Image Processing-Based Phase-Sensitive Optical Time-Domain Reflectometer and Signal Demodulation Method and System

Abstract

This invention relates to an image processing-based phase-sensitive optical time-domain reflectometer and signal demodulation method and system, comprising: a laser, an electro-optic modulator, an arbitrary function generator, a first circulator, a filter, a first erbium-doped fiber amplifier, a bandpass filter, a semiconductor optical amplifier, a second erbium-doped fiber amplifier, a second circulator, an optical fiber under test, a photodetector, and a data acquisition card. The method of this invention facilitates high-precision demodulation over a wide range of strain or temperature conditions with high spatial resolution. This method combines image block matching and image denoising. Image block matching utilizes multiple frequency and spatial location information, improving demodulation accuracy when strain or temperature changes are large and helping to expand the measurement range. The image denoising method denoises the matched function image, reducing random noise in the system and improving demodulation accuracy at high spatial resolution.

Description

Technical Field

[0001] This invention relates to the field of fiber optic sensing technology, specifically to a phase-sensitive optical time-domain reflectometer based on image processing and a signal demodulation method and system. Background Technology

[0002] The development of distributed fiber optic sensing has made significant contributions to many applications in fields such as structural health monitoring, safety monitoring, and the energy industry. Among various distributed fiber optic sensors, phase-sensitive optical time-domain reflectometers are renowned for their ultra-high sensitivity and measurement resolution. Benefiting from their extreme sensitivity to changes in external quantities, phase-sensitive optical time-domain reflectometers have been applied to high-sensitivity temperature and strain measurements.

[0003] In the demodulation process of a phase-sensitive optical time-domain reflectometer (OTDR), the shift in optical frequency is typically determined by calculating the cross-correlation of the backscattered Rayleigh spectra before and after strain or temperature changes, thus obtaining the strain or temperature change. However, when the strain or temperature change is large, the similarity between the two spectra decreases, and spurious peaks appear in the cross-correlation spectrum, leading to erroneous demodulation results. This phenomenon reduces measurement accuracy and increases cumulative errors over a wide range of measurements, limiting the measurable strain and temperature range. Furthermore, as the spatial resolution of the system increases, the influence of random noise in the system becomes more significant, further reducing measurement accuracy. Therefore, a high-precision signal demodulation method for phase-sensitive OTDs is urgently needed. Summary of the Invention

[0004] This invention addresses the problems of existing technologies by proposing an image-processing-based phase-sensitive optical time-domain reflectometer and signal demodulation method. This invention is achieved through the following technical solutions:

[0005] The present invention discloses an image processing-based phase-sensitive optical time-domain reflectometer, comprising: a laser (1-1), an electro-optic modulator (1-2), an arbitrary function generator (1-3), a first circulator (1-4), a filter (1-5), a first erbium-doped fiber amplifier (1-6), a bandpass filter (1-7), a semiconductor optical amplifier (1-8), a second erbium-doped fiber amplifier (1-9), a second circulator (1-10), an optical fiber under test (1-11), a photodetector (1-12), and a data acquisition card (1-13).

[0006] Further, the laser (1-1) is connected to the electro-optic modulator (1-2), and the arbitrary function generator (1-3) is connected to the electro-optic modulator (1-2). The light output from the laser (1-1) is modulated by the electro-optic modulator (1-2) into a double-sideband modulated signal with suppressed carrier. This double-sideband modulated signal with suppressed carrier enters the filter (1-5) through the first circulator (1-4) to filter out one of the sidebands, resulting in a single-sideband modulated signal. This single-sideband modulated signal enters the first erbium-doped fiber amplifier (...). 1-6) After amplification, the signal is filtered out by the bandpass filter (1-7) to remove spontaneous emission noise; the semiconductor optical amplifier (1-8) is used to modulate the filtered single-sideband modulation signal into a pulse light sequence with different frequencies; the second erbium-doped fiber amplifier (1-9) amplifies the pulse light sequence, and the pulse light sequence is injected into the fiber under test (1-11) through the second circulator (1-10); the backscattered Rayleigh signal in the fiber under test is converted into an electrical signal by the photodetector (1-12) and then acquired by the data acquisition card (1-13).

