050] Figure 19 This is a schematic diagram showing the first arrival result identified in embodiment 3 of the present invention on the data after the inter-channel time difference jump is suppressed. Detailed Implementation

[0051] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0052] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, and / or combinations thereof.

[0053] For first-arrival acquisition of mountain earthquake data, the first step is elevation correction, a standardized preprocessing method used to suppress abrupt changes in inter-trace time differences and enhance the lateral continuity of first arrivals. First-arrival acquisition can be viewed as a signal detection problem. The first-arrival data received by the source detector contains statistically relevant information, which manifests in space as a (approximately) linear first-arrival co-axial structure (if there is no statistically relevant information between data points, then the data contains no information and therefore no structure, such as white noise). When processing weak signals, we must fully utilize the statistically relevant information within the signal to effectively extract its structure. In other words, during first-arrival acquisition, we must fully utilize the spatially relevant statistical information of the first-arrival signal to extract the first-arrival structure more effectively.

[0054] PCA (Programmatic Conversion A) is a well-known method in multivariate analysis and data mining. Proposed by Karl Pearson in 1901 and first applied to statistics, it is based on the KL transform, a special type of orthogonal transform. PCA plays a crucial role in many practical applications, such as data compression, eigenvalue extraction, noise filtering, signal recovery, and classification. Its main limitations include low accuracy in matrix approximation, sensitivity to outliers, and high computation time and memory requirements.

[0055] This invention is based on the understanding of the spatial correlation of first arrival signals and the basic element of seismic data, the seismic wavelet. It uses PCA within a local data window to generate a good wavelet template by truly utilizing information from multiple channels (not just two adjacent channels). Then, it uses the correlation coefficient method to calculate the spatial properties that truly utilize spatial correlation and have good noise resistance for first arrival picking.

[0056] like Figure 1 As shown, Figure 1 This is a flowchart of the first-arrival identification method based on spatial attribute changes according to the present invention. The first-arrival identification method based on spatial attribute changes includes:

[0057] Step 1: Input the seismic gathers from which the first arrivals need to be picked.

[0058] Step 2: Elimination of inter-track time difference based on small smooth reference surface.

[0059] Step 3: Use the PCA method to reconstruct the gather using the spatial correlation of the gather, extract the wavelet template from the PCA reconstructed gather, and then use the correlation coefficient method to calculate the spatial properties of each point in the gather.

[0060] Step 4: Identify the first arrival wave based on changes in spatial properties.

[0061] Specifically, it includes:

[0062] In step 1, the wavelet convolution reflection coefficient is used as the (approximate) linear first-arrival data after inter-trace time difference elimination. By default, the elevations of all traces are equal. Nonlinear inter-trace time differences are randomly applied to the data, and the elevation is calculated based on these nonlinear inter-trace time differences at a given surface velocity. Simultaneously, noise is applied to simulate a seismic gather requiring first-arrival picking on a complex surface with known surface elevations (drastic changes in inter-trace time differences and low signal-to-noise ratio). This data is used as the input seismic gather.

[0063] In step 2, we calculate the surface elevation and smooth it to generate a smoothed surface elevation. Then, we use the formula... (elev smooth : Smoothed elevation surface; elev real : Actual ground elevation; v smooth (Smoothing speed) corrects the data from the real ground surface to a smooth elevation surface, resulting in a gather.

[0064] This makes the initial arrival (approximately) linear, thus enhancing the spatial correlation between data.

[0065] In step 3, we perform spatial attribute extraction. In step 2, we performed inter-channel time difference elimination, which enhanced the spatial correlation between first-arrival data. Here, we perform spatial attribute extraction. The correlation coefficient method is a method that considers multi-channel information. In its specific implementation, a wavelet template is first selected, and then the wavelet template is matched to obtain attributes that characterize the similarity to the wavelet template. Its effectiveness is highly dependent on the selection of the template.

[0066]

[0067] Among them, a i and b i Let w represent two discrete signals, and w represent the length of the time window.

[0068] a Represents the wavelet template. ai This represents the amplitude value at each sampling point of the wavelet template; b This represents the signal with the same window length as the wavelet template on the corresponding seismic trace in the original data. biThis represents the amplitude value at each sampling point of the signal. w represents the length of the time window. This formula characterizes the similarity between two time series of equal length, ranging from 0 to 1. The closer to 1, the greater the similarity between the two time series.

