Denoising method and measurement method for optical frequency domain reflectometry image signal based on deconvolution

Abstract

The present invention provides an optical frequency domain reflectometer image signal denoising method and measurement method based on deconvolution. The denoising method includes: using FFT to convert reference and measurement signals of OFDR into distance domain reference and measurement signals; dividing the measurement scattering medium into multiple segments according to the spatial resolution of OFDR, and using STFT to convert the distance domain reference and measurement signals of each segment into time domain reference and measurement signals; calculating the cross-correlation spectrum curve of the time domain reference and measurement signals corresponding to each segment to obtain a cross-correlation two-dimensional image; performing autocorrelation on the time domain reference signals of each segment to obtain an autocorrelation two-dimensional image; using the main lobe of the cross-correlation peak corresponding to each segment in the cross-correlation two-dimensional image as the wavelength cut-off range of the segment, and cutting out an image within the wavelength cut-off range of each segment of the autocorrelation two-dimensional image as the point spread function (PSF); and using deconvolution to denoise the cross-correlation two-dimensional image based on the PSF. The present invention solves the problem that OFDR cannot perform image signal denoising based on deconvolution.

Description

Technical Field

[0001] The present invention belongs to the field of optical frequency domain reflectometry signal denoising, and in particular relates to an optical frequency domain reflectometry image signal denoising method and a measurement method based on deconvolution. Background Art

[0002] Optical frequency domain reflectometry (OFDR) boasts high spatial resolution, sensitivity, and robustness against electromagnetic interference, making it widely used in structural health monitoring, shape sensing, and industrial monitoring. OFDR technology was first proposed by M. Eickhoff in 1981. In 1998, Froggatt and colleagues studied Rayleigh scattering in optical fibers, treating it as a weak Bragg grating with random periodicity. They found that by calculating the changes in the Rayleigh scattering spectrum before and after strain application, distributed static strain could be determined through cross-correlation. Consequently, cross-correlation, which measures the shift in the Rayleigh backscattered spectrum and determines strain, has been widely adopted in OFDR sensing systems. Achieving high spatial resolution is a key reason for choosing an OFDR system. However, increasing spatial resolution introduces various noise challenges to the cross-correlation method. This noise arises from the system itself, environmental interference, and signal processing, manifesting as distortion, broadening, and increased sidelobes of the cross-correlation peak, ultimately leading to erroneous peak detection.

[0003] To address these noise challenges, image processing technology has emerged as an effective solution. Soto and colleagues first introduced image denoising techniques in Brillouin optical time-domain analysis, successfully eliminating noise from fiber optic sensors and improving sensing performance. Similar methods have also been applied to phase-sensitive optical time-domain reflectometry and Raman distributed temperature sensing systems, achieving significant results. These successful applications have promoted in-depth research and widespread application of image denoising techniques in OFDR systems.

[0004] In recent years, a variety of image denoising techniques have been applied to OFDR sensing systems to improve performance. Denoising effectiveness is generally proportional to the complexity of the denoising algorithm. Methods such as non-local means wavelet denoising, block matching 3D, non-local Haar transform, independent component analysis, and convolutional neural networks have been continuously reported. These techniques exploit the high similarity and redundancy of the multidimensional information measured by distributed fiber optic sensors to enhance the system's signal-to-noise ratio (SNR), thereby improving measurement accuracy.

[0005] Wiener deconvolution is a classic frequency-domain image denoising method that can flexibly adapt to various noise and blur conditions. By considering the combined effects of noise and blur in the frequency domain and using statistical models, Wiener deconvolution can effectively denoise and restore signal details under different conditions. Compared with other denoising algorithms, it has greater adaptability and signal recovery capabilities, demonstrating a wide range of application potential. For example, in Brillouin optical time-domain reflectometry (BOTDR) systems, the Wiener deconvolution method has been successfully applied to improve spatial resolution. Its point spread function (PSF) is obtained by two-dimensional convolution of the pulse envelope with a short-time Fourier transform window function, which is similar to the convolution of the impulse response in the time domain and frequency domain.

[0006] However, unlike the pulsed light sources used in BOTDR systems, OFDR systems use swept-frequency light sources as input, making it impossible to construct a PSF through pulse convolution in the time domain. Therefore, methods for constructing the PSF remain unclear. Consequently, effective applications of Wiener deconvolution in OFDR systems have yet to be reported. The challenge lies in obtaining an effective PSF to minimize the loss of spatial resolution during image processing. Summary of the Invention

[0007] The present invention provides an optical frequency domain reflectometer image signal denoising method and measurement method based on deconvolution, so as to solve the problem that the current optical frequency domain reflectometer (OFDR) cannot denoise its image signal based on deconvolution.

