A denoising method and denoising system for distributed fiber optic sensing data

Abstract

This invention discloses a denoising method and system for distributed optical fiber sensor data, relating to the field of distributed optical fiber sensor data processing technology. The method includes: acquiring distributed optical fiber sensor data; removing low-frequency common-mode noise from the distributed optical fiber sensor data; and removing random noise from the distributed optical fiber sensor data using compressed sensing sparsity technology. Before utilizing compressed sensing sparsity technology to remove random noise from the distributed optical fiber sensor data, low-frequency common-mode noise is removed first. This achieves effective removal of random and common-mode noise while preserving signal details, meeting the denoising requirements of distributed optical fiber sensor data in high-precision application scenarios.

Description

Technical Field

[0001] This invention relates to the field of distributed optical fiber sensor data processing technology, and in particular to a denoising method and denoising system suitable for distributed optical fiber sensor data. Background Technology

[0002] Distributed Acoustic Sensing (DAS) is an innovative sensing method that uses optical fiber as the sensing medium to measure scattered light along the fiber, achieving high-precision, high-resolution monitoring of specific areas. In seismic exploration in mining areas of my country, the terrain is complex and varied, and traditional seismic instruments are costly to deploy. DAS technology, however, utilizes a single optical cable to acquire data over a large area, effectively solving the problems of terrain obstacles and high costs associated with deploying survey lines in mining areas. However, DAS signal acquisition is significantly affected by environmental noise, instrument noise, and transmission loss, resulting in lower data quality and impacting the accuracy of subsequent signal processing and interpretation.

[0003] 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 signal's detailed features, resulting in a decrease in signal fidelity. Summary of the Invention

[0004] To address the technical problems existing in the background art, this invention proposes a denoising method and denoising system suitable for distributed optical fiber sensing data.

[0005] In a first aspect, the present invention proposes a denoising method suitable for distributed optical fiber sensing data, comprising:

[0006] Acquire distributed fiber optic sensing data;

[0007] Remove low-frequency common-mode noise from distributed fiber optic sensing data;

[0008] Random noise in distributed fiber optic sensing data is removed using compressed sensing sparsity techniques.

[0009] Preferably, removing low-frequency common-mode noise from distributed fiber optic sensing data specifically includes:

[0010] Median filtering is used to remove low-frequency common-mode noise from distributed fiber optic sensing data.

[0011] Preferably, removing low-frequency common-mode noise from distributed fiber optic sensing data specifically includes:

[0012] Low-frequency common-mode noise in distributed fiber optic sensing data is removed using mean filtering.

[0013] Preferably, removing low-frequency common-mode noise from distributed fiber optic sensing data specifically includes:

[0014] Low-pass filtering is used to remove low-frequency common-mode noise from distributed fiber optic sensing data.

[0015] Preferably, random noise in distributed fiber optic sensing data is removed using compressed sensing sparsity technology, specifically including:

[0016] The curve transform method is used to remove random noise from distributed fiber optic sensing data.

[0017] Preferably, the curvelet transform method is used to remove random noise from distributed fiber optic sensing data, specifically including:

[0018] The distributed fiber optic sensing data is transformed from the spatiotemporal domain to the sparse domain to obtain the sparse coefficient matrix.

[0019] A threshold function is used to process the sparse coefficient matrix;

[0020] By using inverse curve transform, the sparse coefficient matrix after threshold function processing is restored to the spatiotemporal domain, resulting in distributed fiber optic sensing data with low-frequency common-mode noise and random noise removed.

[0021] Preferably, the threshold function is a hard threshold function.

[0022] Preferably, the hard threshold function is:

[0023] In the formula, T λ Let λ represent the hard threshold function, λ represent the threshold, and x represent the sparse coefficient matrix.

[0024] Preferably, the curvelet transform method is a second-generation curvelet transform method, and the curvelet coefficients of this curvelet transform method are:

[0025]

[0026] In the formula, c(j,l,k) represents the curvature coefficient, j represents the scale, l represents the direction angle, k represents the position parameter, and f represents the signal. Describe the basis functions;

[0027] The inverse transform form of the curvelet transform method is as follows:

[0028]

[0029] In the formula, This represents the curve coefficients after thresholding. This represents the inflection wave basis function.