[0069] In this patent, the larger the value of the formula, the more similar the windowed time series on the original data is to the wavelet template on the reconstructed gather, that is, the greater the probability that the center point of the time series on the original data is the first arrival.

[0070] Conventional correlation methods for first arrival picking often select a data window on the time axis containing the wavelet near the shot point as the wavelet template. However, seismic wavelets are spatially variable; they change as they move away from the shot point. Therefore, this method of selecting a template for correlation analysis is ineffective at extracting attributes far from the shot point. Another approach is to open a time axis data window containing the wavelet at the previous first arrival picking position during the picking process, using this as the template for the next correlation calculation, calculating and adjusting it step by step. However, since a segment of signal is selected from the original data as the template, the wavelet template is usually noisy. This method often lacks accuracy and performs particularly poorly at long offsets with low signal-to-noise ratios.

[0071] The innovation of this application lies in the following: Based on the understanding of the spatial correlation of first-arrival data and the seismic wavelet, the most basic element of seismic data, a new wavelet template is generated within the template matching framework by combining the PCA method with the unique meta-mode seismic wavelet of seismic data. This ensures that the correlation coefficient attribute calculated under the new wavelet template is specific to the seismic data, contains multi-channel first-arrival information, and is a sufficiently robust spatial attribute, which is then used for first-arrival identification. In the generation of the new wavelet template, we use the PCA method. Because the new wavelet template is generated on the PCA reconstructed gather using the spatial (i.e., lateral) correlation of first arrivals, each wavelet template has a high signal-to-noise ratio and contains relevant lateral information from multiple first arrivals within the two-dimensional window participating in the PCA calculation. Furthermore, it retains some spatially variable information of the wavelet (gradually changing wavelet morphology during PCA reconstruction is preserved). Therefore, the spatial attributes generated by subsequently calculating the correlation coefficient between the corresponding wavelet template and the original data channel by channel and time window are relatively robust and have good noise resistance.

[0072] This invention presents a novel application of PCA in the generation of new wavelet templates. Within a template matching framework, it combines seismic wavelets with PCA to generate new feature templates (i.e., extracting wavelet templates corresponding to each trace from the reconstructed gather). Based on these new feature templates, a correlation coefficient metric is used to generate new correlation coefficients. Compared to conventional correlation coefficient methods, this method updates the selection of wavelet templates, generating multiple new wavelet templates containing multi-trace information using PCA. Using these templates for calculation significantly improves the robustness and accuracy of the correlation coefficient method. Since spatial information about first arrivals is utilized in the generation of new wavelet templates, the correlation coefficient attribute generated under the new wavelet template can be considered a spatial attribute, its physical meaning being the similarity between the signal within a certain time window in the original data and the new wavelet template corresponding to that seismic trace. As can be seen from the above process, this invention directly combines PCA with seismic wavelets during feature generation to produce new wavelet templates containing first arrival features. This is a novel approach, completely different from the conventional application of PCA for multi-feature selection after feature generation.

[0073] In step 4, for the output of step 3, find the maximum value point of the spatial attribute in each path, that is, the position of the initial arrival wave.

[0074] The following are several specific embodiments of the application of the present invention.

[0075] Example 1

[0076] In a specific embodiment 1 of the present invention, the first arrival identification method based on spatial attribute changes includes the following steps:

[0077] In step 1, input the seismic gathers from which the first arrivals need to be picked. The first arrival data is approximated as linear after inter-trace time difference elimination, using the wavelet convolution reflection coefficient. By default, the elevations of all traces are equal. Nonlinear inter-trace time differences are randomly applied to the data, and the elevations are calculated based on these nonlinear inter-trace time differences at a given surface velocity of 2000 m / s. Simultaneously, noise is applied to achieve a signal-to-noise ratio (SNR) of -3 dB (the SNR calculation formula is...). s represents the signal and n represents the noise, thus simulating the first-arrival picking of seismic gathers under complex terrain conditions with known elevations (drastic changes in inter-trace time difference and low signal-to-noise ratio). Figure 2 As shown, there is a time difference between the arrival times of the first arrivals. It can be seen that the time difference between the arrival times of the two traces is non-linear, and the geometric structure of the first arrival wave is also non-linear, exhibiting an undulating pattern. This data is used as the input seismic gather.

[0078] In step 2, inter-trace time difference elimination is performed based on a small smoothed reference surface. The surface elevation is statistically analyzed and smoothed to generate a smoothed surface elevation, which is then calculated using the formula... (elev smooth: Smoothed elevation surface; elev real : Actual ground elevation; v smooth (Smoothing speed) corrects the data from the real ground surface to a smooth elevation surface.