[0008] According to a first aspect of an embodiment of the present invention, a method for denoising an optical frequency domain reflectometry image signal based on deconvolution is provided, comprising: step S100, collecting a reference signal and a measurement signal of an optical frequency domain reflectometry (OFDR), performing fast Fourier transform (FFT) on the reference signal and the measurement signal, respectively, to obtain a distance domain reference signal and a measurement signal;

[0009] Step S200: Divide the measurement scattering medium in the OFDR into multiple segments according to the spatial resolution of the OFDR, and process the distance domain reference signal and measurement signal of each segment using a short-time Fourier transform (STFT) to obtain a time domain reference signal and measurement signal of the segment;

[0010] Step S300: For each segment, a cross-correlation spectrum curve between the time-domain reference signal and the measurement signal corresponding to the segment is calculated, and the cross-correlation spectrum curves corresponding to each segment are sequentially arranged along the scattering medium to obtain a cross-correlation two-dimensional image, where the two axes of the cross-correlation two-dimensional image are the distance to the scattering medium and the color gradient value used to represent the wavelength;

[0011] Step S400: Arrange the time domain reference signals corresponding to the segments in step S200 in sequence along the scattering medium to obtain a time domain reference signal corresponding to the scattering medium, perform autocorrelation on the time domain reference signal corresponding to the scattering medium, and obtain an autocorrelation two-dimensional image, where two axes of the autocorrelation two-dimensional image represent the distance to the scattering medium and a color gradient value representing the wavelength, respectively.

[0012] Step S500, extracting the point spread function (PSF): for the cross-correlation two-dimensional image, the main lobe of the cross-correlation peak of each segment is used as the wavelength cutoff range of the segment, and an image within the wavelength cutoff range of each segment of the autocorrelation two-dimensional image is extracted as the point spread function (PSF);

[0013] Step S600 : De-noising the cross-correlated two-dimensional image using a deconvolution algorithm according to the point spread function (PSF) to obtain a denoised two-dimensional image signal.

[0014] In an optional implementation, step S300 specifically includes:

[0015] When the scattering medium is arranged in a straight line, for each segment, a cross-correlation spectrum curve between the time domain reference signal and the measurement signal corresponding to the segment is calculated, and the cross-correlation spectrum curves corresponding to each segment are arranged in sequence along the layout path of the scattering medium to obtain a two-dimensional cross-correlation curve. The two-dimensional cross-correlation curve is converted into a two-dimensional cross-correlation image, where the two axes of the two-dimensional cross-correlation image are the distance along the layout path of the scattering medium and the color gradient value used to represent the wavelength;

[0016] When the scattering medium is arranged in a non-linear manner, for each segment, a cross-correlation spectrum curve between the time domain reference signal and the measurement signal corresponding to the segment is calculated, and the cross-correlation spectrum curves corresponding to each segment are arranged in sequence along the arrangement path of the scattering medium to obtain a three-dimensional cross-correlation curve;

[0017] The three-dimensional cross-correlation curve is projected onto a two-dimensional plane perpendicular to the overall layout direction of the scattering medium to form a two-dimensional cross-correlation image. The two axes of the two-dimensional cross-correlation image are the distance in the overall layout direction of the scattering medium and the color gradient value used to represent the wavelength size.

[0018] In another optional implementation, step S400 specifically includes:

[0019] When the scattering medium is arranged in a straight line, the time domain reference signals corresponding to the various segments in step S200 are sequentially arranged along the arrangement path of the scattering medium to obtain a time domain reference signal corresponding to the scattering medium. Autocorrelation is performed on the time domain reference signal corresponding to the scattering medium to obtain a two-dimensional autocorrelation curve. The two-dimensional autocorrelation curve is converted into a two-dimensional autocorrelation image, where two axes of the two-dimensional autocorrelation image are the distance along the arrangement path of the scattering medium and a color gradient value representing the wavelength, respectively.