[0030] Secondly, the present invention also proposes a denoising system suitable for distributed optical fiber sensing data, comprising:

[0031] The acquisition module is used to acquire distributed fiber optic sensing data;

[0032] The first noise reduction module is used to remove low-frequency common-mode noise from distributed fiber optic sensing data.

[0033] The second denoising module is used to remove random noise from distributed fiber optic sensing data using compressed sensing sparsity technology.

[0034] The proposed denoising method and system for distributed fiber optic sensing data in this invention removes low-frequency common-mode noise from the distributed fiber optic sensing data before utilizing compressed sensing sparsity technology to remove random noise. This lays the foundation for subsequent compressed sensing sparse reconstruction, helps improve the accuracy and stability of sparse domain processing, and effectively removes random noise and common-mode noise while preserving signal details, thus meeting the denoising requirements of distributed fiber optic sensing data in high-precision application scenarios. Attached Figure Description

[0035] Figure 1 This is a flowchart illustrating a denoising method for distributed optical fiber sensing data proposed in this invention. Detailed Implementation

[0036] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0037] Firstly, referring to Figure 1 The present invention proposes a denoising method for distributed optical fiber sensing data, comprising:

[0038] Acquire distributed fiber optic sensing (DAS) data;

[0039] Remove low-frequency common-mode noise from distributed fiber optic sensing data;

[0040] Random noise in distributed fiber optic sensing data is removed using compressed sensing sparsity techniques.

[0041] This invention lays the foundation for subsequent compressed sensing sparse reconstruction by removing low-frequency common-mode noise from distributed fiber optic sensing data before using compressed sensing sparsity technology to remove random noise. This helps improve the accuracy and stability of sparse domain processing and achieves effective removal of random and common-mode noise while preserving signal details, thus meeting the denoising requirements of distributed fiber optic sensing data in high-precision application scenarios.

[0042] In one specific embodiment, removing low-frequency common-mode noise from distributed fiber optic sensing data specifically includes:

[0043] Median filtering is used to remove low-frequency common-mode noise from distributed fiber optic sensing data, thereby eliminating in-channel uniformity noise while preserving nonlinear abrupt changes in the data.

[0044] In another specific embodiment, removing low-frequency common-mode noise from distributed fiber optic sensing data specifically includes:

[0045] The mean filtering method is used to remove low-frequency common-mode noise from distributed fiber optic sensing data, so as to reduce the impact of common-mode noise on the overall signal through smoothing.

[0046] In addition, in certain specific embodiments, removing low-frequency common-mode noise from distributed fiber optic sensing data specifically includes:

[0047] Low-pass filtering is used to remove low-frequency common-mode noise from distributed fiber optic sensing data.

[0048] In this embodiment, random noise in distributed fiber optic sensing data is removed using compressed sensing sparsity technology, specifically including:

[0049] The curve transform method is used to remove random noise from distributed fiber optic sensing data.

[0050] To remove random noise from distributed fiber optic sensing data, this embodiment employs curvelet transform to remove random noise from the distributed fiber optic sensing data, specifically including:

[0051] The distributed fiber optic sensing data is transformed from the spatiotemporal domain to the sparse domain to obtain the sparse coefficient matrix.

[0052] A threshold function is used to process the sparse coefficient matrix;

[0053] Then, the sparse coefficient matrix after thresholding is restored to the spatiotemporal domain by inverse curve transform, resulting in distributed fiber optic sensing data with low-frequency common-mode noise and random noise removed.

[0054] In this embodiment, the curvelet transform method used is the second-generation curvelet transform method. It makes up for the shortcomings of wavelet transform in representing line singularities, and can obtain a sparser representation than wavelet transform. It is simpler to implement than the first-generation curvelet transform and can achieve a near-optimal representation for smooth singular curves.

