Data denoising method and device, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention relates to the technical field of data processing, and discloses a data denoising method and device, electronic equipment, a storage medium and a program product, and the method comprises the steps: obtaining to-be-processed DAS data in a mining area environment; curvelet transformation is carried out on the to-be-processed DAS data to obtain a curvelet coefficient set, the curvelet coefficient set corresponds to multiple scales, the multiple scales comprise at least one key scale, and the at least one key scale is determined based on a frequency band to which an effective signal in the mining area environment belongs; the density of multiple directions corresponding to each key scale in a key direction range is higher than the density in a non-key direction range, and the key directions are determined based on geological features of a mining area environment; performing denoising processing on the curvelet coefficient set to obtain a denoised curvelet coefficient set; and generating first DAS data based on the de-noised curvelet coefficient set. In this way, the quality of the DAS data is improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a data denoising method, apparatus, electronic device, storage medium, and program product. Background Technology

[0002] Distributed optical fiber sensing technology uses optical fiber as a sensing medium to measure the scattered light along the optical fiber to obtain DAS (Distributed Acoustic Sensing) data, which can then be used to achieve high-precision, high-resolution monitoring of specific areas.

[0003] During seismic exploration in mining areas, the terrain is complex and varied, making the deployment of traditional seismic instruments costly. Distributed fiber optic sensing technology, however, utilizes optical cables to acquire large-scale DAS data, effectively solving the problems of terrain obstacles and high costs associated with setting up survey lines in mining areas. However, DAS data acquisition is significantly affected by noise, resulting in low data quality and impacting the accuracy of subsequent signal processing and interpretation.

[0004] How to denoise DAS data to improve data quality is an urgent problem to be solved. Summary of the Invention

[0005] The purpose of this application is to provide at least one data denoising method, apparatus, electronic device, storage medium, and program product, which can at least solve the problem of denoising DAS data and at least achieve the effect of improving the quality of DAS data.

[0006] To address the aforementioned technical problems, at least one embodiment of this application provides a data denoising method, comprising: acquiring distributed acoustic wave sensor (DAS) data to be processed in a mining environment; performing curvelet transform on the DAS data to be processed to obtain a set of curvelet coefficients, wherein the set of curvelet coefficients corresponds to multiple scales, the multiple scales including at least one key scale, the at least one key scale being determined based on the frequency band to which the effective signal in the mining environment belongs, and the density of multiple directions corresponding to each key scale within the key direction range being higher than the density within the non-key direction range, the key direction being determined based on the geological characteristics of the mining environment; performing denoising processing on the set of curvelet coefficients to obtain a denoised curvelet coefficient set; and generating first DAS data based on the denoised curvelet coefficient set.

[0007] Optimizing the core parameters (scale and direction) of curvelet transform based on the effective signal frequency bands and geological characteristics of the mining area environment makes the denoising tool no longer general-purpose but "tailor-made" for the mining area, fundamentally improving the accuracy of denoising. By identifying and densifying directional subbands within key directional ranges at critical scales, the resolution and fidelity of effective signals propagating along the main geological structures of the mining area are greatly improved, better preserving geologically significant effective signals. Non-uniform directional partitioning means that computation can be reduced in signal-insensitive directions, thereby improving the overall efficiency of the algorithm without sacrificing key information, making it more suitable for processing large-scale data generated by DAS.

[0008] In some optional embodiments, acquiring the distributed acoustic wave sensor (DAS) data to be processed in the mining environment includes: acquiring the raw DAS data in the mining environment, the raw DAS data including multiple time series; constructing a common-mode noise vector based on the preset quantiles of the amplitude of the multiple time series at each time point; determining the cross-correlation coefficient between each time series and the common-mode noise vector; and removing the product of the cross-correlation coefficient corresponding to the time series and the common-mode noise vector from each time series of the raw DAS data to obtain the DAS data to be processed.

[0009] DAS data contains common-mode noise caused by instruments, etc. Preprocessing directly targets the in-phase noise inherent in the DAS system that affects all channels, facilitating subsequent denoising and improving the quality of the first DAS data after denoising. Common-mode noise is easily confused with effective signals in the sparse domain; prioritizing common-mode noise removal prevents it from interfering with subsequent curve transform and thresholding, significantly improving the effectiveness and reliability of sparse domain denoising. Using preset quantiles instead of averages to estimate common-mode noise is less sensitive to outliers (such as strong effective signals on individual channels), resulting in a more robust estimate. Adaptive weighted removal using cross-correlation coefficients avoids damage to effective signals caused by a one-size-fits-all approach.

[0010] In some optional embodiments, the denoising process of the curvelet coefficient set to obtain a denoised curvelet coefficient set includes: denoising the curvelet coefficient set according to the following hard threshold function:

[0011] in, The curve wave coefficients are the curve wave coefficients in the set of curve wave coefficients. For the set of denoised curved coefficients and The corresponding curve coefficient;

[0012] for , One of them, or is and The product; The geological weights are determined based on the geological characteristics of the mining area environment. yes The function, For location; The statistical weights are determined based on the local mathematical and statistical properties of the curve coefficients. yes The function, As a scale, For direction; For based on The basic threshold is determined by the statistical distribution of the curvature coefficients within the sub-band.

[0013] In sparse domain denoising, a dynamic hard threshold function is employed. This threshold is not fixed but determined by a base threshold, geological weights, and / or statistical weights. These geological and statistical weights enable adaptive denoising that is "time- and location-dependent," improving denoising accuracy. The dynamic threshold mechanism overcomes the contradiction of fixed thresholds failing to balance denoising intensity and signal detail globally, preserving valuable weak but effective signals even in noisy environments. Integrating geological features as weights into the purely data-driven algorithm gives the denoising process physical guidance, resulting in results that better conform to geological patterns.