[0007] Furthermore, the filter (1-5) is a reflective fiber Bragg grating; the semiconductor optical amplifier (1-8) has a high extinction ratio of 50dB; the first erbium-doped fiber amplifier (1-6) is a continuous optical amplifier, and the second erbium-doped fiber amplifier (1-9) is a pulsed optical amplifier.

[0008] The present invention also relates to a signal demodulation method using the above-mentioned image processing-based phase-sensitive optical time-domain reflectometer, the signal demodulation method comprising the following steps:

[0009] Step S1: Collect the first characteristic image of the backscattering Rayleigh spectrum before strain and the second characteristic image of the backscattering Rayleigh spectrum after strain, respectively;

[0010] Step S2: Based on the influence of image patch size on the matching degree of image patches in the first feature map and the second feature map, calculate the change of matching degree of image patch in the first feature map and matching image patch in the second feature map with the image patch size under different strain changes, and select the image patch size a with the maximum matching degree to perform the image patch matching process.

[0011] Step S3: The difference between the second-dimensional coordinates of the image block of the first feature map and the second-dimensional coordinates of the matching image block of the second feature map is the frequency offset. The spatial domain and the frequency offset domain are used as the first and second-dimensional coordinates, respectively, and the matching function value is used as the third-dimensional coordinate. The matching function curves of all spatial locations are used to construct a three-dimensional matching function graph.

[0012] Step S4: Project the three-dimensional matching function graph into a two-dimensional image in the spatial domain-frequency offset domain, and use the BM3D image denoising method to denoise the image.

[0013] Step S5: Extract the second-dimensional coordinates corresponding to the maximum value of the matching function at each spatial location in the denoised image, and determine the frequency offset at each location of the fiber under test based on the coordinate value; solve for the temperature based on the relationship between the frequency offset and the temperature change, and solve for the strain based on the relationship between the frequency offset and the strain change.

[0014] Furthermore, step S1 specifically includes:

[0015] The data acquisition card acquires the backscattering Rayleigh spectrum of a pulse sequence at N frequencies at the first moment. Using the spatial domain and frequency domain as the first and second dimensions respectively, and the signal intensity as the third dimension, a three-dimensional backscattering spectrum is obtained and projected into the spatial-frequency domain as a first feature map. At the second moment, the data acquisition card acquires the backscattering spectrum of a pulse sequence at N frequencies. Using the spatial domain and frequency domain as the first and second dimensions respectively, and the signal intensity as the third dimension, a three-dimensional backscattering spectrum is obtained and projected into the spatial-frequency domain as a second feature map.

[0016] Furthermore, in step S2, the matching function that measures the matching degree is:

[0017]

[0018] Where Δf is the frequency offset, f is the laser frequency, A is the image block in the first feature map, B is the image block in the second feature map, x is the spatial position of the image block, M is the number of positions in the fiber under test, and N is the number of modulation signals generated by the arbitrary function generator.

[0019] Further, in step S3, image block matching is performed on the two feature maps. In the first feature map, an image block of size a×a is selected at the spatial domain coordinate x, and the coordinates of the lower left corner of the image block are used as the coordinates of the image block. In the second feature map, an image block with the initial position and initial frequency as coordinates is selected, and the matching function between this image block and the image block at the spatial domain coordinate x in the first feature map is calculated. The image block in the second feature map is moved along the frequency domain, and the matching function is calculated for each step. After the image block in the second feature map moves to the maximum value in the frequency domain, it returns to the initial frequency and moves along the spatial domain to the next position. The image block in the second feature map is moved along the frequency domain again, and the matching function is calculated for each step. This process is repeated to traverse the entire second feature map, calculate the matching function between the image block of size a×a at each spatial position in the second feature map and the image block at the spatial domain coordinate x in the first feature map, and obtain the position coordinates of the position with the maximum matching function. The difference in the second dimension of the coordinate is the frequency offset under the spatial coordinate, and the spatial domain curve where the coordinate is located is the matching function curve at the spatial domain coordinate x.

[0020] The image block in the first feature map is moved to the next position along the spatial domain, and the matching function between each a×a image block in the second feature map and the image block at the given position is calculated. In this way, the matching function between the image block at each spatial position in the first feature map and the image block in the second feature map is obtained. The difference between the second-dimensional coordinates of the image block in the first feature map and the second-dimensional coordinates of the matching image block in the second feature map is the frequency offset. The spatial domain and the frequency offset domain are used as the first and second-dimensional coordinates, respectively, and the matching function value is used as the third-dimensional coordinate. The matching function curves of all spatial positions are used to construct a three-dimensional matching function graph.