[0020] When the scattering medium is arranged non-linearly, the time domain reference signals corresponding to the segments in step S200 are sequentially arranged along the arrangement path of the scattering medium to obtain a time domain reference signal corresponding to the scattering medium. Autocorrelation is performed on the time domain reference signal corresponding to the scattering medium to obtain a three-dimensional autocorrelation curve. The three-dimensional autocorrelation curve is projected onto a two-dimensional plane perpendicular to the overall arrangement direction of the scattering medium to form a two-dimensional autocorrelation image. The two axes of the two-dimensional autocorrelation image are the distance in the overall arrangement direction of the scattering medium and the color gradient value used to represent the wavelength.

[0021] In another optional implementation, in step S300, sequentially arranging the cross-correlation spectrum curves corresponding to each segment includes sequentially arranging the cross-correlation peaks in the cross-correlation spectrum curves corresponding to each segment.

[0022] In another optional implementation, in step S600, the deconvolution algorithm is a Wiener deconvolution algorithm.

[0023] According to a second aspect of an embodiment of the present invention, a deconvolution-based optical frequency domain reflectometry measurement method is provided, comprising:

[0024] Step S1: using the above-mentioned deconvolution-based optical frequency domain reflectometry image signal denoising method to denoise the optical frequency domain reflectometry image signal to obtain a denoised two-dimensional image signal;

[0025] Step S2, converting the denoised two-dimensional image signal into a wavelength peak shift curve;

[0026] Step S3: determining the parameters to be measured for each segment on the scattering medium according to the wavelength peak shift of each segment in the wavelength peak shift curve.

[0027] In an optional implementation, step S2 specifically includes:

[0028] When the scattering medium is arranged in a straight line, converting the denoised two-dimensional image signal into a two-dimensional wavelength peak shift curve;

[0029] When the scattering medium is arranged non-linearly, the denoised two-dimensional image signal is converted into a three-dimensional image signal based on the three-dimensional positional relationship between each segment in the denoised two-dimensional image signal and the scattering medium; and the three-dimensional image signal is converted into a three-dimensional wavelength peak shift curve.

[0030] In another optional implementation, the parameter to be measured is strain or temperature.

[0031] The beneficial effects of the present invention are:

[0032] 1. The present invention has efficient noise reduction performance: by taking the main lobe of the corresponding cross-correlation peak of each segment in the cross-correlation two-dimensional image as the wavelength cutoff range of the segment, and cutting out the image within the wavelength cutoff range of the segment from each segment of the autocorrelation two-dimensional image, as the point spread function PSF and applying the deconvolution algorithm, the noise interference in the OFDR system can be effectively reduced, the signal quality is improved, and the strain or temperature change in the optical fiber can still be accurately detected in a high-noise environment; the present invention has spatial resolution: compared with traditional image denoising methods, deconvolution can better restore signal details, avoid signal peak distortion caused by noise, and significantly improve the spatial resolution of the system, which provides strong support for the application of OFDR systems in small-scale, complex strain or temperature change environments; the present invention has flexible adaptability: the deconvolution algorithm has strong adaptability, can work efficiently under different noise and fuzzy conditions, and is suitable for multiple OFDR sensing applications in complex environments, especially in dealing with environmental interference or large system noise, can be flexibly adjusted to obtain the best signal processing effect; the present invention has the characteristic of strong signal recovery ability: compared with other denoising algorithms, the deconvolution algorithm not only performs well in noise reduction, but also can effectively restore the signal peak blurred by noise. This recovery ability is particularly important in OFDR systems and can ensure the accuracy of detection results in high-precision strain sensing; the present invention has broad application potential: the method lays the foundation for achieving high-performance and high-precision applications of OFDR systems, and has a wide range of application prospects in industrial and research fields, especially in structural health monitoring, shape sensing and other high-spatial-resolution distributed sensing applications; thus, the present invention provides an efficient, accurate and flexible solution for image denoising of OFDR systems, significantly improving the overall performance of OFDR systems;

[0033] 2. For scattering media arranged non-linearly, the present invention first arranges the corresponding cross-correlation curves of each segment in sequence along the scattering medium arrangement path to obtain a cross-correlation three-dimensional image. Then, the cross-correlation three-dimensional image is projected onto a two-dimensional plane perpendicular to the overall arrangement direction of the scattering medium to obtain a cross-correlation two-dimensional image. When obtaining an autocorrelation two-dimensional image, the time domain reference signals corresponding to each segment in step S200 are first arranged in sequence along the scattering medium arrangement path to obtain a time domain reference signal corresponding to the scattering medium. Then, cross-correlation is performed on the time domain reference signals of the scattering medium to obtain an autocorrelation three-dimensional curve. Then, the autocorrelation three-dimensional curve is projected onto a two-dimensional plane perpendicular to the overall arrangement direction of the scattering medium to form an autocorrelation two-dimensional image. This shows that even for scattering media arranged non-linearly, the present invention can perform denoising processing on its image signal based on the deconvolution method.

