A high spatial resolution OFDR data processing method based on blind source separation

Abstract

The application discloses a high spatial resolution OFDR data processing method based on blind source separation, which first converts a traditional one-dimensional cross-correlation processing result signal into a two-dimensional image signal, carries out denoising processing on the image based on the two-dimensional image through an independent component analysis algorithm, carries out next operation on the processed image, and finally obtains a high spatial resolution result. The OFDR sensing system of the independent component analysis algorithm provided in the application can improve the spatial resolution of a measurement system, so that the application has greater advantages and broader application in the fields of high-precision monitoring such as aerospace, machines and equipment.

Description

Technical Field

[0001] This invention belongs to the technical field of fiber optic sensing and detection, specifically relating to a high spatial resolution OFDR data processing method based on blind source separation. Background Technology

[0002] In distributed fiber optic sensing, optical fiber serves as both the sensing and transmission medium for measurements. Utilizing the transmission characteristics of light waves in the fiber, including Raman scattering, Rayleigh scattering, and Brillouin scattering, real-time monitoring of the external environment along the fiber's length is achieved. Distributed fiber optic sensing technology boasts advantages such as strong resistance to electromagnetic interference, relatively simple structure, high spatial resolution, and long sensing distance. Based on these advantages, this technology is increasingly being applied in various fields, such as bridge safety monitoring, civil engineering inspection, fire alarms in tunnels and other underground structures, and geological surveying, playing a significant role in social development. As a representative of distributed fiber optic sensing systems, optical frequency domain reflectance (OFDR) technology offers advantages such as light weight, small size, high sensitivity, strong resistance to electromagnetic interference, and high spatial resolution. It can continuously measure changes in external physical quantities such as strain, vibration, and temperature along the fiber optic distance. In recent years, with the development of OFDR technology, shape sensing and acoustic sensing applications have also been realized. OFDR systems are characterized by high spatial resolution, reaching the millimeter level, making them crucial for high-precision monitoring in aerospace and other fields. However, as the spatial resolution of the measurement increases, the cross-correlation between the reference and test signals decreases significantly, leading to multiple peaks and spurious peaks in the cross-correlation results, resulting in incorrect information. Therefore, effectively improving the spatial resolution of OFDR systems is a crucial research direction. Summary of the Invention

[0003] To address the aforementioned problems, this application provides a high spatial resolution OFDR data processing method based on blind source separation, which effectively improves the spatial resolution of the OFDR system without changing the system structure or increasing costs. The technical solution is as follows:

[0004] A high spatial resolution OFDR data processing method based on blind source separation includes the following steps:

[0005] S1. Acquire two signals separately: one signal without strain information, which is the reference signal; and the other signal containing strain information, which is the test signal.

[0006] S2. Divide the reference signal and the test signal into N equal parts in the distance domain according to a certain window size C;

[0007] S3. Apply a fast inverse Fourier transform to each of the range domain information of the reference signal and the test signal;

[0008] S4. Perform cross-correlation calculation between the reference signal after inverse Fourier transform and the test signal to obtain the one-dimensional cross-correlation result;

[0009] S5. Repeat steps S3-S4 to obtain the cross-correlation results for each corresponding position of the optical fiber. Rearrange all the obtained one-dimensional cross-correlation results as a function of the optical fiber distance to form a two-dimensional image signal A.

[0010] S6. Use the simulated graphs with the same statistical distribution as the generated noise-free two-dimensional images of the cross-correlation results as the training set B;

[0011] S7. For the training set B obtained in S6, select a portion of the image blocks as training sub-image blocks, and perform mean removal and whitening processing on them to obtain noise-free data b.

[0012] S8. Process the noise-free data b using the FastICA algorithm to obtain the mixing matrix W. k Then calculate the separation matrix.

[0013] S9. Probability density of each component s i Depend on It is estimated that W i It is the separation matrix of the i-th component; by s i Determine the maximum likelihood function of the contraction function g(u);

[0014] S10. Perform the same preprocessing as step S7 on the two-dimensional image signal A obtained in S5 to obtain noisy data a. Perform independent component analysis transformation on a through y = Wa to obtain the projection y of the two-dimensional image signal A under the separation matrix W.

[0015] S11. Obtain the denoised estimate using the maximum likelihood function of the contraction function g(u) from step S9. Inverse transformation through independent component analysis Obtain low-noise cross-correlation two-dimensional image estimation

[0016] S12. Will By reconstructing the optical fiber, the spectral shift at each position is obtained, thus enabling measurement results with high spatial resolution.

[0017] Preferably, the method for obtaining the spectral shift in step S12 is as follows:

[0018] Will The result is decomposed back to the corresponding position of the fiber using the fiber distance as a function, and the shift of the spectrum at the corresponding fiber position is obtained by finding the shift of the main peak.