[0055] In this embodiment, the sparse representation of the second-generation curvelet transform is implemented in the following form, where the curvelet coefficients can be calculated from the signal f and the basis functions. The inner product representation is as follows:

[0056]

[0057] In the formula, c(j,l,k) represents the curvature coefficient, j represents the scale, l represents the direction angle, and k represents the position parameter.

[0058] The inverse transform form of the second-generation curvelet transform is:

[0059]

[0060] In the formula, This represents the curve coefficients after thresholding. This represents the inflection wave basis function.

[0061] In this embodiment, the threshold function is a hard threshold.

[0062] Curves transform is a type of sparse transform based on a fixed basis. While fixed-basis sparse transforms suffer from poor adaptability and limited expression, they are simple to implement, computationally fast, and advantageous for processing large datasets. After transforming distributed fiber optic sensing data from the spatiotemporal domain to the sparse domain, a sparse coefficient matrix is ​​obtained. By applying a hard threshold function to this sparse matrix and setting a reasonable threshold, the sparse coefficient matrix is ​​thresholded, and then inversely transformed back to the spatiotemporal domain.

[0063] Wherein, the hard threshold function is

[0064] In the formula, T λ Let λ represent the hard threshold function, λ represent a reasonable threshold, and x represent the sparse coefficient matrix obtained after sparse transformation of distributed fiber optic sensing data.

[0065] The hard thresholding function sets values ​​with absolute values ​​less than or equal to a threshold λ to 0, while values ​​greater than λ remain unchanged. The threshold setting is crucial for denoising; a smaller threshold represents weaker sparsity constraints, insufficiently removing noise and resulting in denoised data still containing significant noise, while a larger threshold represents stronger sparsity constraints, potentially damaging the effective signal. Therefore, the threshold setting should reasonably balance denoising strength and signal fidelity. The soft thresholding function is the counterpart to the hard thresholding function, but it modifies all data, often resulting in greater bias. The hard thresholding function, on the other hand, crudely sets sparse coefficients with absolute values ​​less than or equal to the threshold λ to 0, causing abrupt changes in the inverse-transformed data. However, the hard thresholding function typically yields a higher signal-to-noise ratio. Therefore, considering all factors, the hard thresholding function is chosen. The sparse coefficient matrix after hard thresholding is then restored to the spatiotemporal domain using an inverse warp transform, reconstructing the distributed fiber optic sensing data.

[0066] Therefore, the curvelet transform method is used to remove random noise from distributed fiber optic sensing data, specifically including:

[0067] The distributed fiber optic sensing data is transformed from the spatiotemporal domain to the sparse domain to obtain the sparse coefficient matrix.

[0068] A hard thresholding function is used to process the sparse coefficient matrix;

[0069] Then, the sparse coefficient matrix after hard thresholding is restored to the spatiotemporal domain by inverse curve transform, resulting in distributed fiber optic sensing data with low-frequency common-mode noise and random noise removed.

[0070] This embodiment can preserve the geometric features of the original signal while significantly reducing the impact of noise, thereby improving the clarity and signal-to-noise ratio of the data.

[0071] Secondly, the present invention also proposes a denoising system suitable for distributed optical fiber sensing data, comprising:

[0072] The acquisition module is used to acquire distributed fiber optic sensing data;

[0073] The first noise reduction module is used to remove low-frequency common-mode noise from distributed fiber optic sensing data.

[0074] The second denoising module is used to remove random noise from distributed fiber optic sensing data using compressed sensing sparsity technology.

[0075] This embodiment uses the cooperation of the acquisition module, the first denoising module, and the second denoising module to denoise distributed optical fiber sensing data. This achieves effective removal of random noise and common-mode noise in distributed optical fiber sensing data while preserving signal details, thus meeting the denoising requirements of distributed optical fiber sensing data in high-precision application scenarios.

[0076] The first denoising module uses median filtering, mean filtering, or low-pass filtering to remove low-frequency common-mode noise from distributed optical fiber sensing data.