[0034] 3. The present invention arranges the cross-correlation peaks in the cross-correlation spectrum curve in sequence when arranging the cross-correlation spectrum curve, thereby removing part of the noise and improving the signal-to-noise ratio of the cross-correlation two-dimensional image;

[0035] 4. The present invention selects the main lobe of the corresponding cross-correlation peak in each segment of the cross-correlation two-dimensional image as the wavelength cutoff range of the segment, which not only covers the main features of the cross-correlation two-dimensional image but also avoids over-expansion. Then, based on the wavelength cutoff range, the image within the wavelength cutoff range is cut off from each segment of the autocorrelation two-dimensional image as the point spread function (PSF). This can effectively construct a PSF that can efficiently remove noise while retaining valuable information and improving measurement accuracy.

[0036] 5. The present invention performs denoising on the image signal of the optical frequency domain reflectometer (OFDR) based on a deconvolution algorithm, making full use of the similarity and redundancy of multidimensional data in the OFDR system, reducing the impact of noise and enhancing signal quality. The present invention significantly improves the spatial resolution and signal detection accuracy of the OFDR system measurement, and effectively improves the distributed strain sensing performance of the OFDR system. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of an embodiment of a method for denoising an optical frequency domain reflectometry image signal based on deconvolution according to the present invention;

[0038] Figure 2 1 is a schematic diagram of a method for denoising an optical frequency domain reflectometer image signal based on deconvolution according to the present invention;

[0039] Figure 3 This is a comparison chart of the denoising effects of the present invention and traditional Gaussian filtering;

[0040] Figure 4The present invention is used to test the result of denoising at a high spatial resolution of 0.2mm.

[0041] Figure 5 This is a flow chart of an embodiment of the optical frequency domain reflectometry measurement method based on deconvolution of the present invention. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention and to make the above-mentioned purposes, features and advantages of the embodiments of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention are further described in detail below with reference to the accompanying drawings.

[0043] In the description of the present invention, unless otherwise specified and limited, it should be noted that the term "connection" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the internal connection between two elements. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meaning of the above terms can be understood according to the specific circumstances.

[0044] See also Figure 1 , is a flow chart of an embodiment of the method for denoising an optical frequency domain reflectometer image signal based on deconvolution of the present invention. Figure 2 As shown, the optical frequency domain reflectometry image signal denoising method based on deconvolution may include the following steps:

[0045] Step S100: Collect a reference signal and a measurement signal from an optical frequency domain reflectometer (OFDR), and perform fast Fourier transforms (FFTs) on the reference signal and the measurement signal, respectively, to obtain a distance domain reference signal and a measurement signal. The reference signal is a signal collected when no event occurs in the scattering medium used for measurement in the OFDR, and the measurement signal is a signal collected when an event occurs in the scattering medium used for measurement in the OFDR. The event may be a spectral peak shift caused by stress, temperature, or other factors. The signal collected by the optical frequency domain reflectometer (OFDR) is a frequency domain signal, which is converted into a distance domain signal through FFT processing. The collected reference signal and measurement signal may be Rayleigh scattered light backreflected by the scattering medium, which may be a distributed sensing fiber.

[0046] Step S200: Divide the measurement scattering medium in the OFDR into multiple segments according to the spatial resolution of the OFDR. For each segment, process the distance domain reference signal and the measurement signal of the segment using short-time Fourier transform (STFT) to obtain the time domain reference signal and the measurement signal of the segment.

[0047] Step S300: For each segment, calculate the cross-correlation spectrum curve between the time domain reference signal and the measurement signal corresponding to the segment, and arrange the cross-correlation spectrum curves corresponding to each segment in sequence along the scattering medium to obtain a cross-correlation two-dimensional image. The two axes of the cross-correlation two-dimensional image are the distance to the scattering medium and the color gradient value used to represent the wavelength size.

[0048] Clearly, the two-dimensional cross-correlation data structure of each segment of the scattering medium contains repetitive data signal patterns, making it effective for noise reduction. Given that OFDR measurement data contains a large amount of highly similar and redundant information, image processing can be used to reduce noise in the two-dimensional image. Compared to traditional one-dimensional filtering methods, image processing is more effective in noise reduction because it utilizes the full range of data signal patterns present in the two-dimensional measurement data. After image denoising, the quality of the cross-correlation curve is improved, thereby enhancing the accuracy of detecting spectral shift peaks caused by strain / temperature changes.