[0014] In some optional embodiments, generating the first DAS data based on the denoised curved wave coefficient set includes: obtaining the confidence weight of the sub-band corresponding to each scale-direction in the denoised curved wave coefficient set, the confidence weight being determined based on the average energy of the sub-band corresponding to each scale-direction; weighting the curved wave coefficients in the denoised curved wave coefficient set based on the confidence weight; and generating the first DAS data based on the weighted and adjusted denoised curved wave coefficient set.

[0015] When reconstructing the signal from the denoised sparse coefficients, a confidence weight is introduced. This weighted reconstruction of subbands at different scales and directions assigns higher weights to subbands rich in effective signal while suppressing noise-dominant subbands. This achieves a second filtering and enhancement during the reconstruction stage, resulting in a final signal with a higher signal-to-noise ratio. The confidence weight is determined based on the average energy of the subbands corresponding to each scale and direction, making it objective and data-driven, thus ensuring the scientific validity and reproducibility of the method.

[0016] In some optional embodiments, generating the first DAS data based on the weighted adjusted set of denoised curve coefficients includes: constructing a constraint formula: ,in, For preset reconstruction signals, for The set of curvelet coefficients after curvelet transform. The set of weighted and adjusted denoised curve coefficients. The notation for L2 norm calculation. For regularization parameters, To Perform total variational processing; with the goal of minimizing the constraints, apply the following... Optimize and complete the optimization. This serves as the first DAS data.

[0017] The reconstruction process is formalized as an optimization problem, with the objective function containing... (The goal is to reconstruct the signal and ensure consistency with the denoising coefficients) fidelity terms and The total variational regularization term (which aims to make the reconstructed signal smooth and edge-preserving) results in a more physically reasonable output signal. Total variational regularization tends to produce piecewise smooth signals, consistent with the characteristics of the seismic signal phase axis (smooth internally with clear boundaries). Therefore, the reconstruction result not only denoises but also preserves the edges of the effective signal, avoiding blurring. (The text then abruptly shifts to a description of sparse domain denoising.) A small amount of noise remains, and the regularization process can further suppress the noise during reconstruction, improving the fault tolerance and robustness of the entire method. This optimization framework mathematically restricts the reconstructed signal, resulting in a higher quality signal that better meets the requirements of subsequent interpretation than directly performing an inverse transform.

[0018] In some optional embodiments, the method further includes: subtracting the first DAS data from the DAS data to be processed to obtain first residual data; performing curvelet transform and low-threshold denoising on the first residual data, and generating a first effective signal based on the obtained result; superimposing the first effective signal with the first DAS data to obtain second DAS data; subtracting the second DAS data from the DAS data to be processed to obtain second residual data; comparing the similarity of the second residual data with each effective signal in the effective signal pattern library corresponding to the mining area environment, and taking the portion of the second residual data with a similarity greater than a similarity threshold as the second effective signal; superimposing the second effective signal with the second DAS data to obtain third DAS data; subtracting the third DAS data from the DAS data to be processed to obtain third residual data; removing components in the third residual data that are lower than the physical background noise in the mining area environment to obtain a third effective signal; and superimposing the third effective signal with the third DAS data to obtain target DAS data.

[0019] This embodiment provides a three-stage residual compensation mechanism to progressively recover potentially deleted valid signals from the denoising residuals, thereby better protecting valid signals. Through three coarse-to-fine residual processing steps, it systematically retrieves weak or uniquely shaped valid signals that may have been lost in the main denoising process. The first compensation (low threshold) achieves gentle recovery, targeting high-frequency details that have been over-thresholded. The second compensation (template matching) achieves precise recovery, using prior knowledge (valid signal pattern library) to identify and recover components that conform to typical valid signal patterns. The third compensation (physical threshold) achieves safe recovery, providing a final layer of protection based on the physical background noise level, ensuring that no valid energy above the noise level is lost. Ultimately, this significantly improves the integrity of the target DAS data. For high-value, low signal-to-noise ratio exploration data, preserving every weak signal is crucial, ensuring the high fidelity and integrity of the final data, and providing a solid foundation for subsequent detailed geological interpretation.

[0020] At least one embodiment of this application also provides a data denoising device, comprising: an acquisition module for acquiring distributed acoustic wave sensor (DAS) data to be processed in a mining environment; a processing module for performing curvelet transform on the DAS data to be processed to obtain a set of curvelet coefficients, the set of curvelet coefficients corresponding to multiple scales, the multiple scales including at least one key scale, the at least one key scale being determined based on the frequency band to which the effective signal in the mining environment belongs, the density of multiple directions corresponding to each key scale being higher in the key direction range than in the non-key direction range, the key direction being determined based on the geological characteristics of the mining environment; a denoising module for performing denoising processing on the set of curvelet coefficients to obtain a denoised curvelet coefficient set; and a generation module for generating first DAS data based on the denoised curvelet coefficient set.

[0021] In some optional embodiments, an acquisition module is used to acquire raw DAS data in the mining area environment, the raw DAS data including multiple time series; construct a common-mode noise vector based on the preset quantile of the amplitude of the multiple time series at each time point; determine the cross-correlation coefficient between each time series and the common-mode noise vector; and remove the product of the cross-correlation coefficient corresponding to the time series and the common-mode noise vector from each time series of the raw DAS data to obtain the DAS data to be processed.

[0022] In some optional embodiments, the denoising module is configured to denoise the set of curvelet coefficients according to the following hard threshold function:

[0023] in, The curve wave coefficients are the curve wave coefficients in the set of curve wave coefficients. For the set of denoised curved coefficients and The corresponding curve coefficient;

[0024] for , One of them, or is and The product; The geological weights are determined based on the geological characteristics of the mining area environment. yes The function, For location; The statistical weights are determined based on the local mathematical and statistical properties of the curve coefficients. yes The function, As a scale, For direction; For based on The basic threshold is determined by the statistical distribution of the curvature coefficients within the sub-band.

