Cable external damage prevention signal noise reduction method, system and terminal

Abstract

The invention relates to the technical field of electric power safety monitoring, and particularly discloses a cable external damage prevention signal noise reduction method and system and a terminal, and the method comprises the steps: carrying out the preprocessing of an original disturbance phase signal, so as to eliminate a part of noise and low-frequency drift in an original signal, and dynamically extracting the disturbance information in the signal, so as to achieve the noise reduction. Namely, the wavelet basis is dynamically selected for wavelet analysis, and the mismatch problem of the traditional fixed wavelet basis is relieved in combination with a decomposition layer number self-adaptive strategy. Meanwhile, an auxiliary scheme based on deep learning is provided, and the adaptive noise reduction capability in a complex scene is enhanced.

Description

Technical Field

[0001] The present application relates to the technical field of power safety monitoring, and more specifically, to a method, system, and terminal for reducing noise in signals for preventing external damage to cables. Background Art

[0002] Due to the characteristic that it is difficult to directly locate faults in buried cables, in the prior art, a distributed fiber optic vibration sensing technology is adopted, and sensing optical fibers are laid along the buried cables. By detecting the laser phase change in the sensing optical fibers, it is determined whether an external damage event occurs outside the cables.

[0003] However, the signal containing disturbance information is vulnerable to external noise. For example, the fiber optic sensing system is vulnerable to environmental temperature drift, mechanical vibration noise, and the noise of the optical path itself, resulting in a decrease in the signal-to-noise ratio (SNR), and weak disturbance signals are easily submerged by noise.

[0004] To realize the subsequent analysis of the disturbance signal and the identification of external damage events, it is necessary to denoise the disturbance signal to extract useful information.

[0005] Traditional denoising methods rely on using fixed wavelet bases and fixed thresholds to process the disturbance phase signal. However, different types of external damage events result in different time-frequency characteristics of the disturbance phase signal, which may lead to the loss of effective disturbance information. Therefore, an optimized technical solution needs to be proposed. Summary of the Invention

[0006] The technical problem to be solved by the present application is: to provide a method, system, and terminal for reducing noise in signals for preventing external damage to cables, which solves the problem in the prior art that the disturbance information extraction strategy cannot be adaptively changed for different types of external damage events.

[0007] The technical problem to be solved by the present application is achieved by the following technical solutions:

[0008] In a first aspect, the present application provides a method for reducing noise in signals for preventing external damage to cables, including: obtaining an original disturbance phase signal collected by a distributed fiber optic vibration sensing module based on φ-OTDR; preprocessing the original disturbance phase signal to obtain a preprocessed disturbance phase signal; dynamically extracting and denoising disturbance information from the preprocessed disturbance phase signal to obtain a restored disturbance denoised signal.

[0009] Further, preprocessing the original disturbance phase signal to obtain a preprocessed disturbance phase signal includes: preprocessing the original disturbance phase signal by using moving average filtering and / or phase difference to obtain the preprocessed disturbance phase signal;

[0010] Among them, the following differential formula is used to perform phase difference processing on the input signal to obtain the output signal; where the differential formula is:

[0011]

[0012] Among them, the value of t ranges from 1 to the length of the input signal, is the value of the output signal at time t, is the value of the input signal at time t, is the value of the input signal at time t - 1, is the value of the output signal at the initial time, is the value of the input signal at the initial time.

[0013] Furthermore, dynamic extraction and noise reduction of the disturbance information are performed on the preprocessed disturbance phase signal to obtain the restored disturbance noise reduction signal, including: dynamically selecting the wavelet basis and the decomposition level, performing wavelet decomposition on the preprocessed disturbance phase signal to obtain multiple detail coefficients; performing hierarchical threshold combination processing on the multiple detail coefficients; based on the multiple detail coefficients after hierarchical threshold combination processing, using dual-tree complex wavelet to reconstruct the noise reduction signal and embedding the phase compensation factor to obtain the restored disturbance noise reduction signal.

[0014] Furthermore, dynamically selecting the wavelet basis and the decomposition level, performing wavelet decomposition on the preprocessed disturbance phase signal to obtain multiple detail coefficients, including: determining the signal category to which the preprocessed disturbance phase signal belongs based on the signal difference degree between the preprocessed disturbance phase signal and each group of sample disturbance phase signals in the sample library; selecting the wavelet basis based on the signal category, performing wavelet decomposition on the preprocessed disturbance phase signal to obtain the multiple detail coefficients; where, during the wavelet decomposition process, an adaptive strategy for the decomposition level is used to control whether the wavelet decomposition terminates.

[0015] Furthermore, the signal category at least includes one of the following: instantaneous impact, continuous vibration, and mixed noise; where, when the signal category to which the preprocessed disturbance phase signal belongs is instantaneous impact, select db4 wavelet and coif3 wavelet as the wavelet basis; when the signal category to which the preprocessed disturbance phase signal belongs is continuous vibration, select sym4 wavelet as the wavelet basis; when the signal category to which the preprocessed disturbance phase signal belongs is mixed noise, select bior3.3 wavelet as the wavelet basis.

[0016] Further, during the wavelet decomposition process, an adaptive decomposition layer strategy is adopted to control whether the wavelet decomposition terminates, including: if the current decomposition layer is less than 2, continue the decomposition; if the current decomposition layer is greater than or equal to 3, then perform the following steps: determine the termination threshold according to the signal category to which the preprocessed perturbed phase signal belongs; calculate the energy change rate of the current decomposition layer; when the energy change rate is less than the termination threshold, terminate the decomposition, otherwise continue the decomposition;

[0017] Among them, the energy change rate is calculated by the following energy change rate quantization formula; the energy change rate quantization formula is:

[0018]

[0019] Among them, is the k-th detail coefficient of the j-th layer, is the number of detail coefficients of the j-th layer, |·| represents calculating the absolute value, is the energy of the j-th layer, is the energy of the (j - 1)-th layer, is the energy change rate of the j-th layer relative to the (j - 1)-th layer;

[0020] Among them, the value of j is restricted by the maximum layer, which is expressed by the formula:

[0021]

[0022] Among them, N is the total number of detail coefficients, is the maximum number of layers, represents the logarithmic function operation with base 2.