[0021] Furthermore, in step S4, the BM3D process is divided into two stages: basic estimation and final estimation.

[0022] Basic estimation stage: The matching function graph is divided into a series of k×k reference blocks, and all points of the image are traversed by raster scanning with a fixed step size. For the current reference block, a similar block search is performed in its surrounding neighborhood (n×n, n>k). The reference block is compared with all blocks in the window. All similar blocks are stacked to form a three-dimensional matrix according to Euclidean distance. Hard threshold collaborative filtering is performed on the three-dimensional matrix. The two-dimensional data blocks in each three-dimensional matrix are transformed using 2D-Bior1.5 wavelet, and a one-dimensional transformation is performed using 1D-Sym8 wavelet in the third dimension of the matrix. The three-dimensional transformation is achieved by combining the two-dimensional transformations. The processed image blocks are obtained through the three-dimensional inverse transformation, and the block estimates are returned to their original positions. The basic estimate of the image is calculated by weighted averaging of all obtained overlapping estimated blocks.

[0023] Final estimation stage: After block matching of the matching function graph and the basic estimate, they are stacked to form a three-dimensional matrix. Both three-dimensional matrices are subjected to three-dimensional transformations. The two-dimensional transformation uses discrete cosine transform, and the one-dimensional transformation of the third dimension uses 1D-Sym8 wavelet transform. Wiener filtering is used to scale the coefficients of the three-dimensional matrix of the matching function graph. These coefficients are obtained from the values ​​of the three-dimensional matrix of the basic estimate and the noise intensity. The estimated values ​​of all image blocks are generated by the three-dimensional inverse transformation, and the estimated values ​​of the blocks are returned to their original positions. The final estimate of the image is calculated by weighted averaging of all the obtained overlapping estimated blocks.

[0024] Furthermore, in step S5, the relationship between the frequency offset and the temperature change is as follows: Where Δf is the frequency offset, f is the laser frequency, ζ is the thermo-optic coefficient, α is the coefficient of thermal expansion, and ΔT is the temperature change; the relationship between the frequency offset and the strain change is as follows: Where Δf is the frequency offset, f is the laser frequency, and p ε ε is the elastic coefficient, and Δε is the change in strain.

[0025] The present invention also relates to a signal demodulation system for a phase-sensitive optical time-domain reflectometer based on image processing, comprising a computer module that runs the aforementioned signal demodulation method for a phase-sensitive optical time-domain reflectometer based on image processing.

[0026] Beneficial effects

[0027] The image processing-based phase-sensitive optical time-domain reflectometer and signal demodulation method of this invention facilitate high-precision demodulation over a wide range of strain or temperature conditions with high spatial resolution. This method combines image block matching and image denoising. Image block matching utilizes multiple frequency and spatial location information, improving demodulation accuracy when strain or temperature changes are large and helping to expand the measurement range. The image denoising method denoises the matching function image, reducing random noise in the system and improving demodulation accuracy at high spatial resolution. The matching function in the image processing-based phase-sensitive optical time-domain reflectometer signal demodulation method of this invention differs from the cross-correlation function, avoiding errors caused by cross-correlation calculations themselves and improving measurement accuracy. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the phase-sensitive optical time-domain reflectometer system provided by the present invention.

[0029] Figure 2 The flowchart shows the signal demodulation method of the phase-sensitive optical time-domain reflectometer based on image processing provided by the present invention.

[0030] Figure 3 This is a diagram showing the relationship between image patch size and matching degree when the strain takes different values, as described in a specific embodiment of the present invention.

[0031] Figure 4 This is a diagram of the three-dimensional matching function in a specific embodiment of the present invention.

[0032] Explanation of icon numbers

[0033] Laser 1-1, Electro-optic modulator 1-2, Arbitrary function generator 1-3, First circulator 1-4, Filter 1-5, First erbium-doped fiber amplifier 1-6, Bandpass filter 1-7, Semiconductor optical amplifier 1-8, Second erbium-doped fiber amplifier 1-9, Second circulator 1-10, Fiber under test 1-11, Photodetector 1-12, Data acquisition card 1-13. Detailed Implementation

[0034] The following is in conjunction with the appendix Figures 1 to 4 The present invention further details the phase-sensitive optical time-domain reflectometer and signal demodulation method based on image processing.