[0025] In some optional embodiments, the generation module is used to obtain the confidence weight of the sub-band corresponding to each scale-direction in the set of denoised curved coefficients, the confidence weight being determined based on the average energy of the sub-band corresponding to each scale-direction; to perform weighted adjustment on the curved coefficients in the set of denoised curved coefficients based on the confidence weight; and to generate the first DAS data based on the weighted and adjusted set of denoised curved coefficients.

[0026] In some optional embodiments, the generation module is used to construct the constraint formula: ,in, For preset reconstruction signals, for The set of curvelet coefficients after curvelet transform. The set of weighted and adjusted denoised curve coefficients. The notation for L2 norm calculation. For regularization parameters, To Perform total variational processing; with the goal of minimizing the constraints, apply the following... Optimize and complete the optimization. This serves as the first DAS data.

[0027] In some optional embodiments, the apparatus further includes: a compensation module, configured to: subtract the first DAS data from the DAS data to be processed to obtain first residual data; perform curvelet transform and low-threshold denoising on the first residual data, and generate a first valid signal based on the obtained result; superimpose the first valid signal with the first DAS data to obtain second DAS data; subtract the second DAS data from the DAS data to be processed to obtain second residual data; compare the similarity of the second residual data with each valid signal in the valid signal pattern library corresponding to the mining area environment, and take the portion of the second residual data with a similarity greater than a similarity threshold as the second valid signal; superimpose the second valid signal with the second DAS data to obtain third DAS data; subtract the third DAS data from the DAS data to be processed to obtain third residual data; remove components in the third residual data that are lower than the physical background noise in the mining area environment to obtain a third valid signal; and superimpose the third valid signal with the third DAS data to obtain target DAS data.

[0028] At least one embodiment of this application also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the data denoising method described above.

[0029] At least one embodiment of this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described data denoising method.

[0030] At least one embodiment of this application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described data denoising method. Attached Figure Description

[0031] One or more embodiments are illustrated by way of example with reference to the accompanying drawings, and these illustrative descriptions do not constitute a limitation on the embodiments.

[0032] Figure 1 This is a flowchart of a data denoising method provided in one embodiment of this application; Figure 2 This is a flowchart of an embodiment of the present application for acquiring distributed acoustic wave sensor (DAS) data to be processed in a mining environment; Figure 3 This is a flowchart illustrating the generation of first DAS data based on a set of denoised curved coefficients, as provided in one embodiment of this application. Figure 4 This is a flowchart of a residual iterative process for processing DAS data provided in one embodiment of this application; Figure 5 This is a schematic diagram of a data denoising apparatus provided in another embodiment of this application. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the various embodiments of this application to help readers better understand this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various changes and modifications based on the following embodiments. The division of the various embodiments below is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0034] To facilitate understanding of the embodiments of this application, relevant content regarding DAS data (signals) will be introduced first.

[0035] Distributed fiber optic sensing technology is an innovative sensing method that uses optical fibers as the sensing medium to measure scattered light along the fiber, achieving high-precision, high-resolution monitoring of specific areas. This technology offers advantages such as wide observation range, convenient installation, and low cost, and has therefore been widely applied in fields such as oil and gas exploration, seismic monitoring, and pipeline safety monitoring.

[0036] During seismic exploration in mining areas, the complex and varied terrain makes the deployment of traditional seismic instruments costly. Distributed fiber optic sensing technology, however, utilizes optical cables to acquire large-scale DAS data, effectively solving the problems of terrain obstacles and high costs associated with setting up survey lines in mining areas. However, DAS data acquisition is significantly affected by environmental noise (e.g., mining equipment, trucks, crushers, and blasting shock waves), instrument noise (e.g., photoelectric conversion circuit noise), and transmission loss (e.g., signal-to-noise ratio decreases exponentially with distance), resulting in lower data quality and impacting the accuracy of subsequent signal processing and interpretation.

[0037] To improve data quality, commonly used denoising methods include traditional methods such as frequency domain filtering and median filtering. However, while these methods remove noise, they often lose some of the detailed features in the DAS data, resulting in a decrease in signal fidelity.

[0038] Furthermore, in response to the denoising requirements of DAS data, most existing denoising methods are based on removing natural noise from traditional seismic data acquired by nodes. However, DAS data not only faces interference from natural noise but also contains unique instrument noise.

[0039] Instrument noise in DAS data mainly originates from minute vibrations in the optoelectronic system. Due to the structural characteristics of the equipment, this type of noise exhibits in-phase characteristics, meaning it simultaneously affects data from all receiving channels, manifesting as a fixed in-phase noise (common-mode noise). The presence of this noise not only affects the effective extraction of signals but also poses a challenge to denoising methods based on sparse representation, as in-phase noise and the representation of the useful signal in the sparse domain are prone to aliasing.

[0040] To address the aforementioned technical problems encountered in DAS data denoising, this application proposes a data denoising method. The implementation details of the data denoising method in this embodiment are described below. The following implementation details are provided for ease of understanding and are not essential for implementing this solution.

[0041] Example 1: The data denoising method of this embodiment can be applied to electronic devices with communication, computing, and data storage capabilities. Its specific process can be as follows: Figure 1 As shown, it includes: S101, acquire the DAS data to be processed in the mining area environment.

[0042] In some cases, the DAS data to be processed can be DAS data directly collected from optical fibers deployed in the mining environment.

[0043] In another embodiment, the DAS data to be processed can be: raw DAS data directly collected from optical fibers deployed in the mining environment, after removing common-mode noise.

[0044] Among the DAS data collected in the mining environment, common-mode noise is mainly caused by the minute vibrations of the photoelectric system that realizes DAS data acquisition.

[0045] In some examples, acquiring DAS data to be processed in a mining environment may include: directly acquiring the DAS data to be processed from the mining environment via an optical fiber connected by an electronic device; or, directly acquiring the raw DAS data from the mining environment via an optical fiber connected by an electronic device; removing common-mode noise from the raw DAS data to obtain the DAS data to be processed.