[0023] Further, hierarchical threshold combination processing is performed on the multiple detail coefficients, including: performing hard threshold processing on the detail coefficients at the decomposition layer of 1; performing semi-soft threshold processing on the detail coefficients at the decomposition layer of 2 or 3; performing soft threshold processing on the detail coefficients at the decomposition layer greater than or equal to 4.

[0024] Further, perform dynamic extraction and noise reduction of disturbance information on the preprocessed disturbance phase signal to obtain a restored disturbance noise reduction signal, including: processing the preprocessed disturbance phase signal using a multi-granularity feature residual stripper to obtain a first-granularity disturbance feature map, a second-granularity disturbance feature map, and a third-granularity disturbance feature map; inputting the first-granularity disturbance feature map and the second-granularity disturbance feature map into a convolutional neural network using a spatial attention mechanism to obtain a first disturbance enhancement feature map and a second disturbance enhancement feature map; processing the first disturbance enhancement feature map, the second disturbance enhancement feature map, and the third-granularity disturbance feature map using a bidirectional dynamic fusion device to obtain a multi-granularity disturbance feature map; and processing the multi-granularity disturbance feature map using a decoder-based noise reducer to obtain the restored disturbance noise reduction signal.

[0025] In a second aspect, the present application further provides a cable anti-external break signal noise reduction system, including: a disturbance signal acquisition module for acquiring an original disturbance phase signal collected by a distributed optical fiber vibration sensing module based on φ-OTDR; a preprocessing module for preprocessing the original disturbance phase signal to obtain a preprocessed disturbance phase signal; and a noise reduction module for performing dynamic extraction and noise reduction of disturbance information on the preprocessed disturbance phase signal to obtain a restored disturbance noise reduction signal.

[0026] In a third aspect, the present application further provides a terminal, including: a processor for being coupled to a memory and reading and executing instructions stored in the memory; and when the processor runs, executing the instructions such that the processor is used to execute the cable anti-external break signal noise reduction method.

[0027] The present application includes at least one of the following beneficial technical effects:

[0028] 1. By preprocessing the original disturbance phase signal to eliminate part of the noise and low-frequency drift in the original signal, a more accurate data source is provided for the extraction of disturbance features.

[0029] 2. During the wavelet analysis process, dynamically select the wavelet basis and the decomposition level to alleviate the mismatch problem of traditional fixed wavelet bases.

[0030] 3. Provide an auxiliary solution based on deep learning to enhance the adaptive noise reduction ability in complex scenarios. Description of the Drawings

[0031] Figure 1 It is a schematic flowchart of the cable anti-external break signal noise reduction method provided by the present application.

[0032] Figure 2 It is a schematic structural diagram of the distributed optical fiber vibration sensing module based on φ-OTDR provided by the present application.

[0033] Figure 3A Flow diagram for obtaining the restored disturbance noise reduction signal provided by this application Figure 1 。

[0034] Figure 3B Flow diagram for obtaining the restored disturbance noise reduction signal provided by this application Figure 2 。

[0035] Figure 4 Structural schematic diagram of the cable anti-external damage signal noise reduction system provided by this application.

[0036] Figure 5 Structural schematic diagram of the terminal provided by this application.

[0037] In the figure:

[0038] Disturbance signal acquisition module 110; preprocessing module 120; noise reduction module 130;

[0039] Processor 10; memory 20; instruction 30. Specific implementation manners

[0040] To facilitate understanding of the technical means, creative features, achieved objectives and effects of this application, the following further elaborates this application in conjunction with specific illustrations.

[0041] Due to its advantages of convenient construction and space intensiveness, buried cables have become the mainstream form of cable laying in modern cities. As the scale of the underground cable network continues to expand, the risk of external damage is also on the rise.

[0042] Cable faults caused by external damage factors such as mechanical construction misoperation and geological settlement will not only cause direct economic losses, but also lead to large-scale power outages, affecting industrial and commercial operations and people's livelihood security.

[0043] In the prior art, the distributed optical fiber vibration sensing technology is adopted to judge whether an external damage event occurs outside the cable by detecting the laser phase change in the sensing optical fiber.

[0044] However, the signal containing disturbance information is vulnerable to external noise, and weak disturbance signals are easily submerged by noise.

[0045] To realize the subsequent analysis of the disturbance signal and the identification of external damage events, it is necessary to reduce the noise of the disturbance signal to extract useful information.

[0046] Traditional noise reduction methods rely on using fixed wavelet bases and fixed thresholds to process the disturbance phase signal. However, different types of external damage events result in different time-frequency characteristics of the disturbance phase signal, which may lead to the loss of effective disturbance information during the noise reduction process.

[0047] Such as Figure 1As shown in the figure, the present application provides a method for reducing noise of signals for preventing external damage of cables to solve the above technical problems. The specific steps include:

[0048] S1. Obtain the original disturbance phase signal collected by the distributed fiber optic vibration sensing module based on φ-OTDR.

[0049] S2. Preprocess the original disturbance phase signal to obtain the preprocessed disturbance phase signal.

[0050] S3. Dynamically extract and reduce noise of the disturbance information from the preprocessed disturbance phase signal to obtain the restored disturbance noise-reduced signal.

[0051] As Figure 2 shown in the figure, the distributed fiber optic vibration sensing module based on φ-OTDR includes a narrow linewidth laser source, a first coupler, an electro-optic modulator, an acousto-optic modulator, a signal amplifier, a filter, a circulator, a sensing optical fiber, a second coupler, a balanced detector, a photodetector, and an analog-to-digital converter.

[0052] Among them, the distributed fiber optic vibration sensing module based on φ-OTDR (Phase-sensitive Optical Time Domain Reflectometry) detects external vibration events by collecting the modulation signal of the backward scattered light in the optical fiber.

[0053] Due to the elasto-optic effect and the strain effect, external vibration will cause slight changes in the refractive index or length of the optical fiber, thereby affecting the phase information of the backward scattered light signal in the optical fiber. In the technical solution of the present application, obtaining the original disturbance phase signal collected by the distributed fiber optic vibration sensing module based on φ-OTDR can provide a data source for the dynamic extraction of disturbance information.

