Similarity measurement method and system based on dynamic waveform extraction and elastic alignment

Abstract

The application provides a similarity measurement method and system based on dynamic waveform extraction and elastic alignment, and relates to the technical field of data processing.The method comprises the following steps: taking a candidate wave peak position on a main seismic trace as a center, adaptively determining left and right boundaries, and extracting a reference waveform segment; traversing each offset within a preset offset range for the candidate wave peak position on a neighboring seismic trace, and extracting a candidate waveform segment; performing weighted fusion on the candidate waveform segment and the reference waveform segment from shape, local change rate, amplitude intensity and time domain span characteristics; and taking the highest similarity as a similarity matching result.The technical problem of poor identification accuracy and stability of the same phase axis caused by inaccurate alignment and single characteristics in the prior art seismic waveform fixed window matching is solved.The multi-dimensional feature fusion similarity calculation through dynamic waveform adaptive interception and elastic offset alignment improves the technical effects of seismic wave same phase axis matching accuracy and reliability.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and specifically to a similarity measurement method and system based on dynamic waveform extraction and flexible alignment. Background Technology

[0002] Automatic seismic phase axis tracking is a fundamental task in seismic data interpretation. Its core lies in automatically extending from manually labeled seed points within a 3D seismic data volume to adjacent seismic traces, thereby determining the corresponding positions of the same stratum reflection interface on different seismic traces. In existing technologies, the method for determining whether two wave peaks on two seismic traces belong to the same phase axis, i.e., whether they match, mainly relies on waveform similarity calculation.

[0003] However, existing waveform similarity calculation methods mostly employ fixed-length time windows to truncate waveforms and then calculate similarity based on these truncated segments. But because seismic wavelets are affected by factors such as frequency variations, propagation attenuation, and formation absorption, their actual width and shape exhibit significant dynamic changes. Fixed windows easily lead to incomplete waveform truncation or the introduction of interference from adjacent reflections, thus reducing matching reliability. Simultaneously, due to factors such as formation dip angle variations and tectonic undulations, the same phase axis often exhibits natural time shifts on different seismic traces. Existing methods typically compare waveforms at fixed locations directly, lacking an effective peak alignment mechanism, resulting in an underestimation of similarity under true correspondence. Furthermore, existing methods often rely on single indicators such as Pearson correlation coefficients or normalized cross-correlation for similarity evaluation, making it difficult to simultaneously characterize differences in waveform amplitude, structure, and width across multiple dimensions, leading to insufficient discrimination capability under complex geological conditions. Moreover, because seismic wavelets typically possess asymmetric characteristics, and traditional methods ignore the energy distribution differences on both sides of the peak, mismatches are easily generated when absorption attenuation effects are present.

[0004] Existing technologies for fixed-window matching of seismic waveforms suffer from inaccurate alignment and limited features, resulting in poor accuracy and stability in identifying in-phase axes. Summary of the Invention

[0005] The purpose of this application is to provide a similarity measurement method and system based on dynamic waveform extraction and elastic alignment, which can solve the technical problems of inaccurate alignment and single feature in existing seismic waveform fixed window matching, resulting in poor accuracy and stability of phase axis identification.

[0006] In view of the above problems, this application provides a similarity measurement method and system based on dynamic waveform extraction and flexible alignment.

[0007] The first aspect of this application provides a similarity measurement method based on dynamic waveform extraction and elastic alignment. This method includes: adaptively determining the left and right boundaries of the main seismic trace based on the candidate peak position on the main seismic trace, according to priority-based termination conditions, and extracting a reference waveform segment from the main seismic trace; traversing each offset within a preset offset range for candidate peak positions on adjacent seismic traces, and extracting candidate waveform segments at each offset; weighted fusing the candidate waveform segments and reference waveform segments from four dimensions: shape features, local rate of change features, amplitude intensity features, and temporal span features, to obtain a comprehensive similarity at the corresponding offset; and taking the highest similarity among all offsets as the similarity matching result between the adjacent seismic trace and the main seismic trace at the corresponding peak based on the comprehensive similarity at each offset.

[0008] Optionally, the first priority is when the absolute value of the amplitude of the current sampling point drops below a preset percentage of the peak amplitude of the candidate wave peak; the second priority is when the current sampling point is a local minimum and satisfies that the amplitude of the current sampling point has the opposite sign to the amplitude of the candidate wave peak, or the absolute value of the amplitude of the current sampling point is less than a preset minimum percentage of the peak amplitude; the third priority is when the distance of the current sampling point from the starting center has reached a preset maximum search radius.

[0009] Optionally, starting from the candidate peak position as the center point, the exploration proceeds outward point by point to the left and right sides respectively. During the exploration on each side, the termination condition is determined sequentially according to the first priority to the third priority. When any priority termination condition is met, the current side exploration is terminated, and the final boundary of the current side is extended outward by a preset number of additional sampling points to determine the left and right boundaries. Based on the left and right boundaries, continuous sampling points are extracted from the main seismic trace sampling sequence to obtain a reference waveform segment.

[0010] Optionally, the preset percentage is 30%, and the preset minimum percentage is 15%.

[0011] Optionally, the preset offset range is a sampling point, including a negative sampling point, a zero offset, and a positive sampling point. Taking the candidate wave peak position on the adjacent seismic trace as the center, the preset offset range is offset to the left and to the right by a preset maximum offset amount to form a set of continuous integer offsets, including the zero offset. Each offset in the set of continuous integer offsets is traversed, and the offset position is used as a temporary center point to extract candidate waveform segments under each offset. The extraction of the candidate waveform segments is also based on priority sorting to determine the left and right boundaries, and the candidate waveform segments are truncated based on the determined left and right boundaries.

[0012] Optionally, the number of sampling points of the candidate waveform segment and the reference waveform segment are compared. When the number of sampling points of the extracted candidate waveform segment is different from that of the reference waveform segment, the candidate waveform segment is resampled to the same number of points as the reference waveform segment by linear interpolation before similarity calculation is performed.

