Method for reducing the influence of noise on open optical path spectral detection signals

Abstract

The application discloses a method for reducing the influence of open optical path spectrum detection signal noise, and steps of the method comprise the following steps: 1, decomposing the detection signal by VMD to obtain K sub-sequences; 2, screening the effective signal part in each sub-sequence, filtering and smoothing, and then reconstructing the filtered sub-sequences; 3, performing baseline fitting on the reconstructed noise-reduced signal to calculate the absorption curve, performing Voigt fitting on the absorption curve to obtain the absorption signal, and finally calculating the gas concentration after eliminating the open optical path detection noise according to the Beer-Lambert law. The application can solve the flicker noise problem caused by the external environment interference of the instrument without increasing the hardware structure, ensure the measurement accuracy, play the convenience of open space measurement, and thus improve the accuracy of the development space trace gas detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of environmental optics, and in particular to a method for reducing the noise influence of spectral detection signals in an open optical path. BACKGROUND

[0002] Laser absorption spectroscopy technology based on an open optical path has the advantages of high measurement accuracy, good selectivity, and fast system response speed. However, when detecting in a large-scale open space, various noises of different frequencies are caused by the long detection optical path and external environmental factors, including vibration, temperature, and other interfering gases, which can cause the effective signal to be submerged in noise, making the detection accuracy not ideal. Therefore, effectively reducing the noise influence in spectral detection signals is of great significance for ensuring accurate measurement.

[0003] Many existing studies mostly analyze the laser spectral signal in the time domain and apply digital filtering methods to reduce noise. However, these methods are mostly applied to spectral signals of small-scale indoor detection systems and cannot be directly applied to detection signals in large-scale open areas. There are few reports on methods for eliminating the noise of detection signals in large-scale open optical paths. SUMMARY

[0004] In view of the deficiencies of existing technical methods, the present application proposes a method for reducing the noise influence of spectral detection signals in an open optical path, which can effectively improve the accuracy of open space micro-gas detection without increasing the hardware structure.

[0005] The technical solution of the present application to solve the above problems is as follows:

[0006] The method for reducing the noise influence of spectral detection signals in an open optical path is characterized by being applied to a detection system composed of a laser, a beam splitter, a laser controller, a signal generator, a transceiver telescope, an angle reflector, a signal processing module, and a collection module connected in sequence, and the following steps are performed:

[0007] Step 1. The beam splitter splits the detection laser emitted by the laser and passes through the area to be measured, and then the collection module collects the detection signal D(x) in the time domain, where x represents the sampling sequence.

[0008] Step 2. The signal processing unit calibrates the wavelength of the detection signal D(x) to obtain the detection signal D(v) in the frequency domain, where v represents the signal frequency.

[0009] Step 3. After pre-decomposition processing of the detection signal D(v) by using the VMD algorithm, M sub-sequences are obtained;

[0010] Step 4. Sample entropy values of the M sub-sequences are calculated respectively, and the sample entropy value corresponding to the turning point tending to be stable in the broken line connected by the M sample entropy values is taken as the decomposition layer number of the VMD algorithm; 1<K<M;

[0011] Step 5. According to the decomposition layer number K, the detection signal D(v) is decomposed again to obtain K sub-sequences;

[0012] Step 6. Correlation coefficients between the K sub-sequences and the detection signal D(v) are calculated respectively, and according to the correlation coefficient of each sub-sequence, the K sub-sequences are divided into effective signal sub-sequences and noise signal sub-sequences;

[0013] Step 7. After filtering and smoothing processing of each effective signal sub-sequence, the effective signal sub-sequence is reconstructed to obtain a reconstructed signal;

[0014] Supposing that the window width of the filter is 2h+1, after sampling any effective signal sub-sequence according to the window width, 2h+1 data points are obtained, and then a W-1 order polynomial is used to fit the 2h+1 data points in the window to complete the filtering and smoothing processing of the sub-sequence in the window, so that the filtering and smoothing processing of the corresponding effective signal sub-sequence is completed by continuously moving the window, and finally the filtered and smoothed sub-sequence is reconstructed to obtain a reconstructed signal;

[0015] Step 8. The least square method shown in formula (1) is used to fit the reconstructed signal to obtain a background light intensity signal R(v) of the reconstructed signal;

[0016] R(v)=a0+a1v+a2v 2 +a3v 3 (1)

[0017] In formula (1), a0, a1, a2 and a3 represent four fitting parameters; v represents a signal frequency;

[0018] Step 9. The ratio of is calculated, and the natural logarithm of the ratio is calculated to obtain an integral absorption curve B(v) of the detection laser;

[0019] Step 10. The Voigt fitting is performed on the integral absorption curve B(v), and the Beer-Lambert law formula is substituted to perform inversion, and finally the gas concentration after eliminating the open light path detection noise is obtained.