[0035] Please see Figure 1The image processing-based phase-sensitive optical time-domain reflectometer of the present invention includes, structurally, a laser 1-1, an electro-optic modulator 1-2, an arbitrary function generator 1-3, a first circulator 1-4, a filter 1-5, a first erbium-doped fiber amplifier 1-6, a bandpass filter 1-7, a semiconductor optical amplifier 1-8, an erbium-doped fiber amplifier 1-9, a second circulator 1-10, a fiber under test 1-11, a photodetector 1-12, and a data acquisition card 1-13.

[0036] The connection relationships and principles of each component are as follows:

[0037] Laser 1-1 is connected to electro-optic modulator 1-2, and arbitrary function generator 1-3 is connected to electro-optic modulator 1-2. The light output from laser 1-1 is modulated by electro-optic modulator 1-2 into a suppressed-carrier double-sideband modulated signal. This suppressed-carrier double-sideband modulated signal passes through the first circulator 1-4 and enters the filter 1-5 to filter out one sideband, resulting in a single-sideband modulated signal. This single-sideband modulated signal is amplified by the first erbium-doped fiber amplifier 1-6 and then passes through the bandpass amplifier. Filters 1-7 filter out spontaneous emission noise; the semiconductor optical amplifier 1-8 modulates the filtered single-sideband modulation signal into a pulse light sequence with N different frequencies; the second erbium-doped fiber amplifier 1-9 amplifies the pulse light sequence, which is then injected into the fiber under test 1-11 via the second circulator 1-10; the fiber under test 1-11 contains M location points, and the backscattered Rayleigh signal in the fiber under test is converted into an electrical signal by a photodetector 1-12 and then acquired by a data acquisition card 1-13.

[0038] Filters 1-5 are implemented using reflective fiber Bragg gratings; semiconductor optical amplifiers 1-8 have a high extinction ratio.

[0039] Preferably, the first erbium-doped fiber amplifier 1-6 is a continuous optical amplifier, and the second erbium-doped fiber amplifier 1-9 is a pulsed optical amplifier.

[0040] Example

[0041] In this embodiment, the laser 1-1 in the image processing-based phase-sensitive optical time-domain reflectometer has an output light wavelength of 1550nm and a linewidth of 1kHz; the electro-optic modulator 1-2 has a bandwidth of 20GHz; the arbitrary function generator 1-3 generates a modulation signal with a frequency range of 9GHz-10GHz and a frequency change interval of 20MHz; and the semiconductor optical amplifier 1-8 modulates the continuous light into a pulse sequence with a pulse width of 10ns.

[0042] It should be noted that the image processing-based phase-sensitive optical time-domain reflectometer of the present invention is not limited to the above-mentioned parameter design, and any parameter within the feasible range is acceptable.

[0043] Please see Figure 2 The present invention also relates to a signal demodulation method for a high-precision phase-sensitive optical time-domain reflectometer based on image processing, the method comprising the following steps:

[0044] S1. Collect the first characteristic map of the back Rayleigh scattering spectrum before strain and the second characteristic map of the back Rayleigh scattering spectrum after strain, respectively.

[0045] S2. Analyze the influence of image patch size on the matching degree of image patches in the first feature map and the second feature map. Calculate the change of matching degree between the image patch in the first feature map and the matching image patch in the second feature map with the image patch size under different strain changes. Select the image patch size a with the maximum matching degree to perform the image patch matching process.

[0046] The matching function that measures the matching degree is:

[0047]