[0046] In some examples, acquiring unprocessed DAS data in a mining environment may include receiving the unprocessed DAS data sent by other devices.

[0047] S102, Perform curvelet transform on the DAS data to be processed to obtain a set of curvelet coefficients.

[0048] When performing curvelet transform on DAS data, multi-scale and multi-directional curvelet transforms are performed to decompose the DAS data into curvelet coefficients of different scales and directions, resulting in a set of curvelet coefficients. In some cases, the curvelet transform can be a second-generation curvelet transform.

[0049] Among them, the set of curve coefficients corresponds to multiple scales, including at least one key scale. The key scale is determined based on the frequency band to which the effective signal in the mining area belongs. The density of multiple directions corresponding to each key scale is higher than the density in the non-key direction range. The key direction is determined based on the geological characteristics of the mining area environment.

[0050] The embodiments of this application do not limit the specific number of scales for the at least one key scale, but determine it based on the frequency band to which the effective signal in the mining environment belongs. The frequency band to which the effective signal in the mining environment belongs can be determined based on prior knowledge. For example, based on prior knowledge, the contents shown in Table 1 below can be obtained.

[0051] Table 1

[0052] As shown in Table 1 above, the enriched frequency band for effective microseismic signals is 30Hz-80Hz, meaning that the effective frequency band for seismic signals is 30Hz-80Hz. Based on prior knowledge, the frequency bands of all effective signals in the mining area environment can be determined.

[0053] Since different scales correspond to different frequency bands during curvelet transform, the corresponding scale can be determined based on the frequency band to which the effective signal belongs. Among the multiple scales corresponding to the curvelet coefficient set, the scale corresponding to the frequency band to which the effective signal belongs is the key scale. The specific number of scales or which scales correspond to the frequency band to which the effective signal belongs is not limited in the embodiments of this application and can be determined according to the method of decomposition of scales and the number of scales during curvelet transform.

[0054] Regarding the density of multiple directions corresponding to each critical scale that is higher within the critical direction range than within the non-critical direction range, the critical directions are determined based on the geological characteristics of the mining area environment. These geological characteristics represent the geological features of the mining area environment, including the dominant joints, fissures, or fault strikes of the mining area's rock mass. Geological characteristics can influence which directions in the mining area are more likely to generate effective signals; for example, effective signals are more likely to be generated along the dominant joints, fissures, or fault strikes of the mining area's rock mass. Therefore, the range of critical directions can be determined based on the geological characteristics of the mining area environment. For example, the critical direction ranges included in the mining area environment may be at least one of the following: the direction range corresponding to the dominant joint direction of the mining area's rock mass, the direction range corresponding to the fissure strike, or the direction range corresponding to the fault strike.

[0055] Taking the directional range corresponding to the fault strike in a mining area environment as an example, determining the directional range corresponding to the fault strike can include: first determining the fault strike; then, using the fault strike as a guideline, the directional range corresponding to the surrounding preset angle range is the key directional range. The specific value of this preset angle range is not limited in the embodiments of this application. For example, the preset angle range is ±15°.

[0056] In the entire direction corresponding to each scale, the reverse range, except for the critical direction range, is a non-critical direction range.

[0057] The embodiments of this application do not limit the specific proportional relationship between the density of directions used for decomposing DAS data within the critical direction range and the density of directions used for decomposing DAS data within the non-critical direction range. For example, taking the entire direction range corresponding to each scale as 0~180°, the critical direction range is 0~30°, and 30°~180° is the non-critical direction range. Taking the number of multiple directions corresponding to the critical scale as 16, we can uniformly select 8 directions within 0~30° and 8 directions within 30°~180°, for a total of 16 directions to decompose DAS data. It can be seen that the density of multiple directions corresponding to the critical scale within 0~30° is higher than the density within 30°~180°.

[0058] By determining the number of scales for the curvelet transform and selecting key scales from multiple scales, and then determining the number of directions corresponding to each scale, when dividing the directions for the key scales, dense division is performed within the key direction range, and the spacing is appropriately widened in the non-key range. Then, the curvelet transform is performed on the DAS data to be processed using these multiple scales and the multiple directions corresponding to each scale to obtain the set of curvelet coefficients.

[0059] S103, Denoise the set of curved wave coefficients to obtain a denoised set of curved wave coefficients.

[0060] In some examples, denoising the set of curvelet coefficients to obtain a denoised set of curvelet coefficients can include: denoising the set of curvelet coefficients using a hard threshold function to obtain a denoised set of curvelet coefficients.

[0061] In other examples, the set of curvelet coefficients is denoised to obtain a denoised set of curvelet coefficients, including: The curvelet coefficient set is denoised using the following hard thresholding function:

[0062] in, The curve wave coefficients are the curve wave coefficients in the set of curve wave coefficients. As a scale, As direction, For location, For the set of denoised curved wave coefficients and The corresponding curve coefficient.

[0063]

[0064] for , One of them, or is and The product; The geological weights are determined based on the geological characteristics of the mining area environment. yes The function; The statistical weights are determined based on the local mathematical and statistical properties of the curve coefficients. yes The function; For based on The basic threshold is determined by the statistical distribution of the curvature coefficients within the sub-band.

[0065] Among them, the geological weights determined based on the geological characteristics of the mining area environment may include: dividing the area in the mining area into different regions according to the geological characteristics, including unstable rock mass regions, stable rock mass regions and general regions.

[0066] Unstable rock mass areas are those prone to rock mass vibration, such as faults and fissures. Stable rock mass areas are generally areas that do not experience rock mass vibration and can be identified based on geological characteristics or pre-selected as stable rock mass areas. In a mining area, the areas other than unstable and stable rock mass areas are considered general areas.