[0020] The method for reducing the influence of open light path spectral detection signal noise provided by the application is also characterized in that the step 3 comprises: ​

[0021] Step 3.1, calculate the mth modal component u m the analytic signal δ m of (N) :

[0022]

[0023] In formula (1), m represents the decomposition layer number, N represents the sampling point number of the signal, ψ(·) represents the Dirac function, J is the imaginary unit, u m (N) represents the mth modal component;

[0024] Step 3.2, obtain the corresponding spectrum δ m of the mth modal component u

[0025]

[0026] In formula (2), ω is the center frequency index of the mth modal component u m (N), and J is the imaginary unit;

[0027] Step 3.3, introduce a quadratic penalty factor α and a Lagrange multiplier λ, and construct a constraint formula by using formula (3):

[0028]

[0029] In formula (3), f(N) represents the sampling signal, u m represents the mth modal component, ω m represents the center frequency of the mth modal component, represents the derivative of the function with respect to time;

[0030] Step 3.4, define the current iteration number as n, and initialize n = 1, and initialize the mth modal component of the n th iteration as the center frequency of the mth modal component of the n th iteration the Lagrange multiplier λ of the n th iteration n ;

[0031] Step 3.5, obtain the spectrum of the mth modal component in the n+1 th iteration according to formula (4)

[0032]

[0033] In formula (4), in formula (4), and are the Fourier transforms of f(N), λ n , respectively, ω m n represents the center frequency of the mth modal component in the nth iteration;

[0034] Step 3.6, determining the center frequency of the mth modal component in the n+1th iteration according to formula (5)

[0035]

[0036] Step 3.7, updating the Lagrange multiplier of the nth iteration according to formula (6) obtaining the Lagrange multiplier of the n+1th iteration

[0037]

[0038] In formula (6), τ represents noise tolerance;

[0039] Step 3.8, when the difference of the modal components of two adjacent iterations is less than a set threshold σ, that is, formula (7) is established, the loop is ended, and the eigenmodal function component with the fixed mth center frequency is obtained, otherwise, n+1 is assigned to n, and the step 3.5 is returned to be sequentially executed:

[0040]

[0041] In formula (7), σ represents the set threshold.

[0042] The electronic device comprises a memory and a processor, and the memory is used for storing a program supporting the processor to execute the method for reducing the influence of the noise of the open optical path spectrum detection signal.

[0043] The computer readable storage medium stores a computer program, and the computer program is executed by the processor to execute the steps of the method for reducing the influence of the noise of the open optical path spectrum detection signal.

[0044] Compared with the prior art, the beneficial effects of the present application are reflected in:

[0045] 1. In the present application, without increasing the hardware structure, the flicker noise problem caused by the detection optical path length and the external environmental interference is solved through the variational modal decomposition, filtering and smoothing and reconstruction processing, which is simple, fast and has strong applicability.

[0046] 2. The method for reducing the influence of the noise of the open optical path spectrum detection signal can not only be used in the spectrum signal of a small-range detection system, but also be directly applied to the detection signal of a large-scale open area.

[0047] 3. The noise reduction method proposed in this invention utilizes VMD to decompose the detection signal, effectively distinguishing between the noise part and the effective signal part. By directly discarding the noise part, the noise reduction effect of the detection signal is greatly improved, thus effectively improving the detection accuracy. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of a gas detection system for an open space in an embodiment of the present invention; Figure 2 A flowchart of a method for reducing the influence of noise in open optical path spectral detection signals according to the present invention;

[0049] Figure 3 This is a comparison diagram of the two methods for removing background noise in an embodiment of the present invention;

[0050] Figure 4 This is the Voigt fitting plot in an embodiment of the present invention;

[0051] The following components are labeled in the diagram: 1. Signal generator, 2. Laser controller, 3. Laser, 4. Fiber beam splitter, 5. Collimator, 6. Telescope, 7. Data acquisition module, 8. Reference optical path beam collimator, 9. Standard gas chamber, 10. Photodetector, 11. Data processing module, 12. Computer. Detailed Implementation

[0052] In this embodiment, as Figure 1 As shown, a method for reducing the influence of noise on open optical path spectral detection signals is applied to a detection system consisting of a laser 3, an optical fiber beam splitter 4, a laser controller 2, a signal generator 1, a collimator 5, a telescope 6, a data acquisition module 7, and a data processing module 11 connected in sequence.