[0054] Specifically, a narrow-linewidth laser source is used to generate a continuous optical signal; a first coupler is connected to the narrow-linewidth laser source and is used to split the continuous optical signal into two parts: a local optical signal and a signal optical signal; an electro-optic modulator is connected to the first coupler and is used to process the signal optical signal via the electro-optic modulator for electro-optic modulation; an acousto-optic modulator is connected to the electro-optic modulator and is used to send the signal into the acousto-optic modulator for chopping and frequency shifting; a signal amplifier is connected to the acousto-optic modulator and is used to enhance the optical power of the signal; a filter is connected to the signal amplifier and is used to remove the spontaneous emission noise generated during the amplification process from the amplified signal; a circulator is respectively connected to the filter and the sensing optical fiber and is used to inject the noise-reduced signal into the sensing optical fiber laid along the cable; a second coupler is respectively connected to the first coupler and the circulator and is used to perform beat frequency between the scattered light and the local optical signal; a balanced detector is connected to the second coupler and is used to reduce the noise of the beat frequency signal; a photodetector is connected to the balanced detector and is used to convert the noise-reduced signal into an electrical signal; an analog-to-digital converter is connected to the photodetector and is used to convert the analog electrical signal into a digital electrical signal (original perturbation phase signal).

[0055] Next, preprocessing is performed on the original perturbation phase signal to eliminate part of the noise and low-frequency drift in the original signal and provide high-quality data for subsequent processing.

[0056] Specifically, the original perturbation phase signal is preprocessed by using moving average filtering and / or phase difference to obtain a preprocessed perturbation phase signal.

[0057] Here, moving average filtering aims to smooth the signal and reduce the influence of high-frequency random noise (such as electronic noise and environmental interference) on the cable line perimeter.

[0058] Among them, moving average filtering smooths the signal by calculating the mean value of the data within the window. The formula is as follows:

[0059]

[0060] Among them, L is the window length, that is, the number of data within the window, is the (n + 1)-th data within the signal window after moving average filtering, and s[n - i] is the (n - i + 1)-th data within the original input signal window.

[0061] Among them, the window length can be dynamically adjusted according to the signal characteristics in practical applications. Generally, the selection of the window length depends on the signal sampling rate and the highest frequency.

[0062] Preferably, in the technical solution of this application, the window length is an odd number, and this odd number is greater than the quotient of the signal sampling rate and twice the highest frequency. Here, an odd number is preferentially selected to avoid phase shift, and zero padding is performed on the signal to handle the boundary effect.

[0063] Here, the phase difference is aimed at removing the low-frequency trend terms caused by factors such as temperature drift and slow changes in the optical path, so as to highlight the high-frequency vibration signal.

[0064] In the technical solution of this application, the following difference formula is used to perform phase difference processing on the input signal to obtain the output signal; where the difference formula is:

[0065]

[0066] where the value of t ranges from 1 to the length of the input signal, is the value of the output signal at time t, is the value of the input signal at time t, is the value of the input signal at time t - 1.

[0067] where, is the value of the output signal at the initial time, is the value of the input signal at the initial time. That is, the value at the initial time is filled with zeros or the original value is retained.

[0068] It is worth mentioning that since there is not enough historical data at the beginning stage of the signal to calculate to fill the window, if not processed, it will lead to the length of the output signal being shorter than the input signal, resulting in incomplete boundary data.

[0069] Common filling methods are zero-padding or truncating the window. Among them, zero-padding sets the missing data points to 0 to keep the length of the output signal consistent with the input signal. Truncating the window can use the initial value of the input signal as the initial value of the output signal, that is, retain the original value. Its advantage is to avoid introducing false data and retain the characteristics of the original signal, which is suitable for scenarios with high requirements for boundary accuracy.

[0070] The core purpose of filling the boundary initial value is to ensure that the output length of the filter is the same as the input and there is a smooth transition at the boundary. In the actual operation process, the specific filling method needs to be selected in combination with the signal characteristics and application scenarios.

[0071] This difference processing is equivalent to high-pass filtering and can suppress low-frequency components (frequency f < fs / 2).

[0072] Then, in a specific embodiment of this application, as Figure 3A shown, the following steps are executed to dynamically extract and denoise the perturbation information from the preprocessed perturbation phase signal:

[0073] S301. Dynamically select the wavelet basis and the decomposition level, and perform wavelet decomposition on the preprocessed perturbation phase signal to obtain multiple detail coefficients;

[0074] S302. Perform hierarchical threshold combination processing on the multiple detail coefficients;

[0075] S303. Based on multiple detail coefficients after hierarchical threshold combination processing, use dual-tree complex wavelet reconstruction to denoise the signal and embed a phase compensation factor to obtain a restored perturbed denoised signal.

[0076] It should be understood that in cable anti-external damage monitoring, the vibration signals of different external damage events (such as instantaneous impact, continuous vibration, and mixed noise) have significant time-frequency characteristic differences. Adopt the dynamic wavelet decomposition method to optimize the decomposition effect for different signal types through the combination of hybrid wavelet bases and hierarchical strategies.

[0077] The specific implementation method of step S301 includes: First, determine the signal category to which the preprocessed perturbed phase signal belongs (including at least one of instantaneous impact, continuous vibration, and mixed noise) based on the signal difference degree between the preprocessed perturbed phase signal and each group of sample perturbed phase signals in the sample library.

[0078] Next, select a wavelet basis based on the signal category and perform wavelet decomposition on the preprocessed perturbed phase signal to obtain multiple detail coefficients.

[0079] In the embodiments of the present application, the signal difference degree is calculated using correlation analysis, statistical methods, or signal distance measurement methods.

[0080] More specifically, correlation analysis measures the relationship between two signals by calculating the correlation coefficient between the two signals, such as the Pearson correlation coefficient or the Spearman correlation coefficient.

[0081] Statistical methods, such as mean, variance, standard deviation, etc., are used to quantify the statistical characteristics of the signal to compare the statistical differences between two signals.

[0082] Signal distance measures the difference between them based on the numerical difference between the signal waveforms. For example, calculate the average value of the absolute value of the difference between two signals at corresponding moments, or use more complex distance measurement methods, such as Euclidean distance, Manhattan distance, etc.