[0013] Optionally, the reference waveform segment and the candidate waveform segment are standardized with zero mean and unit variance, so that the mean of the two sequences is zero and the standard deviation is one. The Pearson correlation coefficient of the two standardized sequences is calculated, and the normalized contour difference between the standardized sequences is also calculated. Based on the Pearson correlation coefficient and the normalized contour difference, a shape similarity score is calculated. The first-order difference sequences of the reference waveform segment and the candidate waveform segment are calculated respectively, and a local change similarity score is obtained based on the first-order difference similarity. The maximum amplitude of the reference waveform segment is extracted as the reference peak value, and the maximum amplitude of the candidate waveform segment is extracted as the candidate peak value. The amplitude similarity score is obtained by comparing the reference peak value and the candidate peak value. The number of sampling points of the reference waveform segment and the candidate waveform segment are counted respectively, and the smaller value and the larger value are compared and analyzed to obtain a time span similarity score. According to the configured weights of each dimension, the shape similarity score, local change similarity score, amplitude similarity score, and time span similarity score are weighted and fused to obtain a comprehensive similarity score under the corresponding offset.

[0014] Optionally, the energy distribution difference between the reference waveform segment and the candidate waveform segment on the left and right sides with their respective peak positions as the dividing line is calculated to obtain their respective asymmetry indices; the absolute value of the difference between the asymmetry indices is calculated, and the difference between 1 and the product of the absolute value and the preset positive coefficient is obtained to obtain a penalty factor between 0 and 1, wherein if the result is negative, it is taken as 0; the weighted fusion obtained comprehensive similarity is multiplied by the penalty factor to obtain the final comprehensive similarity under the corresponding offset.

[0015] Optionally, when traversing each offset within the preset offset range, for each candidate waveform segment extracted under each offset, the ratio of the number of sampling points to the number of sampling points of the reference waveform segment is calculated; if the ratio exceeds the preset length ratio threshold, the corresponding offset is skipped directly, the comprehensive similarity under this offset is not calculated, and the comprehensive similarity of the corresponding offset is recorded as zero, wherein the preset length ratio threshold is 2.0.

[0016] The second aspect of this application provides a similarity measurement system based on dynamic waveform extraction and elastic alignment. This system includes: a boundary determination module, used to adaptively determine the left and right boundaries based on priority-based termination conditions, centered on the candidate peak positions on the main seismic trace, and extract reference waveform segments from the main seismic trace; an offset traversal module, used to traverse each offset within a preset offset range for the candidate peak positions on adjacent seismic traces, and extract candidate waveform segments at each offset; a segment fusion module, used to perform weighted fusion of the candidate waveform segments and reference waveform segments from four dimensions: shape features, local rate of change features, amplitude intensity features, and temporal span features, to obtain a comprehensive similarity at the corresponding offset; and a matching result acquisition module, used to take the highest similarity among all offsets as the similarity matching result between the adjacent seismic trace and the main seismic trace at the corresponding peak, based on the comprehensive similarity at each offset.

[0017] One or more technical solutions provided in this application have at least the following technical effects or advantages: By adaptively determining the left and right boundaries of the main seismic trace using candidate wave crest positions as the center and prioritizing termination conditions, reference waveform segments of the main seismic trace are extracted. For candidate wave crest positions on adjacent seismic traces, each offset is traversed within a preset offset range, and candidate waveform segments are extracted at each offset. The candidate waveform segments and reference waveform segments are weighted and fused from four dimensions: shape features, local rate of change features, amplitude intensity features, and temporal span features, to obtain the comprehensive similarity at the corresponding offset. Based on the comprehensive similarity at each offset, the highest similarity among all offsets is taken as the similarity matching result between the adjacent seismic trace and the main seismic trace at the corresponding wave crest. This achieves the technical effect of improving the accuracy and reliability of seismic wave phase axis matching through multi-dimensional feature fusion similarity calculation by dynamically adaptively truncating waveforms and aligning with elastic offset.

[0018] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 A flowchart illustrating the similarity measurement method based on dynamic waveform extraction and elastic alignment provided in this application.

[0021] Figure 2 The waveform diagram of the dynamic wave packet extraction process in the similarity measurement method based on dynamic waveform extraction and elastic alignment provided in this application.

[0022] Figure 3 The waveforms of fixed window and dynamic wave packet extraction are compared in the similarity measurement method based on dynamic waveform extraction and elastic alignment provided in this application.

[0023] Figure 4 The waveform diagram of the elastic alignment process in the similarity measurement method based on dynamic waveform extraction and elastic alignment provided in this application. Figure 5 A schematic diagram of the similarity measurement system based on dynamic waveform extraction and elastic alignment provided in this application.

[0024] Figure labeling: Boundary determination module 11, offset traversal module 12, fragment fusion module 13, matching result acquisition module 14. Detailed Implementation

[0025] This application provides a similarity measurement method and system based on dynamic waveform extraction and elastic alignment, which addresses the technical problems of inaccurate alignment and single feature in existing seismic waveform fixed-window matching, resulting in poor accuracy and stability of phase axis identification. It achieves the technical effect of improving the accuracy and reliability of seismic wave phase axis matching through multi-dimensional feature fusion similarity calculation via adaptive dynamic waveform truncation and elastic offset alignment.

[0026] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. It should also be noted that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, not all of them.

[0027] Example 1, as Figure 1As shown, this application provides a similarity measurement method based on dynamic waveform extraction and flexible alignment, which includes: Centered on the candidate wave peak position on the main seismic trace, the left and right boundaries are adaptively determined according to the termination conditions arranged by priority, and reference waveform segments of the main seismic trace are extracted.

[0028] Furthermore, the termination conditions for priority ranking include: the first priority is when the absolute value of the amplitude of the current sampling point drops below a preset percentage of the peak amplitude of the candidate wave peak; the second priority is when the current sampling point is a local minimum and satisfies the condition that the amplitude of the current sampling point has the opposite sign to the amplitude of the candidate wave peak, or the absolute value of the amplitude of the current sampling point is less than a preset minimum percentage of the peak amplitude; the third priority is when the distance of the current sampling point from the starting center has reached a preset maximum search radius.

[0029] Furthermore, the preset percentage is 30%, and the preset minimum percentage is 15%.