[0048] S3. Perform image patch matching on the two feature maps. In the first feature map, select an image patch of size a×a at spatial coordinate x, and use the coordinates of the lower left corner of the image patch as its coordinates. In the second feature map, select an image patch with the initial position and initial frequency as coordinates, and calculate the matching function between this image patch and the image patch at spatial coordinate x in the first feature map. Move the image patch in the second feature map along the frequency domain, calculating the matching function for each move. After the image patch in the second feature map moves to the maximum value in the frequency domain, it returns to the initial frequency and moves the image patch along the spatial domain to the next position. Repeat this process, moving the image patch in the second feature map along the frequency domain again, calculating the matching function for each move. This process is repeated for the entire second feature map, calculating the matching function between the image patch of size a×a at each spatial position in the second feature map and the image patch at spatial coordinate x in the first feature map. The matching function of the image patch is obtained, and the position coordinates of the maximum matching function are obtained. The difference in the second dimension of the coordinate is the frequency offset in the spatial coordinate. The spatial domain curve where the coordinate is located is the matching function curve at the spatial domain coordinate x. The image patch in the first feature map is moved to the next position along the spatial domain, and the matching function of each a×a image patch in the second feature map with the image patch at the position is calculated. The matching function of each spatial position of the image patch in the first feature map with the image patch in the second feature map is obtained in this way. The difference in the second dimension coordinates of the image patch in the first feature map and the matching image patch in the second feature map is the frequency offset. The spatial domain and the frequency offset domain are the first and second dimension coordinates, respectively. The matching function value is the third dimension coordinate. The matching function curves of all spatial positions are constructed into a three-dimensional matching function map.

[0049] S4. Denoise the matching image and project the three-dimensional matching function graph into a two-dimensional image in the spatial domain-frequency offset domain; use BM3D to denoise the image, which is divided into two stages: basic estimation and final estimation.

[0050] In the basic estimation stage, the matching function graph is divided into a series of k×k reference blocks, and all points in the image are traversed through a raster scan sequence with a fixed step size. For the current reference block, a similar block search is performed in its surrounding neighborhood (n×n, n>k). The reference block is compared with all blocks in the window. All similar blocks are stacked to form a three-dimensional matrix based on Euclidean distance. Hard thresholding collaborative filtering is applied to the three-dimensional matrix. The two-dimensional data blocks in each three-dimensional matrix are transformed using 2D-Bior1.5 wavelet, and a one-dimensional transformation is performed using 1D-Sym8 wavelet in the third dimension of the matrix. The three-dimensional transformation is achieved by combining the two-dimensional transformations. The processed image blocks are obtained through the inverse three-dimensional transformation, and the block estimates are returned to their original positions. The basic estimate of the image is calculated by weighted averaging of all obtained overlapping estimated blocks.

[0051] In the final estimation stage, the matching function graph and the basic estimate are stacked after block matching to form a three-dimensional matrix. Both three-dimensional matrices are subjected to three-dimensional transformations. The two-dimensional transformation uses discrete cosine transform, and the one-dimensional transformation of the third dimension uses 1D-Sym8 wavelet transform. Wiener filtering is used to scale the coefficients of the three-dimensional matrix of the matching function graph. These coefficients are obtained from the values ​​of the three-dimensional matrix of the basic estimate and the noise intensity. The estimated values ​​of all image blocks are generated by the three-dimensional inverse transformation, and the estimated values ​​of the blocks are returned to their original positions. The final estimate of the image is calculated by weighted averaging of all the obtained overlapping estimated blocks.

[0052] S5. Extract the second-dimensional coordinates corresponding to the maximum values ​​of the matching functions at each spatial location in the denoised image, and determine the frequency offset at each location of the fiber under test based on these coordinate values. Calculate the strain change, where Δf is the frequency shift, f is the laser frequency, and p ε ε is the elastic coefficient, and Δε is the change in strain.

[0053] Figure 3 The figure shows the relationship between image patch size and matching degree when the strain changes are 0.5με, 1με, 2με and 3με.

[0054] Figure 4 The figure shows the three-dimensional matching function plots for each spatial location when the strain change at the 4-10m position is 3με and there is no change at other positions.

[0055] The above description is only one embodiment of the present invention, and not all or the only embodiment. Any equivalent modifications made by those skilled in the art to the technical solution of the present invention by reading the present invention specification are covered by the claims of the present invention.