[0083] Each group of sample perturbed phase signals in the sample library is generated by simulation software or obtained through actual acquisition. The sample perturbed phase signals in the sample library correspond to at least one of instantaneous impact, continuous vibration, and mixed noise.

[0084] Among them, when the signal category to which the preprocessed perturbed phase signal belongs is instantaneous impact, select db4 wavelet and coif3 wavelet as the wavelet basis.

[0085] When the signal category to which the preprocessed perturbed phase signal belongs is continuous vibration, select sym4 wavelet as the wavelet basis.

[0086] When the signal category to which the preprocessed perturbation phase signal belongs is mixed noise, the bior3.3 wavelet is selected as the wavelet basis.

[0087] Specifically, instantaneous impacts (such as excavation) have time-frequency characteristics of high frequency, short-time mutation, and energy concentration. The db4 wavelet basis has core characteristics of a compact support (short filter length), 4th-order vanishing moments, and asymmetry. Coif3 has core characteristics of approximate symmetry, a longer support (18 points), and 6th-order vanishing moments.

[0088] Thus, in the technical solution of this application, first, the instantaneous impact signal is shallowly decomposed (db4 decomposed to 3 layers), quickly extracting high-frequency transient components (D1 - D3), capturing the rising edge and peak of the impact signal, and retaining the low-frequency approximation (A3) for subsequent deep refinement.

[0089] For example, set the decomposition level to 3 layers (covering the 0 - 500 Hz frequency band, assuming a sampling rate of 1 kHz). The output results are high-frequency detail coefficients D1 (500 - 250 Hz), D2 (250 - 125 Hz), D3 (125 - 62.5 Hz), and the low-frequency approximation coefficient: A3 (0 - 62.5 Hz).

[0090] Then, deep refinement is performed (coif3 decomposes A3 to 3 layers), further decomposing A3 to extract potentially hidden weak impact features (such as low-frequency vibration afterwaves), generating deep high-frequency details (D4 - D6) and the final low-frequency approximation (A6).

[0091] For example, set the decomposition level to 3 layers (covering the sub-band of 0 - 62.5 Hz). The output results are deep high-frequency details D4 (62.5 - 31.25 Hz), D5 (31.25 - 15.625 Hz), D6 (15.625 - 7.8125 Hz), and the final low-frequency approximation A6 (0 - 7.8125 Hz).

[0092] Finally, coefficient merging is carried out, adopting a merging strategy of merging the shallow high-frequency details (D1 - D3) with the deep high-frequency details (D4 - D6), such as weighted averaging, to retain all high-frequency information, and the low-frequency approximation (A6) is directly used for reconstruction.

[0093] Specifically, continuous vibrations (such as vehicle rolling) have time-frequency characteristics of low frequency, periodicity, and smooth continuity. The sym4 wavelet basis has core characteristics of approximate symmetry, high vanishing moments (4th order), and a medium support length (8 points).

[0094] Thus, in the technical solution of this application, the continuous vibration signal is decomposed 4 times based on the sym4 wavelet basis (covering the 0 - 62.5 Hz frequency band when the sampling rate is 1 kHz).

[0095] For example, the frequency band is divided into D1 (500 - 250 Hz), D2 (250 - 125 Hz), D3 (125 - 62.5 Hz), D4 (62.5 - 31.25 Hz), and A4 (0 - 31.25 Hz), and the low - frequency periodic components are retained.

[0096] Among them, the low - pass lpd and high - pass hpd are used for decomposition filtering, and symmetric extension (mode ='sym') reduces the edge effect. It reduces the phase distortion during signal decomposition and reconstruction, effectively suppresses low - frequency interference, accurately extracts the main frequency components of continuous vibration, and balances the time - frequency resolution.

[0097] Specifically, the mixed noise (complex environment) has the time - frequency characteristics of multi - band, non - stationary, and composite interference. Bior3.3 has the core characteristics of biorthogonality (the decomposition and reconstruction filters are independently designed and support linear phase), long support (10 - point decomposition / 6 - point reconstruction, high - frequency band subdivision ability), and 3 - order vanishing moment (balancing high - frequency noise suppression and weak signal retention).

[0098] In this way, in the technical solution of this application, a 5 - layer decomposition based on the bior3.3 wavelet basis is performed on the mixed - noise signal (covering the 0 - 15.625 Hz low - frequency band when the sampling rate is 1 kHz) to separate the low - frequency geological vibration and high - frequency wind - rain noise.

[0099] For example, the frequency band is divided into D1 (500 - 250 Hz), D2 (250 - 125 Hz), D3 (125 - 62.5 Hz), D4 (62.5 - 31.25 Hz), D5 (31.25 - 15.625 Hz), A5 (0 - 15.625 Hz). Among them, A5 contains the low - frequency components of geological activities.

[0100] Among them, the lo_d (low - pass decomposition) and hi_d (high - pass decomposition) of the decomposition filter bior3.3, and the lo_r (low - pass reconstruction) and hi_r (high - pass reconstruction) of the reconstruction filter bior3.3 are established to adapt to the complex noise environment, avoid reconstruction distortion, and balance high - frequency noise suppression and weak signal retention.

[0101] In addition, in the technical solution of this application, a decomposition - layer adaptive strategy is adopted during wavelet decomposition to control whether wavelet decomposition terminates.

[0102] More specifically, it includes:

[0103] If the current decomposition layer is less than 2, continue decomposition;

[0104] If the current decomposition layer is greater than or equal to 3, then perform the following steps:

[0105] Determine the termination threshold according to the signal category to which the preprocessed perturbed phase signal belongs;

[0106] Calculate the energy change rate of the current decomposition layer;

[0107] When the energy change rate is less than the termination threshold, the decomposition terminates; otherwise, continue the decomposition;

[0108] Among them, the energy change rate is calculated by the following energy change rate quantization formula; the energy change rate quantization formula is:

[0109]

[0110] Among them, is the k-th detail coefficient of the j-th layer, is the number of detail coefficients of the j-th layer, |·| represents calculating the absolute value, is the energy of the j-th layer, is the energy of the (j - 1)-th layer, is the energy change rate of the j-th layer relative to the (j - 1)-th layer.