[0079] Δt This is the time difference correction amount. Based on the formula above, it is calculated using the actual elevation and smoothed elevation of each seismic trace. Thus, each seismic trace has a time difference correction amount, which is applied to the original data to correct for time difference; that is, each seismic trace is adjusted along the time axis. Δt Length translation allows data collected on a smooth elevation surface, eliminating arrival time jumps between tracks. For example, for... Figure 2 Based on the actual elevation and smoothed elevation in each trace head, the data shown allows for the calculation of an elevation-corrected time-shift for each seismic trace. Δt Based on this time shift, each path is processed on the time axis. Δt By shifting the length vertically, the nonlinear arrival time difference can be eliminated, aligning the axes in the same direction to achieve linearity, resulting in... Figure 3 The results are shown.

[0080] This makes the initial arrival (approximately) linear, thus enhancing the spatial correlation between data points. Figure 3 As shown, the inter-channel time difference abrupt change suppression method based on a small smooth reference surface is used for preprocessing to eliminate the inter-channel time difference and restore the first arrival wave to a (approximately) linear first arrival.

[0081] In step 3, firstly, the output of step 2 is used to select principal component reconstruction data using PCA, such as... Figure 4 As shown, the reconstructed trace set obtained by PCA reconstruction of the output trace set in step 2 is... Figure 3The trace gathers obtained by PCA reconstruction of (approximately) linear first arrivals are used as templates, with the wavelet extracted from the PCA-reconstructed trace gathers serving as templates. Since the first arrivals have been made approximately linear through inter-trace time difference elimination, a linear predictor (PCA in this case) can be used to reconstruct the data using statistically relevant information from multiple first arrival signals to generate wavelet templates. Then, for each seismic trace, the location of the first arrival is found in the PCA-reconstructed trace gather (the reconstructed trace gathers contain only principal components and are essentially noise-free, making it very easy to extract linear structures). Windows are opened above and below the time axis as wavelet templates, and the correlation coefficient between the wavelet template of each trace and the corresponding windowed data output from step 2 is calculated using correlation methods. For example, using a time window near the first arrival wave of the PCA reconstructed gather as the wavelet template for that gather, a time window is opened on the first gather of the output data in step 2, and the calculation is performed traversing that gather to obtain the spatial attributes corresponding to the first gather. Similarly, using a time window near the first arrival wave of the PCA reconstructed gather as the wavelet template for that gather, a time window is opened on the second gather of the output data in step 2, and the calculation is performed traversing that gather to obtain the spatial attributes corresponding to the second gather. This process is repeated for each gather to obtain the spatial attributes extracted from the output data in step 2. Figure 5 As shown, this is a schematic diagram of the spatial attributes extracted in step 3 based on the results of step 2. Figure 4 The wavelet template of each trace is extracted from the reconstructed gather. Figure 3 The spatial attribute variation map is obtained by calculating the data according to step three. The wavelet template is generated in this way. Due to the low template noise and the fact that each channel has a corresponding wavelet template, some of the spatial variations of the wavelet are preserved, resulting in good noise resistance and accuracy of the spatial attribute, and it still performs well under low signal-to-noise ratio conditions.

[0082] First arrival picking is not performed directly on the reconstructed PCA gather because the first arrival co-axial direction is not completely linear after time difference abrupt suppression, but approximately linear. The PCA reconstructed gather will linearize the first arrival co-axial direction, such as... Figure 3 After time difference elimination, the collection of channels and Figure 4 Comparison of reconstructed gathers. If the first arrival is extracted directly from the PCA reconstructed gather, it deviates from the first arrival of the original data. Therefore, the method of this patent only extracts the wavelet template from the PCA reconstructed gather.

[0083] In step 4, for the output of step 3, find the maximum value point of the spatial attribute in each path, i.e., the position of the initial arrival wave, such as... Figure 6 and Figure 7 As shown, Figure 6 The initial arrival results identified within the attribute space (the black line represents the initial arrival identification results). Figure 7 The results of the first arrival identified by utilizing changes in spatial attributes are displayed on the data after the inter-track time difference jump is suppressed (the black line represents the first arrival identification results).