[0067] Regarding geological weights, the geological weights corresponding to unstable rock mass areas can be set to values ​​greater than 0 and less than 1, the geological weights corresponding to stable rock mass areas can be set to values ​​greater than 1, and the geological weights for general areas can be set to 1. This setting method is merely an example. However, the overall relationship is: geological weight corresponding to rock mass areas > geological weight corresponding to general areas > geological weight corresponding to unstable rock mass areas.

[0068] Regarding statistical weights, the local mathematical statistical properties of the curvelet coefficients refer to the fact that a curvelet coefficient... The curve coefficients formed by the curves and other curve coefficients in the surrounding preset area The mathematical and statistical properties of the corresponding set of local curvature coefficients.

[0069] in, Other curve coefficients within the surrounding preset area refer to those located within the same sub-band (same scale and direction). The curvature coefficient within a predetermined surrounding area. The embodiments of this application do not limit the specific type of this predetermined surrounding area; for example, using... Centered on the area, the surrounding 5x5 area is The surrounding preset area.

[0070] In some examples, the statistical weights determined based on the local mathematical statistical properties of the curve coefficients may include: calculating the variance of all curve coefficients in a preset region around the curve coefficient as the center; and determining the statistical weight corresponding to the curve coefficient based on the variance.

[0071] High variance indicates that the curve coefficients within the surrounding preset area vary drastically, with some being very large and others very small, thus indicating the presence of a spatially coherent and effective signal in the area (e.g., an earthquake phase axis passing through this window).

[0072] A low variance indicates that all curve coefficients within the surrounding preset area are roughly the same size (either very small or very large). If they are all small, it is basically a noise area; if the values ​​are all large and flat, it may be some kind of background interference.

[0073] Therefore, it can be configured such that if the variance corresponding to the curvature coefficient is high, the statistical weight is less than 1 (e.g., 0.8) to protect the effective signal. If the variance corresponding to the curvature coefficient is low, the statistical weight is greater than 1 (e.g., 1.2) to more aggressively remove noise.

[0074] The embodiments of this application do not limit how to determine whether the variance is high or low. For example, a variance threshold can be set, and the variance can be compared with the variance threshold. If the variance is greater than the variance threshold, it is considered high; otherwise, it is considered low. As another example, two variance thresholds, one large and one small, can be set. If the variance is less than the smaller variance threshold, it is considered low; if the variance is greater than the larger variance threshold, it is considered high. If the variance is between the small and large variance thresholds, it is considered normal variance, and the statistical weight corresponding to this variance can be set to 1.

[0075] The variance threshold can be determined based on prior knowledge and experience.

[0076] about , The statistical distribution of the subband curvature coefficient refers to Statistical distribution of all curvelet coefficients within the sub-band. Based on this... The statistical distribution of all curve coefficients within the subband was determined. It can include: based on All curve coefficients within the sub-band were determined using the median absolute deviation method. The base threshold corresponding to the sub-band .

[0077] In some cases, it is also possible to All curve coefficients within a sub-band are sorted, and the curve coefficient at a set quantile is selected as the base threshold. Regarding the specific quantile setting, the embodiments of this application do not impose limitations and can determine it based on experience and prior knowledge. For example, if the sorting method is from smallest to largest, the preset quantile can be 40%; if the sorting method is from largest to smallest, the preset quantile can be 60%.

[0078] In sparse domain denoising, a dynamic hard threshold function is employed. This threshold is not fixed but determined by a base threshold, geological weights, and / or statistical weights. These geological and statistical weights enable adaptive denoising that is "time- and location-dependent," improving denoising accuracy. The dynamic threshold mechanism overcomes the contradiction of fixed thresholds failing to balance denoising intensity and signal detail globally, preserving valuable weak but effective signals even in noisy environments. Integrating geological features as weights into the purely data-driven algorithm gives the denoising process physical guidance, resulting in results that better conform to geological patterns.

[0079] S104, Generate the first DAS data based on the set of denoised curved wave coefficients.

[0080] In some examples, generating the first DAS data based on the set of denoised curvelet coefficients may include performing an inverse curvelet transform on the set of denoised curvelet coefficients to obtain the first DAS data.

[0081] Optimizing the core parameters (scale and direction) of curvelet transform based on the effective signal frequency bands and geological characteristics of the mining area environment makes the denoising tool no longer general-purpose but "tailor-made" for the mining area, fundamentally improving the accuracy of denoising. By identifying and densifying directional subbands within key directional ranges at critical scales, the resolution and fidelity of effective signals propagating along the main geological structures of the mining area are greatly improved, better preserving geologically significant effective signals. Non-uniform directional partitioning means that computation can be reduced in signal-insensitive directions, thereby improving the overall efficiency of the algorithm without sacrificing key information, making it more suitable for processing large-scale data generated by DAS.

[0082] In some embodiments, acquiring distributed acoustic wave sensor (DAS) data to be processed in a mining environment can be achieved through methods such as... Figure 2 The steps S201-S204 shown are implemented.

[0083] S201, Obtain raw DAS data in the mining area environment. The raw DAS data includes multiple time series.

[0084] Raw DAS data is DAS data directly collected from optical fibers deployed in the mining environment. In some cases, acquiring raw DAS data in the mining environment may include: directly acquiring the raw DAS data through optical fibers connected to electronic devices; or receiving the raw DAS data sent by other devices.

[0085] The raw DAS data includes time series from multiple channels, that is, multiple time series (these multiple time series have the same length). For example, the multiple channels are n signals and the length of the time series is m.

[0086] S202 constructs a common-mode noise vector based on the preset quantiles of the amplitudes of multiple time series at each time point.

[0087] The amplitude of a time series at a given time point is the value of that time series at that time point. For example, if the time series in a DAS dataset is {a1, a2, a3}, then the time series corresponds to three time points: the amplitude at the first time point is a1, the amplitude at the second time point is a2, and the amplitude at the third time point is a3.