[0111] Here, the energy gradient method is used to establish an adaptive strategy for the number of decomposition layers to automatically adapt to the signal characteristics of different cable external damage events. In each layer of wavelet decomposition, the energy of the detail coefficients (high-frequency components) is used to measure the activity of the signal at that layer. In the technical solution of this application, represents the energy of the j-th layer, and represents the energy change rate between adjacent decomposition layers to determine whether to continue the decomposition.

[0112] In a specific example, when ΔE j <the termination threshold (such as 5%), it is considered that the energy change tends to be stable and the decomposition stops. Among them, the adaptive decomposition process first performs an initial decomposition and performs at least 3 layers of decomposition to avoid premature termination.

[0113] At the same time, a constraint is placed on the maximum number of layers:

[0114]

[0115] Among them, N is the total number of detail coefficients, is the maximum number of layers, represents the logarithmic function operation with base 2. In this way, it is ensured that enough data points (at least 8 samples) are retained after decomposition.

[0116] In an embodiment of this application, the initial termination threshold is set to 3% - 10% and is dynamically adjusted according to the signal type. For example, the threshold for high-frequency impact signals (such as mechanical excavation) is lower (3%), allowing more layers of decomposition; the threshold for low-frequency continuous vibration (such as vehicle rolling) is higher (8%), reducing excessive decomposition.

[0117] Then, in step S302, hierarchical threshold combination processing is performed on the high-frequency detail coefficients of different decomposition levels. Specifically, it includes: performing hard thresholding on the detail coefficients at decomposition level 1; performing semi-soft thresholding on the detail coefficients at decomposition levels 2 or 3; and performing soft thresholding on the detail coefficients at decomposition levels greater than or equal to 4.

[0118] Here, wavelet decomposition separates the signal layer by layer from high to low frequencies, and the high-frequency detail coefficients of different decomposition levels correspond to the noise and effective signals in different frequency bands. The hierarchical threshold combination processing realizes aggressive denoising in the shallow layer (high-frequency band) by dynamically adjusting the threshold rules and functions to eliminate random noise; balances noise suppression and signal retention in the middle layer (mid-high frequency band); and performs conservative processing in the deep layer (low-frequency band) to avoid loss of effective signals.

[0119] For the high-frequency detail coefficient Dj after wavelet decomposition, the mathematical definition of the hard threshold (Hard Thresholding) is:

[0120]

[0121] where Dj is the detail coefficient of the j-th layer of wavelet decomposition, λ is the preset threshold, usually based on the noise standard deviation (e.g., λ = 3σ), is the coefficient after threshold processing.

[0122] Among them, the hard threshold function has the characteristic of retaining mutation features. The coefficients with absolute values exceeding the threshold are completely retained, which is suitable for detecting transient signals (such as mechanical shocks). It has noise resistance and significant suppression effect on high-frequency noise.

[0123] Typical applications are in the high-frequency noise-dominated layer (such as the first layer of wavelet decomposition). The high-frequency band has high noise energy, and the hard threshold can quickly filter out most of the noise, which is suitable for detecting the initial shock signal of cable external break events in the φ-OTDR system. It can process strong transient signals, retain the amplitude integrity of pulse signals, and avoid the amplitude attenuation problem of soft thresholds.

[0124] The mathematical definition of the soft threshold (Soft Thresholding) is:

[0125]

[0126] where Dj is the detail coefficient of the j-th layer of wavelet decomposition, λ is the preset threshold, is the coefficient after threshold processing.

[0127] The soft threshold function can effectively suppress noise components, especially in wavelet denoising signal processing. By weakening or setting to zero the low-intensity signal coefficients, it can remove high-frequency noise and make the main features of the signal more prominent.

[0128] Compared with the hard thresholding method, the soft thresholding method shows smoother signal processing near the threshold. Since the coefficients are linearly shrunk (instead of simply set to zero), it helps to retain more signal details and features.

[0129] The threshold of the soft threshold can be adjusted according to different noise levels and signal characteristics. This makes the soft thresholding method more adaptable in various applications.

[0130] The mathematical definition of the semi-soft thresholding is:

[0131]

[0132] where \(D_j\) is the detail coefficient of the \(j\)-th layer of wavelet decomposition, is the coefficient after thresholding processing, and are the first preset threshold and the second preset threshold respectively, is the sign function.

[0133] The main feature of the semi-soft threshold function is that within a certain threshold, it still retains certain signal coefficients instead of completely setting them to zero. This enables it to retain a part of the low-intensity signal information while denoising, providing better smoothness and selectivity.

[0134] For those signal coefficients whose boundaries are near the threshold, the semi-soft threshold function can reduce the artifacts caused by noise, thereby improving the stability of the result. This method helps to avoid complete discard at low signal intensities and effectively suppresses the influence of noise.

[0135] The semi-soft threshold does not completely set to zero the coefficients less than the threshold, but reduces them, which enables it to more flexibly adapt to different types of signals, especially when there is more noise or uneven noise levels in the signal.

[0136] Preferably, set = 2 to balance smoothness and detail retention.

[0137] In the embodiments of the present application, in the shallow layer dominated by high-frequency noise (such as the first layer), an aggressive denoising threshold rule is adopted, and a hard threshold function is selected. The preset threshold is determined based on the estimation of the noise standard deviation \(\sigma\): \(\sigma=\text{median}(|D_1|) / 0.6745\), \(\lambda = 3\sigma\). That is, calculate the median of the absolute value of \(D_1\) divided by 0.6745 to obtain the standard deviation, and set the preset threshold to three times the standard deviation.

[0138] In the 2nd - 3rd layers with mixed noise and weak signals, select the semi - soft threshold function and determine the preset threshold by minimizing the Stein Unbiased Risk Estimate: . Determine the optimal threshold by minimizing the Stein Unbiased Risk Estimate.

[0139] Among them, SURE (Stein's Unbiased Risk Estimator) is a mathematical method for estimating the denoising error. Its core idea is that without knowing the true signal, just based on the observed noisy data, it can unbiasedly estimate the mean square error (MSE) after denoising.

[0140] That is, SURE provides an unbiased estimator for evaluating the denoising performance at different thresholds λ without knowing the true signal. By minimizing SURE, the optimal threshold λ can be found.

[0141] In the 4th and higher layers dominated by low - frequency effective signals, adopt a conservative processing (λ = 1.5σ) threshold rule, select the soft threshold function, and determine the preset threshold by dynamically adjusting in combination with the energy proportion: .