[0084] Example 2

[0085] In a specific embodiment 2 of the present invention, the first arrival identification method based on spatial attribute changes includes the following steps:

[0086] In step 1, input the seismic gathers from which first arrivals need to be picked. Here, actual seismic gathers from complex terrains (with drastic variations in inter-trace time differences and low signal-to-noise ratios) are used as the input, such as... Figure 8 As shown, the actual data input has a non-linear time difference between the arrival times of traces, and the geometric structure of the first arrival wave is also non-linear, exhibiting an undulating pattern, corresponding to the seismic gathers that need to be picked up for the first arrival in step 1.

[0087] In step 2, inter-trace time difference elimination is performed based on a small smoothed reference surface. The surface elevation is statistically analyzed and smoothed to generate a smoothed surface elevation, which is then calculated using the formula... (elev smooth : Smoothed elevation surface; elev real : Actual ground elevation; v smooth (Smoothing speed) corrects the data from the real ground surface to a smooth elevation surface.

[0088] This makes the initial arrival (approximately) linear, thus enhancing the spatial correlation between data points. Figure 9 As shown, the results of suppressing the inter-channel time difference jump are similar to those in Example 1. Figure 3 Correspondingly, after the inter-channel time difference jump is suppressed, the nonlinear first-arrival structure becomes (approximately) linear.

[0089] In step 3, firstly, the output of step 2 is used to select principal component reconstruction data using PCA, such as... Figure 10 As shown, for Figure 9The actual data shown is reconstructed using PCA to generate a trace gather. Wavelet templates can be extracted from this gather, and the wavelets extracted from the PCA-reconstructed gather are used as templates. Since the first arrivals have been made approximately linear through inter-trace time difference elimination, a linear predictor (PCA in this case) can be used to reconstruct the data using statistically relevant information from multiple first arrival signals to generate wavelet templates. Then, for each seismic trace, the location of the first arrival is found in the PCA-reconstructed gather (the reconstructed gather contains only principal components and is essentially noise-free, making it very easy to extract linear structures). Windows are opened above and below the time axis as wavelet templates, and the correlation coefficient between the wavelet template of each trace and the corresponding windowed data output from step 2 is calculated using correlation methods. For example, using a time window near the first arrival wave of the PCA reconstructed gather as the wavelet template for that gather, a time window is opened on the first gather of the output data in step 2, and the calculation is performed traversing that gather to obtain the spatial attributes corresponding to the first gather. Similarly, using a time window near the first arrival wave of the PCA reconstructed gather as the wavelet template for that gather, a time window is opened on the second gather of the output data in step 2, and the calculation is performed traversing that gather to obtain the spatial attributes corresponding to the second gather. This process is repeated for each gather to obtain the spatial attributes extracted from the output data in step 2. Figure 11 The diagram shows the extracted spatial attributes. Generating wavelet templates in this manner results in low template noise and preserves some spatial variations of the wavelets, giving the spatial attributes good noise resistance and accuracy, even under low signal-to-noise ratio conditions.

[0090] First arrival picking is not performed directly on the reconstructed PCA gather because the first arrival co-axial direction is not completely linear after time difference abrupt suppression, but approximately linear. The PCA reconstructed gather will linearize the first arrival co-axial direction, such as... Figure 9 After time difference elimination, the collection of channels and Figure 10 Comparison of reconstructed gathers. If the first arrival is extracted directly from the PCA reconstructed gather, it deviates from the first arrival of the original data. Therefore, the method of this patent only extracts the wavelet template from the PCA reconstructed gather.

[0091] In step 4, for the output of step 3, find the maximum value point of the spatial attribute in each path, i.e., the position of the initial arrival wave, such as... Figure 12 and Figure 13 As shown, Figure 12 To identify the initial arrival result within the attribute space (the black line represents the initial arrival identification result), Figure 13 This is a schematic diagram showing the first arrival results displayed on the data after the inter-track time difference jump is suppressed (the black line represents the first arrival identification result).

[0092] Example 3

[0093] In a specific embodiment 3 of the present invention, the first arrival identification method based on spatial attribute changes includes the following steps:

[0094] In step 1, input the seismic gathers from which first arrivals need to be picked. Here, actual seismic gathers from complex terrains (with drastic variations in inter-trace time differences and low signal-to-noise ratios) are used as the input, such as... Figure 14 As shown, the actual input data has a non-linear time difference between traces, and the geometry of the first arrival wave is also non-linear, exhibiting an undulating pattern. Corresponding to the seismic gathers that need to be picked up for the first arrival in step 1, it can be seen that the data contains strong noise and weak first arrivals with long offsets.