[0088] The specific value of the preset quantile is not limited in the embodiments of this application. For example, for a point in time, the amplitudes of multiple time series at that point in time are sorted from smallest to largest, and the value at the 50th quantile of the sorting result is taken as the target amplitude. The target amplitude at each point in time is used to construct the common mode noise vector.

[0089] It should be noted that the preset quantiles can be determined based on whether the sorting is from largest to smallest or smallest to largest. The preset quantiles can be determined based on the sorting method and experience.

[0090] S203, determine the cross-correlation coefficient between each time series and the common-mode noise vector.

[0091] The cross-correlation coefficient is used to represent the proportion of common-mode noise in the time series (signals in the channel).

[0092] In some cases, the cross-correlation coefficient can be a similarity coefficient, which can be calculated by measuring the similarity between each time series and the common-mode noise vector.

[0093] The similarity between each time series and the common-mode noise vector can be calculated using any method that calculates the similarity between vectors.

[0094] In some cases, the cross-correlation coefficient can be the Pearson correlation coefficient, which can be determined by calculating the Pearson correlation coefficient between the time series and the common-mode noise vector to determine the cross-correlation coefficient between each time series and the common-mode noise vector.

[0095] S204. From each time series of the original DAS data, remove the product of the cross-correlation coefficient and the common-mode noise vector corresponding to the time series to obtain the DAS data to be processed.

[0096] Taking a time series from the original DAS data as an example, the time series is S1, the corresponding cross-correlation vector is a, and the common-mode noise vector is Q. Then, the time series S2 corresponding to the time series S1 in the DAS data to be processed is S1-a×Q.

[0097] DAS data contains common-mode noise caused by instruments, etc. Preprocessing directly targets the in-phase noise inherent in the DAS system that affects all channels, facilitating subsequent denoising and improving the quality of the first DAS data after denoising. Common-mode noise is easily confused with effective signals in the sparse domain; prioritizing common-mode noise removal prevents it from interfering with subsequent curve transform and thresholding, significantly improving the effectiveness and reliability of sparse domain denoising. Using preset quantiles instead of averages to estimate common-mode noise is less sensitive to outliers (such as strong effective signals on individual channels), resulting in a more robust estimate. Adaptive weighted removal using cross-correlation coefficients avoids damage to effective signals caused by a one-size-fits-all approach.

[0098] In some embodiments, the first DAS data is generated based on the set of denoised curve coefficients, which can be achieved through methods such as... Figure 3 The steps S301-S303 shown are implemented.

[0099] S301, obtain the confidence weight of the sub-band corresponding to each scale-direction in the set of denoised curved wave coefficients. The confidence weight is determined based on the average energy of the sub-band corresponding to each scale-direction.

[0100] In some cases, the energy of a curvelet coefficient can be represented by the square of that coefficient. The energy of a scale-direction corresponding subband can be represented by the sum of the squares of all curvelet coefficients within that subband. The average energy of a subband is the average energy of the curvelet coefficients within that subband. For example, if a subband contains two curvelet coefficients b1 and b2, then the average energy of the curvelet coefficients is (b1...). 2 +b2 2 ) / 2.

[0101] The higher the average energy of the subband, the greater the confidence weight.

[0102] In some examples, the confidence weight of each subband at each scale-direction is determined based on the average energy of the subband at each scale-direction. This may include: calculating the average of the average energy of the subband at each scale-direction; and dividing the average energy of the subband by the average value to obtain the confidence weight of the subband.

[0103] S302, the curve coefficients in the set of denoised curve coefficients are weighted and adjusted based on the credibility weight.

[0104] Adjust the curve coefficients in the denoised curve coefficient set to the original curve coefficients multiplied by the confidence weight of the sub-band.

[0105] S303, the first DAS data is generated based on the weighted adjusted set of denoised curve coefficients.

[0106] In some cases, the first DAS data can be obtained by directly performing an inverse curvature transform on the weighted and adjusted set of denoised curvature coefficients.

[0107] In other examples, generating the first DAS data based on the weighted adjusted set of denoised curve coefficients may include: constructing constraints: ,in, For preset reconstruction signals, for The set of curvelet coefficients after curvelet transform. This is the set of denoised curve coefficients after weighted adjustment. The notation for L2 norm calculation. For regularization parameters, To Perform total variational processing; with the goal of minimizing the constraint equation, perform... Optimize and complete the optimization. As the first DAS data.

[0108] The reconstruction process is formalized as an optimization problem, with the objective function containing... (The goal is to reconstruct the signal and ensure consistency with the denoising coefficients) fidelity terms and The total variational regularization term (which aims to make the reconstructed signal smooth and edge-preserving) results in a more physically reasonable output signal. Total variational regularization tends to produce piecewise smooth signals, consistent with the characteristics of the seismic signal phase axis (smooth internally with clear boundaries). Therefore, the reconstruction result not only denoises but also preserves the edges of the effective signal, avoiding blurring. (The text then abruptly shifts to a description of sparse domain denoising.) A small amount of noise remains, and the regularization process can further suppress the noise during reconstruction, improving the fault tolerance and robustness of the entire method. This optimization framework mathematically restricts the reconstructed signal, resulting in a higher quality signal that better meets the requirements of subsequent interpretation than directly performing an inverse transform.

[0109] When reconstructing the signal from the denoised sparse coefficients, a confidence weight is introduced. This weighted reconstruction of subbands at different scales and directions assigns higher weights to subbands rich in effective signal while suppressing noise-dominant subbands. This achieves a second filtering and enhancement during the reconstruction stage, resulting in a final signal with a higher signal-to-noise ratio. The confidence weight is determined based on the average energy of the subbands corresponding to each scale and direction, making it objective and data-driven, thus ensuring the scientific validity and reproducibility of the method.

[0110] In some embodiments, such as Figure 4 As shown, data denoising methods may also include: S401-S409.