[0142] Among them, is the total energy of the signal, and β is the adjustment factor (taking values from 0.8 to 1.2).

[0143] Next, in step S303, the Dual - Tree Complex Wavelet Transforms (DTCWT) provides complex - form wavelet coefficients through two parallel real - wavelet trees (generating real and imaginary part coefficients respectively), retaining the phase information of the signal. It reduces the pseudo - Gibbs phenomenon of the traditional DWT, avoids signal edge distortion, and ensures translational invariance.

[0144] The real - part wavelet tree uses the wavelet basis ψ real for decomposition, and the imaginary - part wavelet tree uses the Hilbert transform pair ψ real to generate ψ imag , and finally realizes the dual - tree complex wavelet transform ψ complex (t)=ψ real (t)+jψ imag (t).

[0145] In the φ - OTDR system, the accuracy of the phase signal directly affects the positioning of the disturbance point. The refractive index of the optical fiber changes with temperature / stress, introducing non - linear phase shift; the temperature drift of electronic devices during long - term operation causes phase deviation; low - frequency phase fluctuations caused by wind, rain, and geological activities. The positioning error formula:

[0146]

[0147] wherein, is the phase error, f mod is the modulation frequency, n eff is the effective refractive index of the optical fiber, c represents the speed of light in vacuum, and its standard value in the International System of Units is approximately 2.998×10 8 m / s.

[0148] For example, when Δx is 1.5 m, it is necessary to reduce it through compensation .

[0149] Under the condition of no disturbance, inject a known phase signal and record the output phase of the system for calibration and calculate the phase deviation: . Use the low-frequency approximation coefficient (A j ) to fit the slow-varying phase drift, and extract the trend term through moving average or Kalman filtering .

[0150] Embed the compensation factor for reconstruction, and for the complex coefficient C j = D j real + jD j imag , apply the phase compensation factor .

[0151] wherein, . Use the corrected complex coefficient for inverse DTCWT to obtain the phase-corrected signal: s corrected (t) = IDTCWT({C j corrected}) and perform inverse transform reconstruction.

[0152] In long-distance monitoring, use the adaptive filtering algorithm (LMS) to update θ(t) in real time. The actual phase signal can be expressed as , where θ(t) is the time-varying phase deviation and n(t) is the noise.

[0153] The phase error θ(t) will cause a distance calculation deviation Δx ∝ θ(t). Especially in long-distance (>10 km) monitoring, a small phase drift (such as 0.1 rad) can cause a positioning error of meters. By iteratively updating the filter weights, minimize the mean square error between the output signal and the desired signal to estimate θ(t) in real time.

[0154] Establish a signal model. The reference signal d(t) is a known or estimated signal without phase deviation (such as a calibration signal or a low-frequency trend term), and the input signal x(t) is the actual measured signal with phase deviation . The output of the filter is , where w(t) is the adaptive weight corresponding to the phase compensation factor . The error is calculated as . The weight is updated as , where μ is the step size factor that controls the convergence speed and stability, is the conjugate of x(t) (to be considered when dealing with complex signals).

[0155] Perform low-pass filtering (cutoff frequency 0.1 Hz) on the approximation coefficient A j to obtain , completing the extraction of the low-frequency trend. Periodically inject a known calibration frequency wave, and at this time, d(t) is the ideal phase.

[0156] Initialize the weight w(0) = 1 (no compensation), step size μ = 0.01, and adjust dynamically according to the signal power: μ(t) = 0.1 / (1 + ||x(t)|| 2 ), ensuring 0 < μ < 1 / tr(R), where R is the autocorrelation matrix of the input signal. Update w(t) point by point according to the sampling points, and calculate the instantaneous phase deviation θ(t) = arg(w(t)). Apply the compensation factor j to the complex wavelet coefficient C . Output the reconstructed signal .

[0157] However, even if different wavelet bases can be used for wavelet analysis of different perturbed phase signals, this method has some drawbacks. For example, this method is limited by the preset finite wavelet basis types and the preset hierarchical threshold strategy, which may be difficult to handle unforeseen complex noise reduction scenarios.

[0158] In response, as Figure 3B shown, in step S310, this application expects to use a multi-granularity feature residual stripper to extract the shallow features (first-granularity perturbation feature map, second-granularity perturbation feature map) and deep semantic features (third-granularity perturbation feature map) from the preprocessed perturbed phase signal. This can simultaneously extract the local mutation information and global patterns in the signal (helping to macroscopically understand the event type).

[0159] Specifically, the multi-granularity feature residual stripper first performs convolution, batch normalization, and activation processing on the preprocessed perturbed phase signal to generate a first feature map, and then uses the first feature map as the input of the first residual module, and the first residual module outputs the first-granularity perturbation feature map.

[0160] At the same time, the first-granularity perturbation feature map is used as the input of the second residual module, and the second residual module outputs the second-granularity perturbation feature map. And the second-granularity perturbation feature map is used as the input of the third residual module, and the third residual module outputs the third-granularity perturbation feature map.

[0161] Here, convolution, batch normalization, and activation processing (such as ReLU, LeakyReLU) initially extract the local time-frequency features of the signal. Normalization stabilizes the training, and activation introduces non-linear expression ability. Among them, the first-granularity perturbation feature map extracts local details and describes the minute changes of the signal; the second-granularity perturbation feature map captures medium-granularity information and captures the perturbation features with a wider time domain; the third-granularity perturbation feature map depicts the overall trend of the signal.

[0162] It is worth mentioning that in traditional CNNs, the stacking of ordinary convolutions may cause the loss of shallow features (such as low-frequency information) in the deep layers, while the residual structure can retain the corresponding feature information after each processing. These feature information can be used as the input of the subsequent model and provide an important data source for the enhancement of shallow key features.

[0163] Then, in step S320, the key features of the shallow features are enhanced based on the spatial attention mechanism to automatically suppress non-key regions (such as the frequency bands where environmental noise is concentrated) through attention weights.

[0164] Here, the spatial attention mechanism highlights the key regions related to the external break event and suppresses irrelevant background noise by dynamically calculating the weights of each spatial position in the feature map. That is, it can automatically identify high-probability perturbation regions through attention weights (such as the 0~1 values activated by Sigmoid).