[0019] An OFDR system includes a coupler 1, a coupler 2, a Mach-Zehnder interferometer, a data acquisition card, a polarization controller 1, a polarization controller 2, and a Fresnel ring. The continuous laser output of a tunable laser source is split into two parts by the coupler 1. 10% of the output is incident on an unbalanced Mach-Zehnder interferometer, providing a trigger signal to the data acquisition card. The remaining portion of the light enters the coupler 2. The coupler 2 then splits the output into two parts: 1% of the output is adjusted by the polarization controller 1 to ensure that the "p" and "s" light components have the same power; the remaining 99% passes through the circulator and the polarization controller 2 and enters the sensing fiber for detection. The Fresnel ring is used to suppress Fresnel reflection at the fiber end. The Rayleigh scattering signal is then combined with the 1% laser output from the coupler 3 to obtain an interference signal, which is then decomposed into "p" and "s" components by a polarization beam splitter. Finally, the "p" and "s" light are acquired by the data acquisition card.

[0020] Beneficial effects

[0021] 1) The OFDR sensing system based on the Gaussian filtering denoising algorithm proposed in this invention can improve the spatial resolution of the measurement system, giving it greater advantages and wider applications in high-precision monitoring fields such as aerospace and machinery.

[0022] 2) The OFDR sensing system based on independent component analysis denoising algorithm proposed in this invention can not only improve the spatial resolution of the system by denoising two-dimensional image information, but also effectively remove outliers in the measurement results and improve the accuracy of measurement. Attached Figure Description

[0023] Figure 1 This is a flowchart of the application processing.

[0024] Figure 2 This is a schematic diagram of an OFDR system.

[0025] Wherein 1- is tunable laser; 2- is coupler one; 3- is coupler two; 4- is circulator; 5- is Mach-Zehnder interferometer; 6- is polarization controller one; 7- is polarization controller two; 8- is coupler three; 9- is polarization beam splitter; 10- is balanced detector; 11- is data acquisition card; 12- is sensing fiber; 13- is Fresnel ring.

[0026] Figure 3 The image shows the result of using this method when the sensing fiber is subjected to 100 με at a depth of 70.1 m. The spatial resolution is 0.4 mm.

[0027] Figure 4 The image shows the results of using this method to sense a 100με sensor at a distance of 70.1m in the optical fiber, with a spatial resolution of 0.4mm.

[0028] Figure 5This is a training set of five simulated noise-free cross-correlation images generated by Matlab, where the white stripe in the middle simulates a strain-free location, and the stripes on both sides simulate strained locations. Detailed Implementation

[0029] The following is in conjunction with the appendix Figure 1-5 The present invention will be further described with reference to specific embodiments to aid in understanding the invention.

[0030] A high spatial resolution OFDR data processing method based on blind source separation includes the following steps:

[0031] S1. Acquire two signals separately: one signal without strain information, which is the reference signal; and the other signal with strain information, which is the test signal.

[0032] S2. Divide the reference signal and the test signal into N equal parts in the distance domain according to a certain window size C;

[0033] S3. Apply a fast inverse Fourier transform to each of the range domain information of the reference signal and the test signal;

[0034] S4. By performing cross-correlation between the reference signal after inverse Fourier transform and the test signal, a one-dimensional cross-correlation result can be obtained;

[0035] S5. Repeat steps S3-S4 to obtain the cross-correlation results for each corresponding position of the optical fiber. Rearrange all the obtained one-dimensional cross-correlation results as a function of the optical fiber distance to form a two-dimensional image signal A.

[0036] S6. Use Matlab to generate several noise-free two-dimensional images of the cross-correlation results with the same statistical distribution as the training set B;

[0037] S7. For the training set B obtained in S6, select a portion of the image blocks as training sub-image blocks, and perform mean removal and whitening processing on them to obtain noise-free data b.

[0038] S8. Process the noise-free data b using the FastICA algorithm to obtain the mixing matrix W. k Then calculate the separation matrix.

[0039] S9. Probability density of each component s i Depend on It is estimated that W i It is the separation matrix of the i-th component; by s i Determine the maximum likelihood function of the contraction function g(u), where u has no specific meaning and is just a variable that can refer to s or y above;

[0040] S10. Perform the same preprocessing as step S7 on the two-dimensional image signal A obtained in S5 to obtain noisy data a. Perform independent component analysis transformation on a through y = Wa to obtain the projection y of the two-dimensional image signal A under the separation matrix W.

[0041] S11. Obtain the denoised estimate using the maximum likelihood function of the contraction function g(u) from step S9. Inverse transformation through independent component analysis Obtain low-noise cross-correlation two-dimensional image estimation

[0042] S12. The image processed in step S11 is... By reconstructing the optical fiber, the spectral offset at each position is obtained, thus enabling measurement results with high spatial resolution and improving measurement accuracy.