[0111] S401, subtract the first DAS data from the DAS data to be processed to obtain the first residual data.

[0112] S402, perform curvelet transform and low threshold denoising on the first residual data, and generate the first effective signal based on the obtained result.

[0113] This low-threshold denoising refers to selecting a lower threshold for denoising the first residual data compared to the threshold used when denoising the DAS data to be processed. For example, using a threshold... To denoise the processed DAS data, the following methods are employed. Denoise the first residual data, and for any... , < .

[0114] S403, the first valid signal is superimposed with the first DAS data to obtain the second DAS data.

[0115] The first valid signal is superimposed on the first DAS data, i.e., the first valid signal + the first DAS data.

[0116] S404, subtract the second DAS data from the DAS data to be processed to obtain the second residual data.

[0117] S405, compare the similarity between the second residual data and each effective signal in the effective signal pattern library corresponding to the mining area environment, and take the part of the second residual data with a similarity greater than the similarity threshold as the second effective signal.

[0118] The effective signal pattern library includes various effective signals in the mining environment. In some examples, comparing the similarity between the second residual data and each effective signal in the effective signal pattern library corresponding to the mining environment may include: for each effective signal, determining a sliding window of the same length as the effective signal, sliding the sliding window through the second residual data, and calculating the similarity between the data within each window and the effective signal.

[0119] If the similarity between the data within the sliding window and the valid signal is greater than the similarity threshold, then the data within the sliding window is taken as the valid signal of the window, and all the valid signals of the window form the second valid signal.

[0120] The similarity threshold is determined based on experience and prior knowledge.

[0121] S406, superimpose the second valid signal with the second DAS data to obtain the third DAS data.

[0122] S407, subtract the third DAS data from the DAS data to be processed to obtain the third residual data.

[0123] S408 removes components from the third residual data that are lower than the physical background noise in the mining area to obtain the third effective signal.

[0124] Among them, physical background noise can be determined based on prior knowledge.

[0125] S409, superimpose the third valid signal with the third DAS data to obtain the target DAS data.

[0126] This embodiment provides a three-stage residual compensation mechanism to progressively recover potentially deleted valid signals from the denoising residuals, thereby better protecting valid signals. Through three coarse-to-fine residual processing steps, it systematically retrieves weak or uniquely shaped valid signals that may have been lost in the main denoising process. The first compensation (low threshold) achieves gentle recovery, targeting high-frequency details that have been over-thresholded. The second compensation (template matching) achieves precise recovery, using prior knowledge (valid signal pattern library) to identify and recover components that conform to typical valid signal patterns. The third compensation (physical threshold) achieves safe recovery, providing a final layer of protection based on the physical background noise level, ensuring that no valid energy above the noise level is lost. Ultimately, this significantly improves the integrity of the target DAS data. For high-value, low signal-to-noise ratio exploration data, preserving every weak signal is crucial, ensuring the high fidelity and integrity of the final data, and providing a solid foundation for subsequent detailed geological interpretation.

[0127] Example 2: Another embodiment of this application relates to a data denoising device. The implementation details of the data denoising device in this embodiment are described below. The following implementation details are provided for ease of understanding and are not essential for implementing this solution. A schematic diagram of the data denoising device in this embodiment can be seen as follows: Figure 5 As shown, it includes: an acquisition module 51, used to acquire distributed acoustic wave sensor (DAS) data to be processed in a mining environment; a processing module 52, used to perform curvelet transform on the DAS data to be processed to obtain a set of curvelet coefficients, the set of curvelet coefficients corresponding to multiple scales, including at least one key scale, the at least one key scale being determined based on the frequency band to which the effective signal in the mining environment belongs, and the density of multiple directions corresponding to each key scale within the key direction range being higher than the density within the non-key direction range, the key direction being determined based on the geological characteristics of the mining environment; a denoising module 53, used to denoise the set of curvelet coefficients to obtain a denoised set of curvelet coefficients; and a generation module 54, used to generate first DAS data based on the denoised set of curvelet coefficients.

[0128] In some optional embodiments, the acquisition module 51 is used to acquire raw DAS data in the mining area environment, the raw DAS data including multiple time series; construct a common mode noise vector based on the preset quantile of the amplitude of the multiple time series at each time point; determine the cross-correlation coefficient between each time series and the common mode noise vector; and remove the product of the cross-correlation coefficient corresponding to the time series and the common mode noise vector from each time series of the raw DAS data to obtain the DAS data to be processed.

[0129] In some optional embodiments, the denoising module 53 is used to denoise the set of curvelet coefficients according to the following hard threshold function:

[0130] in, The curve wave coefficients are the curve wave coefficients in the set of curve wave coefficients. For the set of denoised curved coefficients and The corresponding curve coefficient;

[0131] for , One of them, or is and The product; The geological weights are determined based on the geological characteristics of the mining area. yes The function, For location; The statistical weights are determined based on the local mathematical and statistical properties of the curve coefficients. yes The function, As a scale, For direction; For based on The basic threshold is determined by the statistical distribution of the curvature coefficients within the sub-band.

[0132] In some optional embodiments, the generation module 54 is used to obtain the confidence weight of the sub-band corresponding to each scale-direction in the set of denoised curved coefficients, the confidence weight being determined based on the average energy of the sub-band corresponding to each scale-direction; to perform weighted adjustment on the curved coefficients in the set of denoised curved coefficients based on the confidence weight; and to generate the first DAS data based on the weighted and adjusted set of denoised curved coefficients.

[0133] In some alternative embodiments, the generation module 54 is used to construct the constraint formula: ,in, For preset reconstruction signals, for The set of curvelet coefficients after curvelet transform. This is the set of denoised curve coefficients after weighted adjustment. The notation for L2 norm calculation. For regularization parameters, To Perform total variational processing; with the goal of minimizing this constraint, [the expression is then processed]. Optimize and complete the optimization. This is the first DAS data.