[0165] It is worth mentioning that the perturbation features in the shallow layer (such as the transient time-varying characteristics in the impact scenario) are more important than the low-frequency and stable background noise. The spatial attention mechanism simulates the focusing ability of the human eye to pay more attention to important perturbation features from the background noise, making the subsequent model's expression of the comprehensive perturbation features more delicate.

[0166] Then, in step S330, a bidirectional dynamic fuser is used to adaptively adjust the fusion weights between multi-granularity perturbation features, thereby alleviating the fusion conflict caused by the direct mixing of multi-layer perturbation features and obtaining a comprehensive perturbation feature representation.

[0167] Here, in the existing technology, the FPN (Feature Pyramid Networks) network retains the detailed information in the shallow features and the semantic information in the deep features by fusing shallow features and deep features. However, the way FPN fuses data is relatively simple, that is, directly adding and fusing deep and shallow features. This fusion method may cause problems such as information conflict and noise introduction.

[0168] Specifically, deep features usually contain high-level semantic information (such as object categories and overall structures), while shallow features retain more low-level details (such as edges and textures). There are significant differences in the semantic levels between the two, and directly adding them will result in a mismatch in the feature space, potentially introducing noise rather than effective information.

[0169] In addition, shallow features have a small receptive field and focus on local details; deep features have a large receptive field and focus on global context. Direct addition may cause the fused features to lose the advantage of multi-scale information due to scale differences, especially the ability to capture details decreases.

[0170] Meanwhile, if there are contradictions between shallow features and deep semantic features, for example, the shallow features detect an edge while the deep features ignore that area, adding them may blur important features.

[0171] In the technical solution of this application, it is expected to first perform attention-based feature extraction on shallow features using a bidirectional dynamic fuser to retain and strengthen the weak features contained therein; and to adjust and guide the deep features to balance the fusion between shallow features and deep features.

[0172] More specifically, input the second perturbation-enhanced feature map and the third granularity perturbation feature map into the first bidirectional dynamic fusion layer to obtain a perturbation fusion feature map; then input the first perturbation-enhanced feature map and the perturbation fusion feature map into the second bidirectional dynamic fusion layer to obtain a multi-granularity perturbation feature map.

[0173] Among them, the first bidirectional dynamic fusion layer performs point-wise convolution, batch normalization, and activation processing on the second perturbation-enhanced feature map, and performs element-wise multiplication of the processed feature map with the third granularity perturbation feature map to obtain a first fusion feature map; the formula is expressed as follows:

[0174]

[0175] Among them, F2 is the second perturbation-enhanced feature map, F3 is the third granularity perturbation feature map, LR is the Leaky Relu function, sigmoid is the sigmoid function, BN is batch normalization, PConv is point-wise convolution, element-wise multiplication, F r1 is the first fusion feature map;

[0176] Meanwhile, perform global average pooling, fully convolutional, batch normalization, and activation processing on the third granularity perturbation feature map, and perform element-wise multiplication of the processed feature map with the second perturbation-enhanced feature map to obtain a second fusion feature map; the formula is expressed as follows:

[0177]

[0178] Among them, GAP is global average pooling, FC is the fully connected layer, and F r2 is the second fused feature map;

[0179] Finally, the first fused feature map and the second fused feature map are added together according to weights to obtain the perturbed fused feature map.

[0180] Similarly, the second bidirectional dynamic fusion layer fuses the first perturbed enhanced feature map and the perturbed fused feature map through similar processing. In this way, by retaining and enhancing shallow features and adjusting and guiding deep semantic features through attention, and setting weight parameters to balance the information fusion ratio between features, multi-granularity feature representations can be better embedded in each other.

[0181] Finally, in step S340, a decoder is used to construct a noise reducer, which gradually upsamples through a transposed convolutional layer to automatically restore and reconstruct the comprehensive perturbed feature representation. In this way, a restored perturbed noise reduction signal is intelligently generated, converting from a strong manual-dependence type to a data-driven adaptive noise reduction method to better assist professional technicians in cable signal analysis work. This conversion of the expert experience in signal processing into learnable network parameters is especially suitable for cable external break monitoring in complex scenarios.

[0182] It is worth mentioning that the reason for not directly identifying the external break event but denoising the signal is that the discrimination of the model is not interpretable, and denoising the signal can visually display the denoised signal, which is convenient for technicians to use.

[0183] As Figure 4 shown, the present application also provides a cable external break prevention signal noise reduction system, which includes:

[0184] A perturbed signal acquisition module 110, configured to acquire an original perturbed phase signal collected by a distributed optical fiber vibration sensing module based on φ-OTDR;

[0185] A preprocessing module 120, configured to preprocess the original perturbed phase signal to obtain a preprocessed perturbed phase signal;

[0186] A noise reduction module 130, configured to dynamically extract and reduce noise from the preprocessed perturbed phase signal to obtain a restored perturbed noise reduction signal.

[0187] As Figure 5 shown, the present application also provides a terminal, which includes:

[0188] A processor 10, configured to be coupled with a memory 20, and read and execute instructions 30 stored in the memory;

[0189] When the processor 10 runs, it executes the instruction 30, so that the processor 10 is used to execute the cable anti-external break signal noise reduction method.

[0190] Figure 5 Only some components of the terminal are shown. However, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively.

[0191] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as the hard disk or memory of the terminal.

[0192] In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk equipped on the terminal, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc.

[0193] Furthermore, the memory 20 may also include both the internal storage unit of the terminal and the external storage device.

[0194] The memory 20 is used to store the application software installed on the terminal and various types of data, such as the program code installed on the terminal, etc.

[0195] The memory 20 may also be used to temporarily store the data that has been output or will be output. In one embodiment, the instruction 30 is stored on the memory 20, and this instruction 30 can be executed by the processor 10 to execute the cable anti-external break signal noise reduction method.

[0196] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor or other data processing chips, used to run the instruction 30 stored in the memory 20 or process data, such as executing the cable anti-external break signal noise reduction method, etc.

[0197] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited by the above embodiments. Without departing from the spirit and scope of the present application, the present application will have various changes and improvements, and these changes and improvements all fall within the scope claimed by the present application. The scope claimed by the present application is defined by the appended claims and their equivalents.