[0030] Specifically, the candidate peak position detected in the main seismic trace sampling sequence is used as the center point. This candidate peak originates from the local maximum amplitude point obtained after extreme value detection or energy peak detection of the seismic trace signal, and is denoted as the peak amplitude Ap. Using this amplitude point as the center point, i.e., using this center point as the starting index, the sampling is expanded point by point to the left and right according to the sampling sequence number. The sampling interval is determined by the seismic data acquisition system, usually a fixed time sampling interval, corresponding to a discrete sampling point spacing of 1 sampling point. During the left and right expansion process, priority termination conditions are introduced to dynamically determine the left and right boundaries. The priority termination conditions are determined sequentially from strong constraints to weak constraints, including a first priority termination condition, a second priority termination condition, and a third priority termination condition, thereby avoiding premature truncation or over-expansion. During each step of the expansion process, the absolute value of the amplitude of the current sampling point is obtained and compared with the priority termination condition.

[0031] The first priority termination condition is as follows: when the absolute value of the amplitude of the current sampling point, |A(i)|, drops below a preset percentage of the peak amplitude of the candidate wave, i.e., when |A(i)| < α × |Ap|, where α is a preset percentage, preferably 30%, the exploration is immediately terminated. The first priority termination condition indicates that when the energy of the seismic reflected wave decays to a significantly low level of the main event energy, it is considered that the point is close to the tail boundary of the effective waveform, and subsequent sampling points have significantly deviated from the main body of the effective waveform. Continued expansion will introduce adjacent lobes or noise. The second priority termination condition is used to supplement the situation where the waveform structure changes abruptly but the first condition is not triggered. First, it is detected whether the current sampling point is a local minimum, i.e., it is determined that the amplitude of the point is less than that of its adjacent sampling points. When it is less, it indicates that the current sampling point is a local minimum. Then, the sign of the amplitude of the current sampling point is compared with that of the candidate wave peak amplitude. If the sign of the amplitude of the current sampling point is opposite to that of the candidate wave peak amplitude, i.e., a phase reversal occurs, or when the absolute value of the amplitude of the current sampling point is less than a preset minimum percentage of the peak amplitude, preferably 15%, the exploration is terminated. The third priority termination condition is the search distance constraint, i.e., the distance between the current sampling point and the center point. The search stops when the number of samples between them exceeds the preset maximum search radius. The preset maximum search radius is determined based on the dominant frequency and sampling rate of the seismic data, and is the number of sampling points, such as 30-50 sampling points. By using the candidate wave crest position on the main seismic trace as the center, the left and right boundaries are adaptively determined according to the priority-arranged termination conditions, thereby extracting the reference waveform segment of the main seismic trace.

[0032] By employing a priority-based termination mechanism, adaptive boundary identification of seismic reflection events is achieved. Compared to the fixed-window method, this effectively avoids truncating valid waveforms or introducing irrelevant noise, improving the stability and physical consistency of subsequent similarity calculations, thereby enhancing the accuracy and reliability of similarity matching.

[0033] Furthermore, the reference waveform segment of the main seismic trace is extracted, including: starting from the candidate peak position as the center point, exploring outwards point by point to the left and right sides respectively; during the exploration of each side, judging whether the termination condition is met in the order of first priority to third priority; when any priority termination condition is met, the exploration of the current side is terminated, and the final boundary of the current side is extended outwards by a preset number of additional sampling points to determine the left and right boundaries; based on the left and right boundaries, continuous sampling points are extracted from the sampling sequence of the main seismic trace to obtain the reference waveform segment.

[0034] Specifically, the candidate peak positions detected in the main seismic trace sampling sequence are used as the center point. The system then explores outwards point by point to the left and right, visiting adjacent sample values ​​in descending (left) or ascending (right) order of sampling index. During expansion on each side, termination conditions are checked sequentially according to priority levels from first to third. First, the first priority termination condition is checked. If the first priority condition is not met, the second priority condition is checked. If neither of these conditions is triggered, the third priority termination condition is checked. If any priority termination condition is triggered on a side, expansion in that direction is immediately stopped, and the current stopping point is recorded as the initial boundary position. Then, a preset number of additional sampling points are expanded outwards from the initial boundary position. This preset number of additional sampling points is set empirically or obtained through data resolution calibration, for example, ±2 sampling points. This is used for fine-tuning and compensation of the boundaries, determining the final left and right boundaries, and making the waveform segment more complete.

[0035] After determining the left and right boundaries, a continuous sequence of sampling points within this interval is extracted from the main seismic trace sampling sequence to form a reference waveform segment. The reference waveform segment represents a complete local reflection event centered on the candidate wave peak. Figure 2 The complete process of dynamic wave packet extraction centered on the wave peak (sampling point 130, amplitude approximately 1.25) is demonstrated. This process uses multiple priority conditions to explore outward point by point, automatically determine the waveform boundary and extract the effective wave packet.

[0036] like Figure 3 As shown, to verify the adaptive advantage of the dynamic wave packet extraction method, it is compared with the traditional fixed window truncation method: Figure (a) shows the limitations of the traditional fixed window truncation method: with the peak (sampling point 130) as the center, a symmetrical window with a fixed ±18 sampling points (covering sampling points 112 to 148) is used, which results in the window being too wide. The left side contains a large amount of noise base that is not related to the wave packet (approximately sampling points 112 to 120), while the right side truncates the tail of the waveform (approximately sampling points 138 to 148), causing multiple problems such as the window being too wide, loss of side lobes, and truncation of the waveform, and it cannot adapt to changes in different waveform scales.

[0037] Figure (b) illustrates the advantages of this method's dynamic wave packet extraction: Centered on the wave crest, the waveform boundary is automatically explored using multiple priority conditions, determining the left boundary as sampling point 120 (wave trough position) and the right boundary as sampling point 130 (wave trough position). The extraction window is [120, 130], with a length of only 11 sampling points. This window accurately covers the entire waveform (from the left wave trough through the wave crest to the right wave trough), avoiding both noise intrusion caused by an excessively wide fixed window and waveform truncation caused by an excessively narrow window. This achieves adaptive determination of the waveform boundary, ensuring the integrity and purity of the extracted wave packets.

[0038] For example, a candidate peak with a peak amplitude Ap = 12 is detected at sampling point i = 200 in the main seismic trace. The first priority threshold is 3.6 (30%), and the second priority threshold is 1.8 (15%). When exploring to the right, the amplitude of sampling point 201 is 10, the amplitude of sampling point 202 is 8, the amplitude of sampling point 203 is 5, the amplitude of sampling point 204 is 3.5, and the amplitude of sampling point 205 is 2.9. At this point, |A(i)| < 3.6 is satisfied for the first time at sampling point 205, thus triggering the first priority termination, and the initial boundary on the right is point 205. When exploring to the left, the amplitude of sampling point 199 is 9, the amplitude of sampling point 198 is 7, the amplitude of sampling point 197 is 4, and the amplitude of sampling point 196 is -2. It can be seen that the second priority termination condition is triggered at sampling point 196, and the initial boundary on the left is point 196. If the preset additional sampling point number is 2, the final left boundary will extend to 194 points and the right boundary will extend to 207 points. Therefore, the reference waveform segment is a continuous sampling sequence between 195 and 207 points.