[0053] In this configuration, laser controller 2 modulates the output wavelength of laser 3 based on the input signal from signal generator 1. Signal generator 1 scans the selected absorption line;

[0054] The light emitted by the laser 3 of the beam splitter 4 is split into a reference optical path and a probe optical path.

[0055] The reference optical path beam collimator 8 collimates the light and sends it to the standard gas chamber 9, where it is focused onto the photodetector 10. The probe optical path beam collimator 5 collimates the light and sends it through the transceiver telescope 6, through the area to be measured, where it is absorbed by the gas and captured by the photodetector 7. The signal is then converted into an electrical signal and input into the data processing module 11, and displayed by the computer 12.

[0056] See Figure 2 The method for reducing the impact of noise on open optical path spectral detection signals is carried out according to the following steps:

[0057] Step 1. The beam splitter splits the detection laser emitted by the laser, and after passing through the to-be-detected area, the acquisition module acquires the detection signal D(x) in the time domain; x represents the sampling sequence;

[0058] Step 2. The signal processing unit calibrates the detection signal D(x) in the wavelength domain to obtain the detection signal D(v) in the frequency domain;

[0059] Step 3. After the VMD algorithm is used to pre-decompose the detection signal D(v), M sub-sequences are obtained;

[0060] Specifically, the VMD method decomposes the detection signal into a series of IMF modal components, and the implementation of the variational modal decomposition can be divided into two steps, constructing a variational problem and solving the variational problem; in constructing the variational problem, first, based on the Hilbert transform, the unilateral spectrum of each sub-signal is obtained, then the exponential term aliasing of the center frequency ω k of each sub-signal is removed, the spectrum of the sub-sequence is modulated to the base frequency band, and finally the bandwidth of the demodulated signal is estimated by using Gaussian smoothing, and a variational problem with band constraint is finally solved. The specific steps are as follows:

[0061] Step 3.1, the Hilbert transform is used to calculate the analytic signal of the mth modal component u m (N) as shown in formula (1):

[0062]

[0063] In formula (1), m represents the number of decomposition layers, ψ(·) represents the Dirac function, J is the imaginary unit, u m (N) represents each modal component, δ m '(N) represents the analytic signal of the mth modal component.

[0064] Step 3.2, the spectrum of each modal component is modulated to the corresponding base frequency band by using formula (2), and the sum of the frequency bandwidths of each modal component is the frequency width of the original signal:

[0065]

[0066] In formula (2), ω is the center frequency index of the mth modal component u m (N), δ m (N) represents the corresponding spectrum of the mth modal component, N represents the number of sampling points of the signal, and J is the imaginary unit.

[0067] Step 3.3, a quadratic penalty factor α and a Lagrange multiplier λ are introduced, and a constraint formula is constructed by using formula (3):

[0068]

[0069] In formula (3), f(N) represents the sampling signal D(v), u m represents the mth modal component, ω m represents the center frequency of the mth modal component, represents the derivative of a function with respect to time.

[0070] Step 3.4, define the current iteration number as n, and initialize n = 1, and initialize the mth modal component of the nth iteration as the center frequency of the mth modal component of the nth iteration the Lagrange multiplier λ of the nth iteration n ;

[0071] Step 3.5, according to formula (4), the frequency spectrum of the mth modal component under the (n+1)th iteration is obtained

[0072]

[0073] In formula (4), and are the Fourier transforms of f(N), λ n , respectively, represents the frequency center of the mth modal component under the nth iteration, and α represents a quadratic penalty factor to ensure the reconstruction accuracy of the original signal.

[0074] Step 3.6, according to formula (5), the center frequency of the mth modal component under the (n+1)th iteration is determined

[0075]

[0076] Step 3.7, according to formula (6), the Lagrange multiplier of the nth iteration is updated to obtain the Lagrange multiplier of the (n+1)th iteration

[0077]

[0078] In formula (6), τ represents the noise tolerance.

[0079] Step 3.8, when the difference between the modal components of two adjacent iterations is less than a set threshold σ, that is, formula (7) is established, the loop is ended, and the m modal components with fixed center frequencies are obtained, otherwise, n+1 is assigned to n, and then the step 3.5 is returned to be executed in sequence:

[0080]

[0081] In formula (7), σ represents a set threshold value.