[0043] The method for obtaining the spectral shift in step S12 is as follows:

[0044] The two-dimensional image signal reconstructed in step S11 The distance between optical fibers is decomposed into corresponding positions in the fiber, with each pixel representing a unit of distance. The shift of the main peak is then used to obtain the spectral shift of the corresponding optical fiber position.

[0045] Figure 5 In the two-dimensional image simulation, there are two bright stripes. The stripes closer to the center represent locations with no strain information, while those further away from the center represent locations with strain information.

[0046] Example 2

[0047] Figure 3 The image shows the results of a 400 με strain applied at 73.6 m of the sensing fiber without using this method. The spatial resolution is 2 mm. It can be seen that there are many outliers in the results, and the correct strain distribution along the fiber length cannot be obtained.

[0048] As shown in the figure, applying a micro-strain of 400 με to the fiber from 73.6 m to 74 m, without using this method, results in many outliers in the 400 με micro-strain sensing results within this sensing range, due to the system's high spatial resolution (2 mm). This leads to the inability to obtain the correct strain distribution along the fiber length.

[0049] Figure 4 The image shows the results of using this method on a 73.6m sensing fiber subjected to 400με. This method effectively eliminates outliers, improves system resolution, and yields accurate strain distribution results with a spatial resolution of 2mm.

[0050] As can be seen from the figure, when a micro-strain of 400 με is applied to the optical fiber at a distance of 73.6 m to 74 m, the results obtained within the sensing range are mostly around the value of 400 με, even though the spatial resolution of the system is high (2 mm). This indicates that the method effectively eliminates outliers, allowing the system to obtain correct strain distribution results even at high spatial resolution.

[0051] Example 3

[0052] Figure 2 This is a schematic diagram of an OFDR system. An OFDR sensing system based on range-domain compensation includes: a continuous laser output from a tunable laser source is split into two parts by coupler 2 (a 10 / 90 optical coupler). 10% of the output is incident on an unbalanced Mach-Zehnder interferometer 5, providing a trigger signal to the acquisition card 11. The remaining portion of the light enters coupler 3. Coupler 3 (a 1 / 99 optical coupler) then splits the output into two parts. 1% of the output is adjusted by polarization controller 6 to ensure that the "p" and "s" light components have the same power. The remaining 99% passes through circulator 4 and polarization controller 7 and enters the sensing fiber 12 for detection. A Fresnel ring 13 is used to suppress Fresnel reflection at the fiber end. Then, the Rayleigh scattering signal is combined with the 1% laser output from coupler 8 (a 50 / 50 optical coupler) to obtain an interference signal, which is then decomposed into "p" and "s" components by a polarization beam splitter. Finally, the "p" and "s" light are acquired by the acquisition card.

[0053] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A high spatial resolution OFDR data processing method based on blind source separation, characterized in that, Includes the following steps: S1. Acquire two signals separately: one signal without strain information, which is the reference signal; and the other signal containing strain information, which is the test signal. S2. Divide the reference signal and the test signal into N equal parts in the distance domain according to a certain window size C; S3. Apply a fast inverse Fourier transform to each of the range domain information of the reference signal and the test signal; S4. Perform cross-correlation calculation between the reference signal after inverse Fourier transform and the test signal to obtain the one-dimensional cross-correlation result; S5. Repeat steps S3-S4 to obtain the cross-correlation results for each corresponding position of the optical fiber. Rearrange all the obtained one-dimensional cross-correlation results as a function of the optical fiber distance to form a two-dimensional image signal A. S6. Use the simulated graphs with the same statistical distribution as the generated noise-free two-dimensional images of the cross-correlation results as the training set B; S7. For the training set B obtained in S6, select a portion of the image blocks as training sub-image blocks, and perform mean removal and whitening processing on them to obtain noise-free data b. S8. Process the noise-free data b using the FastICA algorithm to obtain the mixing matrix W. k Then calculate the separation matrix. S9. Probability density of each component s i Depend on It is estimated that W i It is the separation matrix of the i-th component; by s i Determine the maximum likelihood function of the contraction function g(u); S10. Perform the same preprocessing as step S7 on the two-dimensional image signal A obtained in S5 to obtain noisy data a. Perform independent component analysis transformation on a through y = Wa to obtain the projection y of the two-dimensional image signal A under the separation matrix W. S11. Obtain the denoised estimate using the maximum likelihood function of the contraction function g(u) from step S9. Inverse transformation through independent component analysis Obtain low-noise cross-correlation two-dimensional image estimation S12. Will By reconstructing the optical fiber, the spectral shift at each position is obtained, thus enabling measurement results with high spatial resolution.

2. The high spatial resolution OFDR data processing method based on blind source separation according to claim 1, characterized in that, The method for obtaining the spectral shift in step S12 is as follows: Will The result is decomposed back to the corresponding position of the fiber using the fiber distance as a function, and the shift of the spectrum at the corresponding fiber position is obtained by finding the shift of the main peak.

Images