[0049] Step S300 may specifically include: when the scattering medium is arranged in a straight line, calculating, for each segment, a cross-correlation spectrum curve between a time-domain reference signal and a measurement signal corresponding to the segment, sequentially arranging the cross-correlation spectrum curves corresponding to the segments along the arrangement path of the scattering medium to obtain a two-dimensional cross-correlation curve, and converting the two-dimensional cross-correlation curve into a two-dimensional cross-correlation image, where two axes of the two-dimensional cross-correlation image are a distance along the arrangement path of the scattering medium and a color gradient value representing a wavelength, respectively;

[0050] When the scattering medium is arranged non-linearly, for each segment, a cross-correlation spectral curve between the time-domain reference signal and the measurement signal corresponding to that segment is calculated. The cross-correlation spectral curves corresponding to each segment are sequentially arranged along the scattering medium to obtain a three-dimensional cross-correlation curve. The three-dimensional cross-correlation curve is then projected onto a two-dimensional plane perpendicular to the overall arrangement direction of the scattering medium to form a two-dimensional cross-correlation image. The two axes of the two-dimensional cross-correlation image are the distance along the overall arrangement direction of the scattering medium and the color gradient value representing the wavelength. In step S300, sequentially arranging the cross-correlation spectral curves corresponding to each segment may include sequentially arranging the cross-correlation peaks in the cross-correlation spectral curves corresponding to each segment. The present invention arranges the cross-correlation spectral curves by sequentially arranging the cross-correlation peaks in the cross-correlation spectral curves, thereby removing some noise and improving the signal-to-noise ratio of the two-dimensional cross-correlation image.

[0051] Step S400: Arrange the time domain reference signals corresponding to the segments in step S200 in sequence along the scattering medium to obtain a time domain reference signal corresponding to the scattering medium, perform autocorrelation on the time domain reference signal corresponding to the scattering medium, and obtain an autocorrelation two-dimensional image, where two axes of the autocorrelation two-dimensional image are the distance to the scattering medium and a color gradient value used to represent the wavelength, respectively.

[0052] The step S400 may specifically include:

[0053] When the scattering medium is arranged in a straight line, the time domain reference signals corresponding to the various segments in step S200 are sequentially arranged along the arrangement path of the scattering medium to obtain a time domain reference signal corresponding to the scattering medium. Autocorrelation is performed on the time domain reference signal corresponding to the scattering medium to obtain a two-dimensional autocorrelation curve. The two-dimensional autocorrelation curve is converted into a two-dimensional autocorrelation image, where two axes of the two-dimensional autocorrelation image are the distance along the arrangement path of the scattering medium and a color gradient value representing the wavelength, respectively.

[0054] When the scattering medium is arranged non-linearly, the time domain reference signals corresponding to the segments in step S200 are sequentially arranged along the arrangement path of the scattering medium to obtain a time domain reference signal corresponding to the scattering medium. Autocorrelation is performed on the time domain reference signal corresponding to the scattering medium to obtain a three-dimensional autocorrelation curve. The three-dimensional autocorrelation curve is projected onto a two-dimensional plane perpendicular to the overall arrangement direction of the scattering medium to form a two-dimensional autocorrelation image. The two axes of the two-dimensional autocorrelation image are the distance in the overall arrangement direction of the scattering medium and the color gradient value used to represent the wavelength.

[0055] The present invention targets scattering media arranged non-linearly. When acquiring a cross-correlation two-dimensional image, the corresponding cross-correlation curves of each segment are first arranged in sequence along the scattering medium arrangement path to obtain a cross-correlation three-dimensional image. The cross-correlation three-dimensional image is then projected onto a two-dimensional plane perpendicular to the overall arrangement direction of the scattering medium to obtain a cross-correlation two-dimensional image. When acquiring an autocorrelation two-dimensional image, the time-domain reference signals corresponding to each segment in step S200 are first arranged in sequence along the scattering medium arrangement path to obtain a time-domain reference signal corresponding to the scattering medium. The time-domain reference signals of the scattering medium are then cross-correlated to obtain a three-dimensional autocorrelation curve. The three-dimensional autocorrelation curve is then projected onto a two-dimensional plane perpendicular to the overall arrangement direction of the scattering medium to form an autocorrelation two-dimensional image. This shows that even for scattering media arranged non-linearly, the present invention can perform denoising processing on its image signal based on the deconvolution method.