[0082] Step 4. The sample entropy values of the M sub-sequences are respectively calculated according to the sample entropy value definition, and a sample entropy value corresponding to a turning point tending to be stable in a broken line connected by the M sample entropy values is taken as the decomposition layer number of the VMD algorithm; 1 < K < M;

[0083] In this embodiment, the specific calculation method of the sample entropy value is as follows:

[0084] Step 4.1. The first-order modal component Imf1 is sequentially composed into a g-dimensional vector according to formula (8), and the distance between different vectors in the g-dimensional vector set is calculated.

[0085]

[0086] In formula (8), N represents the number of sampling points of the signal, c g (N-g+1) represents the N-g+1th vector, d i , and d j respectively represent the ith and jth vectors, 1 < i < N-g+1, 1 < j < N-g+1, d i,j is the absolute value of the maximum difference in the contained elements of the two vectors, l = 1, 2, …, N-g+1.

[0087] Step 4.2. A threshold value v is given, and v > 0, and formula (9) is used to calculate d i,j (v) < v, n i,j (v) is the ratio of the total number of vectors to the total number of vectors, and the average value E g (v) of the ratio is:

[0088]

[0089] In formula (9), n i,j (v) represents the number of d i,j (v) < v, N-g+1 is the total number of vectors, n i,j (v) is the ratio of the total number of vectors to the total number of vectors, E g (v) represents the average value of the ratio.

[0090] Step 4.3. The dimension is increased to g+1, the distance between different vectors in the g-dimensional vector set is calculated, and the ratio and average value and the sample entropy value are recalculated by using formula (10):

[0091]

[0092] In formula (10), When the dimension is g+1, d i,jThe ratio of the number of vectors less than a given threshold v to the number of vectors, E g +1 (v) represents the d i,j The average of the ratio of the number of vectors less than a given threshold v to the number of vectors, SampEn(g, v, N) represents the value of the sample entropy when the given threshold is v, the dimension is g, and the sampling point is N.

[0093] Step 5. According to the number of decomposition layers K, the detection signal D(v) is re-decomposed to obtain K sub-sequences;

[0094] Step 6. The correlation coefficients of the K sub-sequences and the detection signal D(v) are calculated respectively, and the K sub-sequences are divided into sub-sequences of effective signals and sub-sequences of noise signals according to the correlation coefficients of each sub-sequence;

[0095] Step 7. Each sub-sequence of the effective signal is filtered and smoothed, and then reconstructed to obtain a reconstructed signal;

[0096] Suppose the window width of the filter is 2h+1, and after sampling any sub-sequence of the effective signal according to the window width, 2h+1 data points are obtained, and then a W-1 order polynomial is used to fit the 2h+1 data points in the window to complete the filtering and smoothing of the sub-sequence in the window, so as to complete the filtering and smoothing of the corresponding sub-sequence of the effective signal by continuously moving the window. Finally, the filtered and smoothed sub-sequence is reconstructed to obtain a reconstructed signal;

[0097] Step 8. The least square method shown in formula (11) is used to fit the reconstructed signal to obtain the background light intensity signal R(v) of the reconstructed signal;

[0098] R(v) = a0 + a1v + a2v 2 +a3v 3 (11)

[0099] In formula (11), a0, a1, a2, and a3 represent four fitting parameters; v represents the signal frequency;

[0100] Step 9. The ratio of is calculated, and the natural logarithm of the ratio is calculated to obtain the integral absorption curve B(v) of the detection laser;

[0101] Step 10. The Voigt fitting is performed on the integral absorption curve B(v), and the Beer-Lambert law formula is substituted to perform inversion, and finally the gas concentration after eliminating the open light path detection noise is obtained.

[0102] In order to verify the effect of the method, the experiment is carried out by using the established open space gas detection system, the signals after background noise removal are obtained by using the traditional method and the method, and the result graph is as shown in Figure 3 It can be seen that the signal quality after removing the background noise is obviously improved, and the flicker noise is obviously suppressed. The results are compared with the traditional method. Figure 3 The results are synchronously subjected to Voigt fitting, and the signal is as shown in Figure 4 The residual sum of squares (ROD) is 0.06132 and 0.04939 respectively after the adaptive iterative fitting of the traditional method, and the fitting degree is improved by 24%; the Reduced Chi-Sqr (RSS) of the absorption Voigt fitting curve is 1.54847E-4 and 1.24722E-4 respectively after the treatment of the two methods, and the fitting effect is improved by 19%. The experimental results prove that the method can effectively reduce the flicker noise caused by the long optical path and the inherent noise of the instrument during the open space detection, and can effectively improve the accuracy of the gas concentration measurement.