[0039] By combining the termination conditions with priority ordering and the boundary compensation mechanism, the effective time window of seismic reflection events can be adaptively extracted, thereby avoiding the problems of incomplete waveform truncation or the introduction of irrelevant noise commonly found in traditional fixed window methods, and improving the stability and matching accuracy of waveform similarity calculation.

[0040] For candidate wave peak locations on adjacent seismic traces, traverse each offset within a preset offset range and extract candidate waveform segments at each offset.

[0041] Furthermore, for candidate wave peak positions on adjacent seismic traces, each offset is traversed within a preset offset range, and candidate waveform segments are extracted at each offset. This includes: the preset offset range is a sampling point, containing one negative sampling point, zero offset, and one positive sampling point; with the candidate wave peak position on the adjacent seismic trace as the center, a set of continuous integer offsets, including zero offset, is formed by offsetting to the left and right by the preset maximum offset amount within the preset offset range; each offset in the set of continuous integer offsets is traversed, and the offset position is used as a temporary center point to extract candidate waveform segments at each offset; wherein, the extraction of the candidate waveform segments is also based on priority sorting to determine the left and right boundaries, and the candidate waveform segments are truncated based on the determined left and right boundaries.

[0042] Specifically, a preset offset range is set to simulate the slight phase drift or picking errors that may exist in adjacent seismic traces. This preset offset range is one sampling point, including one negative sampling point, zero offset, and one positive sampling point. Candidate peak positions on adjacent seismic traces are determined through extreme value detection or energy peak detection. Adjacent seismic traces refer to data acquired from seismic traces spatially adjacent to the main seismic trace, with sampling intervals consistent with the main seismic trace. Time synchronization is ensured by a unified acquisition device, thereby ensuring time reference consistency for cross-trace comparisons. Then, using the candidate peak position on the adjacent seismic trace as the center, a set of continuous integer offsets, including zero offset, is formed by shifting the candidate peak index position to the left and right by the preset maximum offset amounts of -1, 0, and +1. The offset represents an integer-level translation of the candidate peak index position on the time axis; negative values ​​indicate movement towards the previous sampling point, positive values ​​indicate movement towards the next sampling point, and zero offset corresponds to no adjustment to the peak position.

[0043] The algorithm iterates through each offset in the set of consecutive integer offsets. For each offset, the offset position is used as a temporary center point, and this temporary center point is used as the new waveform extraction center. The algorithm then explores outwards point by point to the left and right, checking for termination conditions in the order of first priority, second priority, and third priority. If any condition is met, the current side exploration is terminated, and the termination boundary is extended outwards by a preset number of additional sampling points to determine the left and right boundaries. Based on these boundaries, continuous sampling points are extracted from the sampling sequences of adjacent seismic traces to form candidate waveform segments for each offset. This ensures that waveform segments extracted from different offsets are generated based on the same rule system, thereby guaranteeing consistency and comparability in cross-offset comparisons. Figure 4 As shown, the flexible alignment process finds the optimal matching position through offset search. Figure 4Each subplot shows a waveform comparison at a specific offset: the thick black line represents the candidate waveform segment extracted from the neighboring seismic trace at the current offset, and the thin gray line represents the main trace reference waveform. The specific search process is as follows: taking the candidate wave peak position of the neighboring seismic trace (sampling point 133) as the center, within the preset offset range [-1, 0, +1], it expands to the left and right to the maximum offset, forming a set of continuous integer offsets including zero offset (-3, -2, -1, 0, +1, +2, +3). For each offset, the offset position is used as a temporary center point, and the left and right boundaries are determined according to a unified priority rule, and equal-length waveform segments are truncated, and then the similarity score between the candidate waveform and the main trace reference waveform is calculated. From the scores marked in each subplot, it can be seen that: when the offset is -3 and +3, the phase deviation is the largest and the score is the lowest (0.773); when the offset converges towards the center, the matching degree gradually increases; the scores are the highest when the offset is -1, 0, and +1 (all 0.997), among which the offset -1 is determined to be the optimal solution. The final elastic alignment result is as follows: the position of the adjacent seismic trace wave crest is shifted to the left by 1 sampling point, achieving the best match with the main trace reference waveform, with an optimal correlation coefficient of 0.997.

[0044] By introducing small-range discrete migration search in adjacent seismic traces and combining it with a dynamic waveform segment extraction mechanism, we can achieve multi-position elastic alignment of seismic reflection events on the time axis. This effectively compensates for small phase shifts caused by picking errors, velocity changes, or differences in inter-trace propagation, thereby improving the accuracy and reliability of similarity calculation for matching the same geological event.

[0045] The candidate waveform segment and the reference waveform segment are weighted and fused from four dimensions: shape features, local rate of change features, amplitude intensity features, and time span features, to obtain the comprehensive similarity under the corresponding offset.

[0046] Furthermore, before performing weighted fusion on the candidate waveform segment and the reference waveform segment to obtain the comprehensive similarity under the corresponding offset, the method further includes: comparing the number of sampling points between the candidate waveform segment and the reference waveform segment; when the number of sampling points of the extracted candidate waveform segment and the reference waveform segment is different, the candidate waveform segment is resampled to the same number of points as the reference waveform segment by linear interpolation before similarity calculation is performed.

[0047] Specifically, before calculating the comprehensive similarity between the candidate waveform segment and the reference waveform segment, the number of sampling points of the candidate waveform segment and the reference waveform segment are compared. When the number of sampling points of the candidate waveform segment is the same as that of the reference waveform segment, the comprehensive similarity calculation is performed directly. When the number of sampling points of the candidate waveform segment is different from that of the reference waveform segment, the reference waveform segment is used as the reference sequence, and the candidate waveform segment is resampled to the same number of points as the reference waveform segment through linear interpolation, thereby realizing the length normalization preprocessing of the candidate waveform segment and the reference waveform segment.