[0134] In some optional embodiments, the apparatus further includes: a compensation module 55, configured to: subtract the first DAS data from the DAS data to be processed to obtain first residual data; perform curve transform and low-threshold denoising on the first residual data, and generate a first valid signal based on the result; superimpose the first valid signal with the first DAS data to obtain second DAS data; subtract the second DAS data from the DAS data to be processed to obtain second residual data; compare the similarity of the second residual data with each valid signal in the valid signal pattern library corresponding to the mining area environment, and take the portion of the second residual data whose similarity with each valid signal is greater than a similarity threshold as the second valid signal; superimpose the second valid signal with the second DAS data to obtain third DAS data; subtract the third DAS data from the DAS data to be processed to obtain third residual data; remove components in the third residual data that are lower than the physical background noise in the mining area environment to obtain a third valid signal; and superimpose the third valid signal with the third DAS data to obtain target DAS data.

[0135] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.

[0136] Example 3: Another embodiment of this application relates to an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the data denoising methods of the above embodiments.

[0137] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0138] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0139] Example 4: Another embodiment of this application relates to a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the method embodiments described above.

[0140] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0141] Example 5: Another embodiment of this application relates to a computer program product, including a computer program. When executed by a processor, the computer program implements the method embodiments described above.

[0142] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.

Claims

1. A data denoising method, characterized in that, include: Acquire distributed acoustic wave sensor (DAS) data to be processed in a mining environment; The DAS data to be processed is subjected to curvelet transform to obtain a set of curvelet coefficients. The set of curvelet coefficients corresponds to multiple scales, including at least one key scale. The at least one key scale is determined based on the frequency band to which the effective signal in the mining area belongs. The density of multiple directions corresponding to each key scale is higher in the key direction range than in the non-key direction range. The key direction is determined based on the geological characteristics of the mining area environment. The set of curved coefficients is denoised to obtain a denoised set of curved coefficients. The first DAS data is generated based on the set of denoised curve coefficients.

2. The data denoising method according to claim 1, characterized in that, The acquisition of distributed acoustic wave sensor (DAS) data to be processed in the mining area environment includes: Obtain raw DAS data in the mining area environment, the raw DAS data including multiple time series; A common-mode noise vector is constructed based on the preset quantile of the amplitude of the multiple time series at each time point; Determine the cross-correlation coefficient between each time series and the common-mode noise vector; From each time series of the original DAS data, the product of the cross-correlation coefficient corresponding to the time series and the common-mode noise vector is removed to obtain the DAS data to be processed.

3. The data denoising method according to claim 1, characterized in that, The denoising process of the set of curved wave coefficients to obtain a denoised set of curved wave coefficients includes: The set of curvelet coefficients is denoised according to the following hard threshold function: in, The curve wave coefficients are the curve wave coefficients in the set of curve wave coefficients. For the set of denoised curved coefficients and The corresponding curve coefficient; for , One of them, or is and The product; The geological weights are determined based on the geological characteristics of the mining area environment. yes The function, For location; The statistical weights are determined based on the local mathematical and statistical properties of the curve coefficients. yes The function, As a scale, For direction; For based on The basic threshold is determined by the statistical distribution of the curvature coefficients within the sub-band.

4. The data denoising method according to claim 1, characterized in that, The generation of the first DAS data based on the denoised curve coefficient set includes: Obtain the confidence weight of the sub-band corresponding to each scale-direction in the set of denoised curved coefficients, wherein the confidence weight is determined based on the average energy of the sub-band corresponding to each scale-direction; The curvature coefficients in the set of denoised curvature coefficients are weighted and adjusted based on the credibility weights. The first DAS data is generated based on the weighted and adjusted set of denoised curve coefficients.

5. The data denoising method according to claim 4, characterized in that, The generation of the first DAS data based on the weighted adjusted set of denoised curve coefficients includes: Constructing constraints: ,in, For preset reconstruction signals, for The set of curvelet coefficients after curvelet transform. The set of weighted and adjusted denoised curve coefficients. The notation for L2 norm calculation. For regularization parameters, To Perform total variational processing; With the goal of minimizing the constraints, for the Optimize and complete the optimization. This serves as the first DAS data.

6. The data denoising method according to claim 1, characterized in that, Also includes: The first residual data is obtained by subtracting the first DAS data from the DAS data to be processed; The first residual data is subjected to curvelet transform and low-threshold denoising, and a first effective signal is generated based on the results. The first valid signal is superimposed on the first DAS data to obtain the second DAS data; The second residual data is obtained by subtracting the second DAS data from the DAS data to be processed; The second residual data is compared with each effective signal in the effective signal pattern library corresponding to the mining area environment. The part of the second residual data with a similarity greater than the similarity threshold with each effective signal is taken as the second effective signal. The second valid signal is superimposed with the second DAS data to obtain the third DAS data; The third residual data is obtained by subtracting the third DAS data from the DAS data to be processed; The third effective signal is obtained by removing components from the third residual data that are lower than the physical background noise in the mining area environment. The third valid signal is superimposed on the third DAS data to obtain the target DAS data.

7. A data denoising device, characterized in that, include: The acquisition module is used to acquire distributed acoustic wave sensor (DAS) data to be processed in the mining environment. The processing module is used to perform curvelet transform on the DAS data to be processed to obtain a set of curvelet coefficients. The set of curvelet coefficients corresponds to multiple scales, including at least one key scale. The at least one key scale is determined based on the frequency band to which the effective signal in the mining area belongs. The density of multiple directions corresponding to each key scale is higher in the key direction range than in the non-key direction range. The key direction is determined based on the geological characteristics of the mining area environment. The denoising module is used to denoise the set of curved coefficients to obtain a denoised set of curved coefficients. The generation module is used to generate first DAS data based on the set of denoised curve coefficients.

8. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the data denoising method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the data denoising method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the data denoising method according to any one of claims 1 to 6.

Images