Claims

1. A method for reducing the influence of noise on open optical path spectral detection signals, characterized in that, It is applied to a detection system consisting of a laser, beam splitter, laser controller, signal generator, transceiver telescope, corner reflector, signal processing module, and acquisition module connected in sequence, and is performed according to the following steps: Step 1. The beam splitter splits the detection laser emitted by the laser into multiple beams, which then pass through the area to be tested and are collected by the acquisition module to obtain the detection signal in the time domain. ; Indicates the sampling sequence; Step 2. The signal processing module processes the detected signal. Wavelength calibration is performed to obtain the detection signal in the frequency domain. ; Indicates signal frequency; Step 3. Use the VMD algorithm to analyze the detected signal. After pre-decomposition, M subsequences are obtained; Step 4. Calculate the sample entropy values ​​of the M subsequences respectively, and take the sample entropy value index K corresponding to the stable inflection point in the broken line formed by connecting the M sample entropy values ​​as the decomposition level K of the VMD algorithm; 1 <K<M; Step 5. Based on the number of decomposition layers K, re-analyze the detection signal. The sequence is decomposed to obtain K subsequences; Step 6. Calculate the K subsequences and the detection signal respectively. The correlation coefficient between them is used to divide the K subsequences into subsequences of effective signal and subsequences of noise signal based on the correlation coefficient of each subsequence; Step 7. Filter and smooth the subsequence of each valid signal before reconstructing it to obtain the reconstructed signal; Let the window width of the filter be 2h+1. After sampling any subsequence of an effective signal according to the window width, 2h+1 data points are obtained. Then, a polynomial of degree W-1 is used to fit the 2h+1 data points within the window to complete the filtering and smoothing of the subsequence within the window. Thus, by continuously shifting the window, the filtering and smoothing of the corresponding subsequence of the effective signal is completed. Finally, the filtered and smoothed subsequence is reconstructed to obtain the reconstructed signal. Step 8. Fit the reconstructed signal according to the least squares method shown in equation (11) to obtain the background light intensity signal of the reconstructed signal. ; (11) In equation (11), , , , This represents four fitting parameters; Indicates signal frequency; Step 9. Calculation The ratio of the two values ​​is then used to calculate the natural logarithm of the ratio, thus obtaining the integral absorption curve of the detected laser. ; Step 10. Analyze the integral absorption curve. After performing Voigt fitting and substituting into the Beer-Lambert law formula for inversion, the gas concentration after eliminating open optical path detection noise is finally obtained.

2. The method for reducing the influence of noise in open optical path spectral detection signals according to claim 1, characterized in that, Step 3 includes: Step 3.1: Calculate the m-th modal component using equation (1). Analyzed signal : (1) In equation (1), m represents the number of decomposition layers, and N represents the number of sampling points of the signal. This represents the Dirac function, where J is the imaginary unit. This represents the m-th modal component; Step 3.2: Use equation (2) to obtain the corresponding spectrum of the m-th modal component. : (2) In equation (2), For the m-th modal component The center frequency index, where J is the imaginary unit; Step 3.3: Introduce a secondary penalty factor and Lagrange multipliers And use equation (3) to construct the constraint formula: (3) In equation (3), Indicates the sampled signal. This represents the m-th modal component. This represents the center frequency of the m-th modal component. This indicates taking the time derivative of the function; Step 3.4: Define the current iteration number as n, and initialize n=1. Initialize the m-th modal component of the n-th iteration as... The center frequency of the m-th modal component in the nth iteration The Lagrange multipliers in the nth iteration ; Step 3.5: Obtain the spectrum of the m-th modal component in the (n+1)-th iteration according to equation (4). ; (4) In equation (4), , and They are respectively , , Fourier transform, This represents the frequency center of the m-th modal component in the n-th iteration; Step 3.6: Determine the center frequency of the m-th modal component in the (n+1)-th iteration according to equation (5). ; (5) Step 3.7: Update the Lagrange multipliers for the nth iteration according to equation (6). We obtain the Lagrange multipliers for the (n+1)th iteration. ; (6) In equation (6), Indicates noise tolerance; Step 3.8: When the difference between the modal components of two adjacent iterations is less than a set threshold... When equation (7) holds true, the loop ends and the m-th eigenmode function component with a fixed center frequency is obtained; otherwise, n+1 is assigned to n, and the process returns to step 3.5 for sequential execution. (7) In equation (7), This indicates the set threshold.

3. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store programs that support the processor in executing the method for reducing the influence of noise on open optical path spectral detection signals as described in claim 1 or 2, and the processor is configured to execute the programs stored in the memory.

4. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program, when run by the processor, performs the steps of the method for reducing the influence of noise on open optical path spectral detection signals as described in claim 1 or 2.

Images