Claims

1. A phase-sensitive optical time-domain reflectometer based on image processing, characterized in that, include: Laser (1-1), electro-optic modulator (1-2), arbitrary function generator (1-3), first circulator (1-4), filter (1-5), first erbium-doped fiber amplifier (1-6), bandpass filter (1-7), semiconductor optical amplifier (1-8), second erbium-doped fiber amplifier (1-9), second circulator (1-10), fiber under test (1-11), photodetector (1-12), and data acquisition card (1-13); The laser (1-1) is connected to the electro-optic modulator (1-2), and the arbitrary function generator (1-3) is connected to the electro-optic modulator (1-2). The light output from the laser (1-1) is modulated by the electro-optic modulator (1-2) into a suppressed-carrier double-sideband modulated signal. This suppressed-carrier double-sideband modulated signal passes through the first circulator (1-4) and enters the filter (1-5) to filter out one of the sidebands, resulting in a single-sideband modulated signal. This single-sideband modulated signal enters the first erbium-doped fiber amplifier (1- 6) After amplification, the spontaneous emission noise is filtered out by the bandpass filter (1-7); the semiconductor optical amplifier (1-8) is used to modulate the filtered single-sideband modulation signal into a pulse light sequence with different frequencies; the second erbium-doped fiber amplifier (1-9) amplifies the pulse light sequence, and the pulse light sequence is injected into the fiber under test (1-11) through the second circulator (1-10); the backscattered Rayleigh signal in the fiber under test is converted into an electrical signal by the photodetector (1-12) and then acquired by the data acquisition card (1-13).

2. The image-processing-based phase-sensitive optical time-domain reflectometer according to claim 1, characterized in that, The filter (1-5) is a reflective fiber Bragg grating; the semiconductor optical amplifier (1-8) has a high extinction ratio of 50dB; the first erbium-doped fiber amplifier (1-6) is a continuous optical amplifier, and the second erbium-doped fiber amplifier (1-9) is a pulsed optical amplifier.

3. A signal demodulation method using an image processing-based phase-sensitive optical time-domain reflectometer as described in any one of claims 1 to 2, characterized in that, The signal demodulation method includes the following steps: Step S1: Collect the first characteristic image of the backscattering Rayleigh spectrum before strain and the second characteristic image of the backscattering Rayleigh spectrum after strain, respectively; Step S2: Based on the influence of image patch size on the matching degree of image patches in the first feature map and the second feature map, calculate the change of matching degree of image patch in the first feature map and matching image patch in the second feature map with the image patch size under different strain changes, and select the image patch size a with the maximum matching degree to perform the image patch matching process. Step S3: The difference between the second-dimensional coordinates of the image block of the first feature map and the second-dimensional coordinates of the matching image block of the second feature map is the frequency offset. The spatial domain and the frequency offset domain are used as the first and second-dimensional coordinates, respectively, and the matching function value is used as the third-dimensional coordinate. The matching function curves of all spatial locations are used to construct a three-dimensional matching function graph. Step S4: Project the three-dimensional matching function graph into a two-dimensional image in the spatial domain-frequency offset domain, and use the BM3D image denoising method to denoise the image. Step S5: Extract the second-dimensional coordinates corresponding to the maximum value of the matching function at each spatial location in the denoised image, and determine the frequency offset at each location of the fiber under test based on the coordinate value; solve for the temperature based on the relationship between the frequency offset and the temperature change, and solve for the strain based on the relationship between the frequency offset and the strain change.

4. The signal demodulation method for a phase-sensitive optical time-domain reflectometer based on image processing according to claim 3, characterized in that, Step S1 is as follows: The data acquisition card acquires the backscattering Rayleigh spectrum of pulse sequences at N frequencies at the first moment. The spatial domain and frequency domain are used as the first and second dimensions, respectively, and the signal intensity is used as the third dimension. The three-dimensional backscattering spectrum is obtained and projected onto the spatial-frequency domain as the first feature map. Backward Rayleigh scattering spectra of pulse sequences at N frequencies are acquired at the second time point. The spatial domain and frequency domain are used as the first and second dimensions, respectively, and the signal intensity is used as the third dimension. The three-dimensional backward Rayleigh scattering spectrum is obtained and projected onto the spatial-frequency domain as the second feature map.

5. The signal demodulation method for a phase-sensitive optical time-domain reflectometer based on image processing according to claim 3, characterized in that, In step S2, the matching function that measures the matching degree is: Where Δf is the frequency offset, f is the laser frequency, A is the image block in the first feature map, B is the image block in the second feature map, x is the spatial position of the image block, M is the number of positions in the fiber under test, and N is the number of modulation signals generated by the arbitrary function generator.