[0095] In step 2, inter-trace time difference elimination is performed based on a small smoothed reference surface. The surface elevation is statistically analyzed and smoothed to generate a smoothed surface elevation, which is then calculated using the formula... (elev smooth : Smoothed elevation surface; elev real : Actual ground elevation; v smooth (Smoothing speed) corrects the data from the real ground surface to a smooth elevation surface.

[0096] This makes the initial arrival (approximately) linear, thus enhancing the spatial correlation between data points. Figure 15 As shown, the results of suppressing the inter-channel time difference jump are similar to those in Example 1. Figure 3 In Example 2 Figure 9 Correspondingly, after the inter-channel time difference jump is suppressed, the nonlinear first-arrival structure becomes (approximately) linear.

[0097] In step 3, firstly, the output of step 2 is used to select principal component reconstruction data using PCA, such as... Figure 16 As shown, for Figure 15The actual data shown is reconstructed using PCA to generate a trace gather. Wavelet templates can be extracted from this gather, and the wavelets extracted from the PCA-reconstructed gather are used as templates. Since the first arrivals have been made approximately linear through inter-trace time difference elimination, a linear predictor (PCA in this case) can be used to reconstruct the data using statistically relevant information from multiple first arrival signals to generate wavelet templates. Then, for each seismic trace, the location of the first arrival is found in the PCA-reconstructed gather (the reconstructed gather contains only principal components and is essentially noise-free, making it very easy to extract linear structures). Windows are opened above and below the time axis as wavelet templates, and the correlation coefficient between the wavelet template of each trace and the corresponding windowed data output from step 2 is calculated using correlation methods. For example, using a time window near the first arrival wave of the PCA reconstructed gather as the wavelet template for that gather, a time window is opened on the first gather of the output data in step 2, and the calculation is performed traversing that gather to obtain the spatial attributes corresponding to the first gather. Similarly, using a time window near the first arrival wave of the PCA reconstructed gather as the wavelet template for that gather, a time window is opened on the second gather of the output data in step 2, and the calculation is performed traversing that gather to obtain the spatial attributes corresponding to the second gather. This process is repeated for each gather to obtain the spatial attributes extracted from the output data in step 2. Figure 17 The diagram shows the extracted spatial attributes. Generating wavelet templates in this manner results in low template noise and preserves some spatial variations of the wavelets, giving the spatial attributes good noise resistance and accuracy, even under low signal-to-noise ratio conditions.

[0098] First arrival picking is not performed directly on the reconstructed PCA gather because the first arrival co-axial direction is not completely linear after time difference abrupt suppression, but approximately linear. The PCA reconstructed gather will linearize the first arrival co-axial direction, such as... Figure 15 After time difference elimination, the collection of channels and Figure 16 Comparison of reconstructed gathers. If the first arrival is extracted directly from the PCA reconstructed gather, it deviates from the first arrival of the original data. Therefore, the method of this patent only extracts the wavelet template from the PCA reconstructed gather.

[0099] In step 4, for the output of step 3, find the maximum value point of the spatial attribute in each path, i.e., the position of the initial arrival wave, such as... Figure 18 and Figure 19 As shown. Figure 18 This is a schematic diagram of identifying the initial arrival result within the attribute space (the black line represents the initial arrival identification result); Figure 19 This diagram illustrates the identification of first arrival results displayed on the data after suppression of inter-track time difference jumps (the black line represents the first arrival identification result).

[0100] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0101] Except for the technical features described in the specification, all other technologies are known to those skilled in the art.

Claims

1. A method for identifying first arrivals based on spatial attribute changes, characterized in that, This first arrival identification method based on spatial attribute changes includes: Step 1: Input the seismic gathers from which the first arrival needs to be picked; Step 2: Perform inter-track time difference elimination based on a small smoothing reference surface; Step 3: Use the PCA method to reconstruct the gather using the spatial correlation of the gather, extract the wavelet template from the PCA reconstructed gather, and use the correlation coefficient method to calculate the spatial attributes of each point in the gather; Step 4: Identify the first arrival wave based on changes in spatial properties.

2. The first arrival identification method based on spatial attribute changes according to claim 1, characterized in that, In step 1, the wavelet convolution reflection coefficient is used as the linear first arrival data after eliminating inter-trace time difference. By default, the elevations of each trace are equal. Nonlinear inter-trace time difference is randomly applied to the data. When a certain surface velocity is given, the elevation is calculated based on the nonlinear inter-trace time difference. At the same time, noise is applied to simulate the seismic gathers that need to be picked up under the complex surface with known surface elevation. This data is used as the input seismic gather.