[0048] The specific process of linear interpolation resampling is as follows: the original sampling sequence of the candidate waveform segment is used as discrete known points, its length is denoted as N2, and it is normalized and mapped at equal intervals on the time axis so that the normalized position t corresponds to the kth sampling point. k =(k 1) / (N2-1), thus forming a discrete function representation in the interval [0,1]. Then, using the length N1 of the reference waveform segment as the target number of resampling points, a new set of sampling positions t is uniformly generated within the same normalized interval. n =(n 1) / (N1 1), where n=1,…,N1. For each target sampling position t n Find the corresponding adjacent original sampling point t in the candidate sequence. k and t k+1 , making t k ≤t n ≤t k+1 Then, based on the linear interpolation formula y(t) n )=y k +(y k+1 -y k )×(t n -t k ) / (t k+1 -t k Calculate the amplitude value at this location, where y k y k+1 t k t k+1 The seismic amplitude values ​​at the sampling points. A resampled candidate waveform sequence of length N1 is obtained by calculating point-by-point for all target locations, thereby achieving consistent alignment with the reference waveform segment in terms of the number of sampling points and time scale.

[0049] By using a resampling alignment mechanism based on linear interpolation, seismic waveform segments of different lengths are unified to the same sampling point scale, thereby eliminating the influence of length differences caused by adaptive window truncation. This enables multidimensional feature weighted similarity calculation to be performed under the same time reference, improving the stability, consistency, and cross-channel comparability of similarity evaluation.

[0050] Furthermore, the candidate waveform segment and the reference waveform segment are weighted and fused from four dimensions: shape features, local rate of change features, amplitude intensity features, and temporal span features, to obtain the comprehensive similarity under the corresponding offset. This includes: standardizing the reference waveform segment and the candidate waveform segment with zero mean and unit variance, so that the mean of the two sequences is zero and the standard deviation is one; calculating the Pearson correlation coefficient between the two standardized sequences; calculating the normalized contour difference between the standardized sequences; and calculating the shape similarity score based on the Pearson correlation coefficient and the normalized contour difference; and calculating the first-order difference between the reference waveform segment and the candidate waveform segment respectively. The sequence is analyzed by obtaining a local change similarity score based on first-order difference similarity; the maximum amplitude of the reference waveform segment is extracted as the reference peak, and the maximum amplitude of the candidate waveform segment is extracted as the candidate peak. The amplitude similarity score is obtained by comparing the reference peak and the candidate peak; the number of sampling points of the reference waveform segment and the candidate waveform segment are counted respectively, and the smaller value and the larger value are compared and analyzed to obtain the temporal span similarity score; according to the configured weight of each dimension, the shape similarity score, local change similarity score, amplitude similarity score, and temporal span similarity score are weighted and fused to obtain the comprehensive similarity under the corresponding offset.

[0051] Specifically, after length normalization preprocessing of the candidate and reference waveform segments, the two waveform segments are quantitatively compared from four dimensions: shape features, local rate of change features, amplitude intensity features, and temporal span features. First, the reference and candidate waveform segments are standardized with zero mean and unit variance, i.e., their respective means are subtracted and divided by the standard deviation, converting them into dimensionless sequences with a mean of 0 and a variance of 1, thus eliminating energy scale differences and retaining only morphological information. The Pearson correlation coefficient is calculated for the two standardized sequences to measure the linear consistency of the two waveforms in the overall trend. Its value ranges from [-1, 1]. A Pearson correlation coefficient closer to 1 indicates a higher degree of positive correlation and more similar shapes; a Pearson correlation coefficient closer to -1 indicates opposite shapes; and a Pearson correlation coefficient close to 0 indicates no linear correlation. Simultaneously, the Euclidean distance between the two standardized sequences is calculated and normalized using the sequence length N1 to obtain the normalized contour difference, which reflects the degree of deviation of the waveform in local morphology. Then, the product of the normalized contour difference and a preset coefficient is subtracted from the Pearson correlation coefficient, and the result is truncated. If the result is negative, it is set to zero to obtain the shape similarity score. The preset coefficient is preferably 0.4.

[0052] Then, the first-order difference sequences of the reference waveform segment and the candidate waveform segment are calculated separately, which are the amplitude differences between adjacent sampling points. Each difference represents the amplitude change between adjacent sampling points. Then, the two first-order difference sequences are processed according to the complete calculation method of the shape feature similarity score described above: zero-mean unit variance standardization is performed on both, the Pearson correlation coefficient of the two standardized difference sequences is calculated, the normalized contour difference between the two standardized difference sequences is calculated, and then the product of the normalized contour difference and a preset coefficient is subtracted from the Pearson correlation coefficient. The result is then truncated to obtain the first-order difference similarity score.

[0053] The maximum amplitude of the reference waveform segment is extracted as the reference peak value, and the corresponding maximum amplitude value is extracted from the candidate waveform segments as the candidate peak value. The reference peak value and the candidate peak value are compared. By calculating the ratio of the smaller value to the larger value and performing a square root operation on this ratio, the resolution capability of the medium difference range is enhanced, thereby obtaining the amplitude similarity score, which reflects the consistency of the two waveforms in terms of energy level. The time-domain span feature is determined based on the original truncation length of the waveform segment, where the number of sampling points is the length of the continuous sampling interval determined by the dynamic boundary before resampling processing. The original number of sampling points of the reference waveform segment and the candidate waveform segment at the corresponding offset is counted respectively. The ratio of the smaller value to the larger value is taken, and the square root operation is performed on this ratio to obtain the time-domain span similarity score.

[0054] Then, according to the configured weights of each dimension, the shape similarity score, local change similarity score, amplitude similarity score, and time span similarity score are weighted and fused to obtain the comprehensive similarity under the corresponding offset. That is, comprehensive similarity = shape feature weight × shape similarity score + local change rate feature weight × local change similarity score + amplitude intensity feature weight × amplitude similarity score + time span feature weight × time span similarity score. Preferably, the shape feature weight is 0.35, the local change rate feature weight is 0.25, the amplitude intensity feature weight is 0.25, and the time span feature weight is 0.15. The corresponding weighted comprehensive score is obtained by multiplying the scores of each dimension by their corresponding weights and summing them.