[0077] Specifically, the compressed sensing sparsity technique is the second-generation curvelet transform method, whose curvelet coefficients are:

[0078]

[0079] In the formula, c(j,l,k) represents the curvature coefficient, j represents the scale, l represents the direction angle, k represents the position parameter, and f represents the signal. Describe the basis functions;

[0080] The inverse transform form of the curvelet transform method is as follows:

[0081]

[0082] In the formula, This represents the curve coefficients after thresholding. This represents the inflection wave basis function.

[0083] The second denoising module, which uses curvelet transform to remove random noise from distributed fiber optic sensing data, specifically includes the following steps:

[0084] The distributed fiber optic sensing data is transformed from the spatiotemporal domain to the sparse domain to obtain the sparse coefficient matrix.

[0085] A threshold function is used to process the sparse coefficient matrix;

[0086] Then, the sparse coefficient matrix after thresholding is restored to the spatiotemporal domain by inverse curve transform, resulting in distributed fiber optic sensing data with low-frequency common-mode noise and random noise removed.

[0087] The threshold function in this embodiment is a hard threshold function.

[0088] Specifically, the hard threshold function is:

[0089] In the formula, T λ Let λ represent the hard threshold function, λ represent the threshold, and x represent the sparse coefficient matrix.

[0090] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A denoising method for distributed fiber optic sensing data, characterized in that, include: Acquire distributed fiber optic sensing data; Remove low-frequency common-mode noise from distributed fiber optic sensing data; Random noise in distributed fiber optic sensing data is removed using compressed sensing sparsity techniques.

2. The denoising method for distributed optical fiber sensor data according to claim 1, characterized in that, Removing low-frequency common-mode noise from distributed fiber optic sensing data specifically includes: Median filtering is used to remove low-frequency common-mode noise from distributed fiber optic sensing data.

3. The denoising method for distributed optical fiber sensing data according to claim 1, characterized in that, Removing low-frequency common-mode noise from distributed fiber optic sensing data specifically includes: Low-frequency common-mode noise in distributed fiber optic sensing data is removed using mean filtering.

4. The denoising method for distributed optical fiber sensing data according to claim 1, characterized in that, Removing low-frequency common-mode noise from distributed fiber optic sensing data specifically includes: Low-pass filtering is used to remove low-frequency common-mode noise from distributed fiber optic sensing data.

5. The denoising method for distributed optical fiber sensing data according to claim 1, characterized in that, The random noise in distributed fiber optic sensing data is removed using compressed sensing sparsity techniques, specifically including: The curve transform method is used to remove random noise from distributed fiber optic sensing data.

6. The denoising method for distributed optical fiber sensing data according to claim 5, characterized in that, The curvelet transform method is used to remove random noise from distributed fiber optic sensing data, specifically including: The distributed fiber optic sensing data is transformed from the spatiotemporal domain to the sparse domain to obtain the sparse coefficient matrix. A threshold function is used to process the sparse coefficient matrix; By using inverse curve transform, the sparse coefficient matrix after threshold function processing is restored to the spatiotemporal domain, resulting in distributed fiber optic sensing data with low-frequency common-mode noise and random noise removed.

7. The denoising method for distributed optical fiber sensing data according to claim 6, characterized in that, The threshold function is a hard threshold function.

8. The denoising method for distributed optical fiber sensing data according to claim 7, characterized in that, The hard threshold function is In the formula, T λ Let λ represent the hard threshold function, λ represent the threshold, and x represent the sparse coefficient matrix.

9. The denoising method for distributed optical fiber sensing data according to claim 5, characterized in that, This curvelet transform method is a second-generation curvelet transform method, and its curvelet coefficients are: In the formula, c(j,l,k) represents the curvature coefficient, j represents the scale, l represents the direction angle, k represents the position parameter, and f represents the signal. Describe the basis functions; The inverse transform form of the curvelet transform method is: In the formula, This represents the curve coefficients after thresholding. This represents the inflection wave basis function.

10. The denoising system for distributed optical fiber sensor data according to claim 1, characterized in that, include: The acquisition module is used to acquire distributed fiber optic sensing data; The first noise reduction module is used to remove low-frequency common-mode noise from distributed fiber optic sensing data. The second denoising module is used to remove random noise from distributed fiber optic sensing data using compressed sensing sparsity technology.

Images