[0056] Step S500, intercepting the point spread function (PSF): for the cross-correlation two-dimensional image, the main lobe of the cross-correlation peak of each segment is used as the wavelength interception range of the segment, and an image within the wavelength interception range corresponding to the segment is intercepted from each segment of the autocorrelation two-dimensional image as the point spread function (PSF).

[0057] In this step, deconvolution is often used to reduce image noise, obtaining the wavelength peak offset along the fiber from the de-noised image. For OFDR systems, the most critical step is constructing the point spread function (PSF). In optical time-domain reflectometry (OTDR) systems, the detection light source is pulsed light, and the detected signal is the convolution of the pulse envelope and the system impulse response. In contrast, because OFDR uses a swept-frequency light source for detection rather than a pulsed light source, representing the process as a convolution in the spatial domain is more difficult. In OFDR's spectral demodulation method, the ideal signal should be infinite, but window truncation of the signal causes spectral leakage, which obscures the quality of the autocorrelation peak in the wavelength domain. In BOTDR systems, the PSF is the convolution of the pulse envelope and a window function. Similarly, in OFDR systems, the PSF can be defined as the convolution of the reference signal and the measured signal with a window function. When the reference signal and the measured signal are identical, the system's PSF is more reliable. Therefore, the autocorrelation function can be used as the PSF of the OFDR system.

[0058] The cross-correlation peak corresponding to each segment in the cross-correlation two-dimensional image includes a main lobe and side lobes located on both sides of the main lobe, so the main lobe of each segment corresponding to the cross-correlation peak can be accurately determined. When selecting a wavelength cutoff range, if it is too large, it will cause the peak of the denoised image signal to be broadened and blurred, thereby reducing spatial resolution and wavelength accuracy. The present invention selects the main lobe of the cross-correlation peak corresponding to each segment in the cross-correlation two-dimensional image as the wavelength cutoff range of that segment, which not only covers the main features of the cross-correlation two-dimensional image but also avoids excessive expansion. Based on this wavelength cutoff range, an image within the wavelength cutoff range is then cut off from each segment of the autocorrelation two-dimensional image as the point spread function (PSF). This effectively constructs a PSF that can efficiently remove noise while retaining valuable information and improving measurement accuracy.

[0059] Step S600: De-noise the cross-correlated two-dimensional image using a deconvolution algorithm based on the point spread function (PSF) to obtain a de-noised two-dimensional image signal. The deconvolution algorithm may be a Wiener deconvolution algorithm.

[0060] It can be seen from the above embodiments that the present invention has efficient noise reduction performance: by taking the main lobe of the corresponding cross-correlation peak of each segment in the cross-correlation two-dimensional image as the wavelength cutoff range of the segment, and cutting out the image within the wavelength cutoff range corresponding to the segment from each segment of the autocorrelation two-dimensional image, as the point spread function PSF and applying the deconvolution algorithm, the noise interference in the OFDR system can be effectively reduced, the signal quality is improved, and the strain or temperature change in the optical fiber can still be accurately detected in a high-noise environment; the present invention has spatial resolution: compared with traditional image denoising methods, deconvolution can better restore signal details, avoid signal peak distortion caused by noise, and significantly improve the spatial resolution of the system, which provides strong support for the application of OFDR systems in small-scale, complex strain or temperature change environments; the present invention has flexible adaptability: the deconvolution algorithm has strong adaptability, can work efficiently under different noise and fuzzy conditions, and is suitable for It can be used for OFDR sensing applications in a variety of complex environments, especially in dealing with environmental interference or large system noise, and can be flexibly adjusted to obtain the best signal processing effect; the present invention has the characteristic of strong signal recovery ability: compared with other denoising algorithms, the deconvolution algorithm not only performs well in noise reduction, but also can effectively restore signal peaks blurred by noise. This recovery ability is particularly important in OFDR systems and can ensure the accuracy of detection results in high-precision strain sensing; the present invention has broad application potential: this method lays the foundation for realizing high-performance and high-precision applications of OFDR systems, and has application prospects in a wide range of industrial and research fields, especially in structural health monitoring, shape sensing and other high-spatial-resolution distributed sensing applications; thus, the present invention provides an efficient, accurate and flexible solution for image denoising of OFDR systems, significantly improving the overall performance of OFDR systems.