3. The first arrival identification method based on spatial attribute changes according to claim 2, characterized in that, In step 1, given a surface velocity of 2000 m / s, the elevation is calculated based on the nonlinear inter-channel time difference, while simultaneously applying noise to achieve a signal-to-noise ratio of -3 dB. The signal-to-noise ratio calculation formula is as follows: s represents the signal and n represents the noise. This is used to simulate the seismic gathers that need to be picked up under the complex surface with known surface elevation. This data is used as the input seismic gather.

4. The first arrival identification method based on spatial attribute changes according to claim 1, characterized in that, In step 2, the surface elevation is statistically analyzed and smoothed to generate a smoothed surface elevation. Then, the data is corrected from the real surface to the smoothed elevation surface using a formula, so that the initial elevation becomes linear and the spatial correlation between the data is enhanced.

5. The first arrival identification method based on spatial attribute changes according to claim 4, characterized in that, In step 2, the formula for correcting the data from the true ground surface to a smooth elevation surface is: Among them, elev smooth : Smoothed elevation surface; elev real : Actual ground elevation; v smooth : Smoothing speed, Δt is the time difference correction amount; According to this formula, the actual elevation and smoothed elevation of each seismic trace are used for calculation, so that each seismic trace has a time difference correction amount. Applying it to the original data to correct the time difference, that is, shifting each seismic trace by a length of Δt on the time axis, we can obtain the data collected on the smoothed elevation surface, thus eliminating the arrival time jump between traces.

6. The first arrival identification method based on spatial attribute changes according to claim 1, characterized in that, In step 3, spatial attributes are extracted, a wavelet template is selected, and then the wavelet template is matched to obtain attributes that characterize the similarity between the wavelet template and the wavelet template.

7. The first arrival identification method based on spatial attribute changes according to claim 6, characterized in that, In step 3, firstly, the principal component reconstruction data is selected using PCA on the output results of step 2, and the wavelet extracted from the PCA-reconstructed gather is used as the template. Then, for each seismic trace, the position of the first arrival wave of each trace is found on the PCA-reconstructed gather, and windows are opened above and below the time axis as wavelet templates. The correlation coefficient between the wavelet template of each trace and the corresponding window of the output data in step 2 is calculated using the correlation method.

8. The first arrival identification method based on spatial attribute changes according to claim 7, characterized in that, In step 3, the formula for calculating the correlation coefficient is: Where w represents the length of the time window. a Represents the wavelet template. ai This represents the amplitude value at each sampling point of the wavelet template; b This represents the signal along the corresponding seismic trace and the wavelet template window length in the original data. bi The amplitude value represents the signal at each sampling point; the correlation coefficient calculation formula characterizes the similarity between the windowed data on the corresponding seismic trace and the wavelet template of that trace; the greater the similarity, the more likely it is to be a first arrival; this formula characterizes the similarity between two time series of equal length, ranging from 0 to 1, with the closer to 1 indicating a greater similarity between the two time series; the larger the value of this formula, the more similar the windowed time series on the original data is to the wavelet template on the reconstructed trace, that is, the greater the probability that the center point of the time series on the original data is a first arrival.

9. The first arrival identification method based on spatial attribute changes according to claim 7, characterized in that, In step 3, a time window is opened near the first arrival wave of the PCA reconstructed gather as the wavelet template for that gather. The time window is opened on the first gather of the output data in step 2 and the gather is traversed to calculate the spatial attributes corresponding to the first gather. A time window is opened near the first arrival wave of the PCA reconstructed gather as the wavelet template for that gather. The time window is opened on the second gather of the output data in step 2 and the gather is traversed to calculate the spatial attributes corresponding to the second gather. The spatial attributes extracted from the output data in step 2 are calculated one gather at a time.

10. The first arrival identification method based on spatial attribute changes according to claim 1, characterized in that, In step 4, for the output of step 3, find the maximum value point of the spatial attribute in each path, that is, the position of the initial wave.

11. A first arrival recognition system based on spatial attribute changes, characterized in that, The first arrival identification system based on spatial attribute changes uses the first arrival identification method based on spatial attribute changes as described in any one of claims 1-10 to identify first arrival waves under complex surface conditions.

Images