[0055] Traditional waveform similarity methods typically rely on a single indicator such as correlation coefficient or Euclidean distance, which is prone to misjudgment: for example, two waveforms may have similar shapes but vastly different energies, or completely different shapes but similar amplitudes. This application achieves a comprehensive judgment of seismic waveform matching results through weighted fusion of four dimensions: shape features, local rate of change, amplitude intensity, and time span. This effectively avoids mismatches caused by a single indicator in complex geological noise environments, and improves the stability and accuracy of cross-seismic event identification and matching.

[0056] Furthermore, after obtaining the comprehensive similarity, the method further includes: calculating the energy distribution difference between the reference waveform segment and the candidate waveform segment on the left and right sides with their respective peak positions as the dividing line, and obtaining their respective asymmetry indices; calculating the absolute value of the difference between the asymmetry indices, and subtracting 1 from the product of the absolute value and a preset positive coefficient to obtain a penalty factor between 0 and 1, wherein if the result is negative, it is taken as 0; multiplying the comprehensive similarity obtained by weighted fusion by the penalty factor to obtain the final comprehensive similarity under the corresponding offset.

[0057] Specifically, after obtaining the initial comprehensive similarity, a penalty mechanism based on waveform asymmetry is introduced to correct the comprehensive similarity. First, the peak position of the reference waveform segment is used as the dividing point, where the peak position refers to the index of the sampling point with the largest absolute amplitude in its respective waveform sequence. This position typically corresponds to the instantaneous response point with the strongest seismic reflection energy. Using this peak position as the dividing line, the waveform is divided into a left interval and a right interval, and energy calculations are performed on both sides separately. The energy calculation method uses the sum of the squares of the amplitudes as the energy metric to obtain the left energy. e With right energy e And calculate the asymmetry index of the reference waveform segment, where the asymmetry index asym = |left e right e | / (left e +right e+ε), where ε is a very small positive number used to avoid the denominator being zero. The asymmetry index is used to characterize the degree of balance in energy distribution on both sides of the peak of the waveform; the larger the value, the more asymmetric the waveform. The same procedure is used to calculate the asymmetry index of the candidate waveform segments.

[0058] After calculating the asymmetry index for both the reference and candidate waveforms, calculate the absolute value of the difference between the two asymmetry indices, asym. diff Then, calculate the penalty factor P according to the following formula: P = max(0, 1.0 - b × asym) diff ), where b is a preset positive coefficient, set between 1.5 and 3.0, preferably b=2.0. Through calculation, a penalty factor P, between 0 and 1, is obtained. diff When P=0, it indicates that the asymmetry of the two waveforms is completely consistent. When P=1, no penalty is applied. When P<0, it is set to 0, thus ensuring that the penalty factor value is limited to the range of [0,1]. The comprehensive similarity obtained by weighted fusion of multidimensional features is multiplied by the penalty factor P to obtain the final comprehensive similarity under the corresponding offset.

[0059] By introducing energy structure constraints on both sides of the peak based on similarity calculations of shape, rate of change, amplitude, and temporal span, matching results can be further screened from the perspective of seismic wave physical propagation characteristics. This effectively suppresses pseudo-matches with similar shapes but inconsistent energy structures, and improves the consistency and reliability of geological matching.

[0060] Based on the comprehensive similarity at each offset, the highest similarity among all offsets is taken as the similarity matching result between the local seismic trace and the main seismic trace at the corresponding wave crest.

[0061] Specifically, after performing multi-position elastic alignment and offset-by-offset comprehensive similarity calculation on candidate peaks of neighboring seismic traces within a preset offset range, a final comprehensive similarity value is obtained for each offset. Then, based on the final comprehensive similarity under each offset, the highest similarity among all offsets, i.e., the maximum similarity value, is taken as the similarity matching result between the current neighboring seismic trace and the main seismic trace at the corresponding peak. This indicates the best alignment position of the current neighboring seismic trace waveform and the main seismic trace waveform on the time axis, i.e., the phase alignment point most likely corresponding to the same geological reflection event.

[0062] By performing maximum similarity selection within a limited offset search space, the uncertainty of seismic waveform time alignment is eliminated. The optimal matching result is automatically selected from multiple candidate alignment schemes, thereby ensuring that the cross-channel correlation corresponding to each wave peak has the highest similarity and the strongest physical consistency, and improving the accuracy, stability and reliability of seismic phase axis tracking and matching.

[0063] Furthermore, traversing each offset within the preset offset range also includes: when traversing each offset within the preset offset range, calculating the ratio of the number of sampling points to the number of sampling points of the reference waveform segment for each candidate waveform segment extracted under each offset; if the ratio exceeds a preset length ratio threshold, the corresponding offset is skipped directly, the comprehensive similarity under this offset is not calculated, and the comprehensive similarity of the corresponding offset is recorded as zero, wherein the preset length ratio threshold is 2.0.

[0064] Specifically, for each offset corresponding to a temporary center point, candidate waveform segments are extracted using a dynamic boundary determination method, and the original number of sampling points for each candidate waveform segment is calculated. Simultaneously, the number of sampling points for the reference waveform segment is also calculated. The number of sampling points represents the length of a continuous effective amplitude sequence at a fixed sampling interval, characterizing the duration span of the seismic reflection event on the time axis. The ratio of the number of sampling points for the candidate waveform segment to the number of sampling points for the reference waveform segment is calculated. If this ratio exceeds a preset length ratio threshold of 2.0, it indicates that the time span of the candidate waveform segment at that offset position significantly deviates from that of the reference waveform segment. The corresponding offset is skipped, the overall similarity at this offset is not calculated, and the overall similarity at the corresponding offset is recorded as zero.

[0065] By introducing a fast pruning mechanism based on the waveform segment length ratio, candidate matching results with severely inconsistent time scales are excluded in the early stages of the offset search, thereby reducing invalid computational overhead and avoiding misjudgment of similarity due to abnormally long waveforms or noise expansion. This improves the stability and computational efficiency of the overall matching process and enhances the physical rationality of seismic phase axis identification.