Claims

1. A method for reducing noise of signals against external damage of cables, characterized in that, Including: Obtaining an original disturbance phase signal collected by a distributed optical fiber vibration sensing module based on φ-OTDR; Preprocessing the original disturbance phase signal to obtain a preprocessed disturbance phase signal; Performing dynamic extraction and noise reduction of disturbance information on the preprocessed disturbance phase signal to obtain a restored disturbance noise reduction signal; Among them, performing dynamic extraction and noise reduction of disturbance information on the preprocessed disturbance phase signal to obtain a restored disturbance noise reduction signal includes: Dynamically selecting a wavelet basis and the number of decomposition levels, and performing wavelet decomposition on the preprocessed disturbance phase signal to obtain a plurality of detail coefficients; Performing hierarchical threshold combination processing on the plurality of detail coefficients; Based on the plurality of detail coefficients after hierarchical threshold combination processing, using a dual-tree complex wavelet to reconstruct a noise reduction signal and embedding a phase compensation factor to obtain the restored disturbance noise reduction signal.

2. The cable anti-external break signal noise reduction method according to claim 1, wherein, Preprocessing the original disturbance phase signal to obtain a preprocessed disturbance phase signal includes: Performing preprocessing on the original disturbance phase signal by using moving average filtering and / or phase difference to obtain the preprocessed disturbance phase signal; Among them, using the following difference formula to perform phase difference processing on the input signal to obtain an output signal; Among them, the difference formula is: where t ranges from 1 to the length of the input signal, is the value of the output signal at time t, is the input signal at time is the value of the input signal at time t - 1, is the value of the output signal at the initial time, is the value of the input signal at the initial time.

3. The cable anti-external damage signal noise reduction method according to claim 1, characterized in that, Dynamically selecting a wavelet basis and the number of decomposition levels, and performing wavelet decomposition on the preprocessed disturbance phase signal to obtain a plurality of detail coefficients, including: Determining the signal category to which the preprocessed disturbance phase signal belongs based on the signal difference degree between the preprocessed disturbance phase signal and each group of sample disturbance phase signals in the sample library; Selecting a wavelet basis based on the signal category and performing wavelet decomposition on the preprocessed disturbance phase signal to obtain the plurality of detail coefficients; Among them, during the wavelet decomposition process, an adaptive strategy for the number of decomposition levels is used to control whether the wavelet decomposition terminates.

4. The cable anti-external damage signal noise reduction method according to claim 3, wherein, The signal category includes at least one of the following: instantaneous impact, continuous vibration, and mixed noise; Among them, when the signal category to which the preprocessed disturbance phase signal belongs is an instantaneous impact, select db4 wavelet and coif3 wavelet as the wavelet basis; When the signal category to which the preprocessed disturbance phase signal belongs is continuous vibration, select sym4 wavelet as the wavelet basis; When the signal category to which the preprocessed disturbance phase signal belongs is mixed noise, select bior3.3 wavelet as the wavelet basis.

5. The cable anti-external-break signal noise reduction method according to claim 4, wherein, Using an adaptive strategy for the number of decomposition levels to control whether the wavelet decomposition terminates during the wavelet decomposition process includes: If the current number of decomposition levels is less than 2, continue the decomposition; If the current number of decomposition levels is greater than or equal to 3, then perform the following steps: Determining a termination threshold according to the signal category to which the preprocessed disturbance phase signal belongs; Calculating the energy change rate of the current decomposition layer; When the energy change rate is less than the termination threshold, the decomposition terminates, otherwise continue the decomposition; Among them, the energy change rate is calculated by the following energy change rate quantization formula; the energy change rate quantization formula is: Among them, is the k-th detail coefficient of the j-th layer, is the number of detail coefficients of the j-th layer, |·| represents calculating the absolute value, is the energy of the j-th layer, is the energy of the (j - 1)-th layer, is the energy change rate of the j-th layer relative to the (j - 1)-th layer; Among them, the value of j is restricted by the maximum layer, which is expressed by the formula: where N is the total number of detail coefficients, is the maximum number of layers, represents the logarithmic function operation with base 2.

6. The cable anti-external damage signal noise reduction method according to claim 5, wherein Performing hierarchical threshold combination processing on the plurality of detail coefficients includes: Performing hard threshold processing on the detail coefficients at the decomposition level of 1; Perform semi-soft threshold processing on the detail coefficients at decomposition levels 2 or 3; Perform soft threshold processing on the detail coefficients at decomposition levels greater than or equal to 4.

7. The cable anti-external damage signal noise reduction method according to claim 1, characterized in that Perform dynamic extraction and noise reduction of disturbance information on the preprocessed disturbed phase signal to obtain a restored disturbed and noise-reduced signal, including: Use a multi-granularity feature residual stripper to process the preprocessed disturbed phase signal to obtain a first-granularity disturbance feature map, a second-granularity disturbance feature map, and a third-granularity disturbance feature map; Input the first-granularity disturbance feature map and the second-granularity disturbance feature map into a convolutional neural network using a spatial attention mechanism to obtain a first disturbance-enhanced feature map and a second disturbance-enhanced feature map; Use a bidirectional dynamic fusion device to process the first disturbance-enhanced feature map, the second disturbance-enhanced feature map, and the third-granularity disturbance feature map to obtain a multi-granularity disturbance feature map; Use a decoder-based noise reducer to process the multi-granularity disturbance feature map to obtain the restored disturbed and noise-reduced signal.

8. A cable anti-external damage signal noise reduction system, characterized in that, Include: A disturbance signal acquisition module for acquiring an original disturbed phase signal collected by a distributed optical fiber vibration sensing module based on φ-OTDR; A preprocessing module for preprocessing the original disturbed phase signal to obtain a preprocessed disturbed phase signal; A noise reduction module for performing dynamic extraction and noise reduction of disturbance information on the preprocessed disturbed phase signal to obtain a restored disturbed and noise-reduced signal.

9. A terminal, characterized in that, Include: A processor for coupling with a memory and reading and executing instructions stored in the memory; When the processor runs, it executes instructions, enabling the processor to execute a cable anti-external break signal noise reduction method.

Images