[0061] Figure 3 This is a comparison chart of the denoising effects of the present invention and traditional Gaussian filtering, where Figure 3 (a) is the strain gradient information before denoising, Figure 3 (b) is the denoising effect of Gaussian filtering. Figure 3 (c) is the deconvolution denoising effect of the present invention. It can be seen from the figure that the denoising effect of the present invention is better.

[0062] Figure 4 The present invention is used to test the noise reduction result at a high spatial resolution of 0.2 mm. Figure 4 (a), (c) and (e) show the signals before denoising. Figure 4 (b), (d) and (f) show the denoised signals. It can be seen from the figures that the denoising effect of the present invention is excellent.

[0063] In addition, the present invention also provides a deconvolution-based optical frequency domain reflectometry measurement method, which may include the following steps:

[0064] Step S1: Using the above-mentioned deconvolution-based optical frequency domain reflectometry image signal denoising method, denoise the optical frequency domain reflectometry image signal to obtain a denoised two-dimensional image signal.

[0065] Step S2: converting the denoised two-dimensional image signal into a wavelength peak shift curve.

[0066] The step S2 may specifically include: when the scattering medium is arranged in a straight line, converting the denoised two-dimensional image signal into a two-dimensional wavelength peak shift curve;

[0067] When the scattering medium is arranged non-linearly, the denoised two-dimensional image signal is converted into a three-dimensional image signal based on the three-dimensional positional relationship between each segment in the denoised two-dimensional image signal and the scattering medium; and the three-dimensional image signal is converted into a three-dimensional wavelength peak shift curve.

[0068] Step S3: Determine a parameter to be measured for each segment on the scattering medium based on the wavelength peak shift of each segment in the wavelength peak shift curve. The parameter to be measured may be strain or temperature. Each segment in the wavelength peak shift curve corresponds to a corresponding segment on the scattering medium.

[0069] As can be seen from the above embodiments, the present invention denoises the image signal of the optical frequency domain reflectometer (OFDR) based on the deconvolution algorithm, fully utilizing the similarity and redundancy of multidimensional data in the OFDR system, reducing the impact of noise and enhancing signal quality. The present invention significantly improves the spatial resolution and signal detection accuracy of the OFDR system measurement, and effectively improves the distributed strain sensing performance of the OFDR system.

[0070] Other embodiments of the present invention will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

[0071] It will be appreciated that the present invention is not limited to the precise construction that has been described above and shown in the accompanying drawings, and that various modifications and variations can be made without departing from its scope, which is governed solely by the appended claims.

Claims

1. A method for denoising optical frequency domain reflectometry image signals based on deconvolution, characterized in that: include: Step S100: collecting a reference signal and a measurement signal of an optical frequency domain reflectometer (OFDR), performing fast Fourier transform (FFT) on the reference signal and the measurement signal, respectively, to obtain a distance domain reference signal and a measurement signal; Step S200: Divide the measurement scattering medium in the OFDR into multiple segments according to the spatial resolution of the OFDR, and process the distance domain reference signal and measurement signal of each segment using a short-time Fourier transform (STFT) to obtain a time domain reference signal and measurement signal of the segment; Step S300: For each segment, a cross-correlation spectrum curve between the time-domain reference signal and the measurement signal corresponding to the segment is calculated, and the cross-correlation spectrum curves corresponding to each segment are sequentially arranged along the scattering medium to obtain a cross-correlation two-dimensional image, where the two axes of the cross-correlation two-dimensional image are the distance to the scattering medium and the color gradient value used to represent the wavelength; Step S400: Arrange the time domain reference signals corresponding to the segments in step S200 in sequence along the scattering medium to obtain a time domain reference signal corresponding to the scattering medium, perform autocorrelation on the time domain reference signal corresponding to the scattering medium, and obtain an autocorrelation two-dimensional image, where two axes of the autocorrelation two-dimensional image represent the distance to the scattering medium and a color gradient value representing the wavelength, respectively. Step S500, extracting the point spread function (PSF): for the cross-correlation two-dimensional image, the main lobe of the cross-correlation peak of each segment is used as the wavelength cutoff range of the segment, and an image within the wavelength cutoff range of each segment of the autocorrelation two-dimensional image is extracted as the point spread function (PSF); Step S600 : De-noising the cross-correlated two-dimensional image using a deconvolution algorithm according to the point spread function (PSF) to obtain a denoised two-dimensional image signal.