[0066] Example 2, based on the same inventive concept as the similarity measurement method based on dynamic waveform extraction and elastic alignment in the foregoing examples, such as... Figure 2 As shown, this application provides a similarity measurement system based on dynamic waveform extraction and flexible alignment, wherein the similarity measurement system based on dynamic waveform extraction and flexible alignment includes: The boundary determination module 11 is used to adaptively determine the left and right boundaries based on the candidate wave peak positions on the main seismic trace and according to the termination conditions arranged by priority, and extract the reference waveform segments of the main seismic trace; the offset traversal module 12 is used to traverse each offset within a preset offset range for the candidate wave peak positions on the adjacent seismic traces, and extract candidate waveform segments under each offset; the segment fusion module 13 is used to perform weighted fusion of the candidate waveform segments and the reference waveform segments from four dimensions: shape features, local rate of change features, amplitude intensity features, and time span features, to obtain the comprehensive similarity under the corresponding offset; the matching result acquisition module 14 is used to take the highest similarity among all offsets as the similarity matching result between the adjacent seismic trace and the main seismic trace at the corresponding wave peak based on the comprehensive similarity under each offset.

[0067] Furthermore, the boundary determination module 11 is also used for: a first priority where the absolute value of the amplitude of the current sampling point drops below a preset percentage of the peak amplitude of the candidate wave peak; a second priority where the current sampling point is a local minimum and satisfies that the amplitude of the current sampling point has the opposite sign to the amplitude of the candidate wave peak, or the absolute value of the amplitude of the current sampling point is less than a preset minimum percentage of the peak amplitude; and a third priority where the distance of the current sampling point from the starting center has reached a preset maximum search radius.

[0068] Furthermore, the boundary determination module 11 is also used to: take the candidate wave peak position as the starting center point, explore outwards point by point to the left and right sides respectively, and in the exploration process of each side, judge whether the termination condition is met in the order of first priority to third priority; when any priority termination condition is met, terminate the current side exploration, extend the current side final boundary outwards by a preset number of additional sampling points, and determine the left and right boundaries; and extract continuous sampling points from the main seismic trace sampling sequence according to the left and right boundaries to obtain reference waveform segments.

[0069] Furthermore, the boundary determination module 11 is also used to: set the preset percentage to 30% and the preset minimum percentage to 15%.

[0070] Furthermore, the offset traversal module 12 is also used for: the preset offset range is a sampling point, including a negative sampling point, a zero offset, and a positive sampling point; taking the candidate wave peak position on the adjacent seismic trace as the center, and offsetting to the left and right by the preset maximum offset amount according to the preset offset range, forming a set of continuous integer offsets including the zero offset; traversing each offset in the set of continuous integer offsets, taking the offset position as a temporary center point, and extracting candidate waveform segments under each offset; wherein, the extraction of the candidate waveform segments is also based on priority sorting to determine the left and right boundaries, and the candidate waveform segments are truncated based on the determined left and right boundaries.

[0071] Furthermore, the segment fusion module 13 is also used to: compare the number of sampling points between the candidate waveform segment and the reference waveform segment; when the number of sampling points of the extracted candidate waveform segment is different from that of the reference waveform segment, the candidate waveform segment is resampled to the same number of points as the reference waveform segment by linear interpolation before similarity calculation is performed.

[0072] Furthermore, the segment fusion module 13 is also used to: standardize the reference waveform segment and the candidate waveform segment with zero mean and unit variance, so that the mean of the two sequences is zero and the standard deviation is one; calculate the Pearson correlation coefficient of the two standardized sequences; calculate the normalized contour difference between the standardized sequences; calculate the shape similarity score based on the Pearson correlation coefficient and the normalized contour difference; calculate the first-order difference sequence of the reference waveform segment and the candidate waveform segment respectively; obtain the local change similarity score based on the first-order difference similarity; extract the maximum amplitude of the reference waveform segment as the reference peak value and extract the maximum amplitude of the candidate waveform segment as the candidate peak value; compare and analyze the reference peak value and the candidate peak value to obtain the amplitude similarity score; count the number of sampling points of the reference waveform segment and the candidate waveform segment respectively; compare and analyze the smaller and larger sampling values ​​to obtain the temporal span similarity score; and perform weighted fusion of the shape similarity score, local change similarity score, amplitude similarity score, and temporal span similarity score according to the configured weights of each dimension to obtain the comprehensive similarity under the corresponding offset.

[0073] Furthermore, the segment fusion module 13 is also used to: calculate the energy distribution difference between the reference waveform segment and the candidate waveform segment on the left and right sides with their respective peak positions as the dividing line, and obtain their respective asymmetry indices; calculate the absolute value of the difference between the asymmetry indices, and obtain a penalty factor between 0 and 1 by subtracting 1 from the product of the absolute value and a preset positive coefficient, wherein if the result is negative, it is taken as 0; multiply the weighted fusion obtained comprehensive similarity by the penalty factor to obtain the final comprehensive similarity under the corresponding offset.

[0074] Furthermore, the offset traversal module 12 is also used to: when traversing each offset within the preset offset range, calculate the ratio of the number of sampling points to the number of sampling points of the reference waveform segment for each candidate waveform segment extracted under each offset; if the ratio exceeds the preset length ratio threshold, directly skip the corresponding offset, do not calculate the comprehensive similarity under this offset, and record the comprehensive similarity of the corresponding offset as zero, wherein the preset length ratio threshold is 2.0.

[0075] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The similarity measurement method and specific examples based on dynamic waveform extraction and elastic alignment in the aforementioned Embodiment 1 are also applicable to the similarity measurement system based on dynamic waveform extraction and elastic alignment in this embodiment. Through the foregoing detailed description of the similarity measurement method based on dynamic waveform extraction and elastic alignment, those skilled in the art can clearly understand the similarity measurement system based on dynamic waveform extraction and elastic alignment in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0076] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0077] Obviously, those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A similarity measurement method based on dynamic waveform extraction and flexible alignment, characterized in that, include: Centered on the candidate wave crest position on the main seismic trace, the left and right boundaries are adaptively determined according to the termination conditions arranged by priority, and the reference waveform segments of the main seismic trace are extracted. For candidate wave peak locations on adjacent seismic traces, traverse each offset within a preset offset range and extract candidate waveform segments at each offset. The candidate waveform segment and the reference waveform segment are weighted and fused from four dimensions: shape features, local rate of change features, amplitude intensity features, and time span features, to obtain the comprehensive similarity under the corresponding offset. Based on the comprehensive similarity at each offset, the highest similarity among all offsets is taken as the similarity matching result between the local seismic trace and the main seismic trace at the corresponding wave crest.