6. The signal demodulation method for a phase-sensitive optical time-domain reflectometer based on image processing according to claim 3, characterized in that, In step S3, image block matching is performed on the two feature maps. In the first feature map, an image block of size a×a is selected at the spatial domain coordinate x, and the coordinates of the lower left corner of the image block are used as the coordinates of the image block. In the second feature map, an image block with the initial position and initial frequency as coordinates is selected, and the matching function between this image block and the image block at the spatial domain coordinate x in the first feature map is calculated. The image block in the second feature map is moved along the frequency domain, and the matching function is calculated for each step. After the image block in the second feature map moves to the maximum value in the frequency domain, it returns to the initial frequency and moves along the spatial domain to the next position. The image block in the second feature map is moved along the frequency domain again, and the matching function is calculated for each step. This process is repeated to traverse the entire second feature map, and the matching function between the image block of size a×a at each spatial position in the second feature map and the image block at the spatial domain coordinate x in the first feature map is calculated. The position coordinate with the maximum matching function is obtained. The difference in the second dimension of this coordinate is the frequency offset under the spatial threshold coordinate, and the spatial domain curve where this coordinate is located is the matching function curve at the spatial domain coordinate x. The image block in the first feature map is moved to the next position along the spatial domain, and the matching function between each a×a image block in the second feature map and the image block at the given position is calculated. In this way, the matching function between the image block at each spatial position in the first feature map and the image block in the second feature map is obtained. The difference between the second-dimensional coordinates of the image block in the first feature map and the second-dimensional coordinates of the matching image block in the second feature map is the frequency offset. The spatial domain and the frequency offset domain are used as the first and second-dimensional coordinates, respectively, and the matching function value is used as the third-dimensional coordinate. The matching function curves of all spatial positions are used to construct a three-dimensional matching function graph.

7. The signal demodulation method for a phase-sensitive optical time-domain reflectometer based on image processing according to claim 3, characterized in that, In step S4, the BM3D process is divided into two stages: basic estimation and final estimation. Basic estimation stage: The matching function graph is divided into a series of k×k reference blocks, and all points of the image are traversed by raster scanning with a fixed step size. For the current reference block, a similar block search is performed in its surrounding neighborhood (n×n, n>k). The reference block is compared with all blocks in the window. All similar blocks are stacked to form a three-dimensional matrix according to Euclidean distance. Hard threshold collaborative filtering is performed on the three-dimensional matrix. The two-dimensional data blocks in each three-dimensional matrix are transformed using 2D-Bior1.5 wavelet, and a one-dimensional transformation is performed using 1D-Sym8 wavelet in the third dimension of the matrix. The three-dimensional transformation is achieved by combining the two-dimensional transformations. The processed image blocks are obtained through the three-dimensional inverse transformation, and the block estimates are returned to their original positions. The basic estimate of the image is calculated by weighted averaging of all obtained overlapping estimated blocks. Final estimation stage: After block matching of the matching function graph and the basic estimate, they are stacked to form a three-dimensional matrix. Both three-dimensional matrices are subjected to three-dimensional transformations. The two-dimensional transformation uses discrete cosine transform, and the one-dimensional transformation of the third dimension uses 1D-Sym8 wavelet transform. Wiener filtering is used to scale the coefficients of the three-dimensional matrix of the matching function graph. These coefficients are obtained from the values ​​of the three-dimensional matrix of the basic estimate and the noise intensity. The estimated values ​​of all image blocks are generated by the three-dimensional inverse transformation, and the estimated values ​​of the blocks are returned to their original positions. The final estimate of the image is calculated by weighted averaging of all the obtained overlapping estimated blocks.

8. The signal demodulation method for a phase-sensitive optical time-domain reflectometer based on image processing according to claim 3, characterized in that, In step S5, the relationship between the frequency offset and the temperature change is as follows: Where Δf is the frequency offset, f is the laser frequency, ζ is the thermo-optic coefficient, α is the coefficient of thermal expansion, and ΔT is the temperature change; the relationship between the frequency offset and the strain change is as follows: Where Δf is the frequency offset, f is the laser frequency, and p ε ε is the elastic coefficient, and Δε is the strain change.

9. A signal demodulation system for a phase-sensitive optical time-domain reflectometer based on image processing, comprising a computer module that runs the signal demodulation method for a phase-sensitive optical time-domain reflectometer based on image processing according to any one of claims 1 to 8.

Images