2. The method for denoising an optical frequency domain reflectometry image signal based on deconvolution according to claim 1, characterized in that: The step S300 specifically includes: When the scattering medium is arranged in a straight line, for each segment, a cross-correlation spectrum curve between the time domain reference signal and the measurement signal corresponding to the segment is calculated, and the cross-correlation spectrum curves corresponding to each segment are arranged in sequence along the layout path of the scattering medium to obtain a two-dimensional cross-correlation curve. The two-dimensional cross-correlation curve is converted into a two-dimensional cross-correlation image, where the two axes of the two-dimensional cross-correlation image are the distance along the layout path of the scattering medium and the color gradient value used to represent the wavelength; When the scattering medium is arranged in a non-linear manner, for each segment, a cross-correlation spectrum curve between the time domain reference signal and the measurement signal corresponding to the segment is calculated, and the cross-correlation spectrum curves corresponding to each segment are arranged in sequence along the arrangement path of the scattering medium to obtain a three-dimensional cross-correlation curve; The three-dimensional cross-correlation curve is projected onto a two-dimensional plane perpendicular to the overall layout direction of the scattering medium to form a two-dimensional cross-correlation image. The two axes of the two-dimensional cross-correlation image are the distance in the overall layout direction of the scattering medium and the color gradient value used to represent the wavelength size.

3. The method for denoising an optical frequency domain reflectometry image signal based on deconvolution according to claim 1 or 2, wherein: The step S400 specifically includes: When the scattering medium is arranged in a straight line, the time domain reference signals corresponding to the segments in step S200 are sequentially arranged along the arrangement path of the scattering medium to obtain a time domain reference signal corresponding to the scattering medium. Autocorrelation is performed on the time domain reference signal corresponding to the scattering medium to obtain a two-dimensional autocorrelation curve. The two-dimensional autocorrelation curve is converted into a two-dimensional autocorrelation image, where two axes of the two-dimensional autocorrelation image are the distance along the arrangement path of the scattering medium and a color gradient value representing the wavelength, respectively. When the scattering medium is arranged non-linearly, the time domain reference signals corresponding to the segments in step S200 are sequentially arranged along the arrangement path of the scattering medium to obtain a time domain reference signal corresponding to the scattering medium. Autocorrelation is performed on the time domain reference signal corresponding to the scattering medium to obtain a three-dimensional autocorrelation curve. The three-dimensional autocorrelation curve is projected onto a two-dimensional plane perpendicular to the overall arrangement direction of the scattering medium to form a two-dimensional autocorrelation image. The two axes of the two-dimensional autocorrelation image are the distance in the overall arrangement direction of the scattering medium and the color gradient value used to represent the wavelength.

4. The method for denoising an optical frequency domain reflectometry image signal based on deconvolution according to claim 1 or 2, wherein: In the step S300, sequentially arranging the cross-correlation spectrum curves corresponding to each segment includes sequentially arranging the cross-correlation peaks in the cross-correlation spectrum curves corresponding to each segment.

5. The method for denoising an optical frequency domain reflectometry image signal based on deconvolution according to claim 1, wherein: In step S600, the deconvolution algorithm is a Wiener deconvolution algorithm.

6. A deconvolution-based optical frequency domain reflectometry measurement method, characterized in that: include: Step S1: using the optical frequency domain reflectometry image signal denoising method based on deconvolution according to any one of claims 1 to 5 to denoise the image signal of the optical frequency domain reflectometry to obtain a denoised two-dimensional image signal; Step S2, converting the denoised two-dimensional image signal into a wavelength peak shift curve; Step S3: determining the parameters to be measured for each segment on the scattering medium according to the wavelength peak shift of each segment in the wavelength peak shift curve.

7. The optical frequency domain reflectometry measurement method based on deconvolution according to claim 6, characterized in that: The step S2 specifically includes: When the scattering medium is arranged in a straight line, converting the denoised two-dimensional image signal into a two-dimensional wavelength peak shift curve; When the scattering medium is arranged non-linearly, the denoised two-dimensional image signal is converted into a three-dimensional image signal based on the three-dimensional positional relationship between each segment in the denoised two-dimensional image signal and the scattering medium; and the three-dimensional image signal is converted into a three-dimensional wavelength peak shift curve.

8. The optical frequency domain reflectometry measurement method based on deconvolution according to claim 6, characterized in that: The parameter to be measured is strain or temperature.

Images