2. The similarity measurement method based on dynamic waveform extraction and elastic alignment according to claim 1, characterized in that, The termination conditions for priority ranking include: The first priority is when the absolute value of the amplitude at the current sampling point drops to below a preset percentage of the peak-to-peak amplitude of the candidate wave; The second priority is that the current sampling point is a local minimum and satisfies the following conditions: the amplitude of the current sampling point is opposite in sign to the amplitude of the candidate peak, or the absolute value of the amplitude of the current sampling point is less than a preset minimum ratio of the peak amplitude. The third priority is when the distance of the current sampling point from the starting center has reached a preset maximum search radius.

3. The similarity measurement method based on dynamic waveform extraction and elastic alignment according to claim 2, characterized in that, Extract reference waveform segments from the main seismic trace, including: Starting from the candidate peak position as the center point, explore outwards point by point to the left and right sides respectively. During the exploration on each side, judge whether the termination condition is met in the order of first priority to third priority. When any priority termination condition is met, the current side exploration is terminated, and the final boundary of the current side is extended outward by a preset number of additional sampling points to determine the left and right boundaries. Based on the left and right boundaries, continuous sampling points are extracted from the main seismic trace sampling sequence to obtain reference waveform segments.

4. The similarity measurement method based on dynamic waveform extraction and elastic alignment according to claim 2, characterized in that, The preset percentage is 30%, and the preset minimum percentage is 15%.

5. The similarity measurement method based on dynamic waveform extraction and elastic alignment according to claim 1, characterized in that, For candidate wave crest locations on adjacent seismic traces, traverse each offset within a preset offset range, and extract candidate waveform segments at each offset, including: The preset offset range is a sampling point, which includes a negative sampling point, a zero offset, and a positive sampling point. Centered on the candidate peak position on the adjacent seismic trace, the preset offset range is offset to the left and to the right by a preset maximum offset amount to form a set of continuous integer offsets, including the zero offset. Traverse each offset in the set of consecutive integer offsets, take the offset position as a temporary center point, and extract the candidate waveform segment under each offset; The extraction of the candidate waveform segments is also based on priority ranking to determine the left and right boundaries, and the candidate waveform segments are then extracted based on the determined left and right boundaries.

6. The similarity measurement method based on dynamic waveform extraction and elastic alignment according to claim 1, characterized in that, Before performing weighted fusion of candidate waveform segments and reference waveform segments to obtain the comprehensive similarity at the corresponding offset, the following steps are also included: The number of sampling points of the candidate waveform segment and the reference waveform segment are compared. When the number of sampling points of the extracted candidate waveform segment is different from that of the reference waveform segment, the candidate waveform segment is resampled to the same number of points as the reference waveform segment by linear interpolation before similarity calculation is performed.

7. The similarity measurement method based on dynamic waveform extraction and elastic alignment according to claim 1, characterized in that, The candidate waveform segment and the reference waveform segment are weighted and fused from four dimensions: shape features, local rate of change features, amplitude intensity features, and temporal span features, to obtain the comprehensive similarity at the corresponding offset, including: The reference waveform segment and the candidate waveform segment are standardized with zero mean and unit variance respectively, so that the mean of the two sequences is zero and the standard deviation is one. The Pearson correlation coefficient of the two standardized sequences is calculated, and the normalized contour difference between the standardized sequences is calculated. Based on the Pearson correlation coefficient and the normalized contour difference, the shape similarity score is calculated. Calculate the first-order difference sequence between the reference waveform segment and the candidate waveform segment respectively, and obtain the local change similarity score based on the first-order difference similarity. The maximum amplitude of the reference waveform segment is extracted as the reference peak value, and the maximum amplitude of the candidate waveform segment is extracted as the candidate peak value. The amplitude similarity score is obtained by comparing and analyzing the reference peak value and the candidate peak value. The number of sampling points for the reference waveform segment and the candidate waveform segment are counted separately. The smaller and larger sampling values ​​are compared and analyzed to obtain the temporal span similarity score. Based on the configuration weights of each dimension, the shape similarity score, local change similarity score, amplitude similarity score, and temporal span similarity score are weighted and fused to obtain the comprehensive similarity under the corresponding offset.

8. The similarity measurement method based on dynamic waveform extraction and elastic alignment according to claim 7, characterized in that, After obtaining the overall similarity score, it also includes: The energy distribution difference between the reference waveform segment and the candidate waveform segment is calculated separately with their respective peak positions as the dividing line, and their respective asymmetry indices are obtained. Calculate the absolute value of the difference between the asymmetry indices, and subtract 1 from the product of the absolute value and the preset positive coefficient to obtain a penalty factor between 0 and 1, wherein if the result is negative, it is taken as 0; Multiply the weighted fusion result by the penalty factor to obtain the final overall similarity at the corresponding offset.

9. The similarity measurement method based on dynamic waveform extraction and elastic alignment according to claim 1, characterized in that, Traversing each offset within the preset offset range also includes: When traversing each offset within the preset offset range, for each candidate waveform segment extracted at each offset, the ratio of the number of sampling points to the number of sampling points of the reference waveform segment is calculated; If the ratio exceeds the preset length ratio threshold, the corresponding offset is skipped directly, the comprehensive similarity under this offset is not calculated, and the comprehensive similarity of the corresponding offset is recorded as zero. The preset length ratio threshold is 2.

0.

10. A similarity measurement system based on dynamic waveform extraction and flexible alignment, characterized in that, The step of implementing the similarity measurement method based on dynamic waveform extraction and elastic alignment according to any one of claims 1 to 9, wherein the similarity measurement system based on dynamic waveform extraction and elastic alignment comprises: The boundary determination module is used to adaptively determine the left and right boundaries based on the candidate wave crest positions on the main seismic trace and the termination conditions arranged according to priority, and to extract the reference waveform segments of the main seismic trace. The offset traversal module is used to traverse each offset within a preset offset range for the candidate wave crest positions on adjacent seismic traces, and extract candidate waveform segments at each offset. The segment fusion module is used to perform weighted fusion of candidate waveform segments and reference waveform segments from four dimensions: shape features, local rate of change features, amplitude intensity features, and temporal span features, to obtain the comprehensive similarity under the corresponding offset. The matching result acquisition module is used to take the highest similarity among all offsets as the similarity matching result between the local seismic trace and the main seismic trace at the corresponding wave crest, based on the comprehensive similarity under each offset.

Images