Wavelet transform-based sub-view-field spatial heterodyne interferogram noise reduction method

Abstract

The invention provides a denoising method for a sub-view-field spatial heterodyne interferogram based on wavelet transform. Comprising the following steps: S1, firstly, establishing a complete spectrum forward modeling model to obtain radiation brightness of different concentrations of a measured target; and S2, taking the radiation brightness of the measured target with different concentrations as an input light source, and obtaining a line-by-line sub-view-field spatial heterodyne simulation interference image of the measured target with different concentrations. And S3, then, adding noise to the obtained sub-view-field spatial heterodyne simulation interference image to simulate a sub-view-field spatial heterodyne interference image in an experimental environment. And S4, performing line-by-line wavelet multilayer decomposition on the obtained noisy sub-view-field spatial heterodyne interferogram to obtain an approximation coefficient and a detail coefficient of each layer. S5, performing noise intensity estimation and threshold processing on the high-frequency detail coefficient; and performing baseline drift modeling and correction on the low-frequency approximation coefficient. And S6, finally, performing wavelet reconstruction on the high-frequency coefficient after line-by-line processing and the corrected low-frequency coefficient, and outputting a noise-reduced sub-view-field spatial heterodyne interference image. According to the method, line-by-line filtering processing is carried out on the sub-view-field space heterodyne interferogram, interferogram information of high-frequency random noise and low-frequency background noise is removed, and each line of interferogram information of different concentrations of a measured target is reserved to a great extent.

Description

(1) Technical field

[0001] The present invention relates to a denoising method for a spatial heterodyne interferogram with a sub-field of view based on wavelet transform, which can be used to efficiently filter out high-frequency random noise and low-frequency background noise from the spatial heterodyne interferogram with a sub-field of view and retain the spectral information of each row of the detection target to the greatest extent, belonging to the technical field of spatial heterodyne spectroscopy. (2) Background technology

[0002] Spatial heterodyne spectroscopy is a new type of spatial heterodyne ultra-high resolution spectral detection technology developed from the Michelson interferometer theory. Due to its advantages such as ultra-high spectral resolution, real-time performance, and high light flux, it has broad applications and development prospects in the fields of atmospheric monitoring, satellite remote sensing, chemical evolution, etc.

[0003] Spatial heterodyne field imaging technology is an improvement on traditional spatial heterodyne spectroscopy technology. By adding a cylindrical mirror to the spatial heterodyne front optical system, the scene within the total field of view is divided into multiple field of view slices. The interference information in each field of view slice is imaged onto several rows of the corresponding detector. Therefore, several rows on the detector surface correspond to the spectral information of a sub-field of view in the field of view.

[0004] Spatial extraheterodyne field imaging technology inverts different spectral information of the target material through each line of the interference pattern, and its accuracy is directly affected by the signal-to-noise ratio of the interference pattern. Noise reduction processing of spatial extraheterodyne spectra can be performed at any stage of data processing, which will make the characteristic peaks of the spectrum appear. However, most current research on this type of data noise reduction focuses on interferograms, and most of them utilize traditional image processing algorithms. Traditional global filtering mixes data of different concentration rows, causing the low-concentration row signals in the interference pattern to be submerged by high-concentration noise, and the concentration gradient information is lost in the averaging operation, reflecting its shortcomings such as weak targeting, poor processing effect and low efficiency. Therefore, it is necessary to find an algorithm suitable for spatial extraheterodyne field imaging interferograms.

[0005] The present invention provides a wavelet-based method for denoising spatial heterodyne interferograms with a sub-field of view. Wavelet decomposition and reconstruction are performed independently on each row of the interferogram (corresponding to a concentration), avoiding blurring of concentration characteristics caused by inter-row data mixing. Independent processing of each row prevents averaging of differences in noise statistical characteristics between concentrations, ensuring that threshold calculation and noise suppression are optimized for the concentration characteristics of the current row. The concentration characteristics of each row of the interferogram are fully preserved during the denoising process, laying the foundation for high-precision spectral inversion. (3) Summary of the invention

[0006] The present invention relates to a denoising method for spatial heterodyne interferograms based on wavelet transform, which can be used to efficiently filter out high-frequency random noise and low-frequency background noise from spatial heterodyne interferograms of the divided field of view, and to retain the spectral information of each row of the detection target to the greatest extent.

[0007] The object of the present invention is achieved like this:

[0008] The present invention relates to a method for reducing noise of a spatial external differential field of view interference pattern based on wavelet transform, comprising the following steps:

[0009] S1. First, a complete spectral forward model is established to obtain the radiation brightness of different concentrations of the measured target.

[0010] S2. Next, the radiation brightness of the target under test with different concentrations is used as the input light source to obtain the sub-viewing field spatial heterodyne simulation interference image of the target under test with different concentrations line by line.

[0011] S3. Then, noise is added to the obtained sub-viewing field spatial heterodyne simulation interference image to simulate the sub-viewing field spatial heterodyne interference pattern in the experimental environment.

[0012] S4. Furthermore, the obtained noisy sub-field spatial heterodyne interferogram is subjected to row-by-row wavelet multi-layer decomposition to obtain the approximate coefficients and detail coefficients of each layer.

[0013] S5. Then, the noise intensity of the high-frequency detail coefficients is estimated and threshold processing is performed; the baseline drift of the low-frequency approximation coefficients is modeled and corrected.

[0014] S6. Finally, the high-frequency coefficients processed row by row and the low-frequency coefficients corrected are reconstructed by wavelet, and a noise-reduced field-of-view spatial heterodyne interference image is output.

[0015] Furthermore, the specific process of S2 includes the following steps:

[0016] By setting reasonable fixed values ​​of atmospheric parameters of the radiation model, the concentration change is used to measure the spectral radiance of the target with different concentrations.

[0017] The radiant brightness of the target under test with different concentrations is used as the input light source for spatial heterodyne modulation. The light source is incident on the beam splitter through a collimating lens. The beam splitter splits the incident light source into two coherent light beams of equal intensity: one beam is reflected on the grating through the beam splitter, and is reflected back to the beam splitter after being diffracted by the grating; the other beam is incident on the grating after passing through the beam splitter, and is returned to the beam splitter after being diffracted by the grating; the two coherent light beams are emitted at different angles, thus generating spatial interference fringes on the output wavefront, and finally an interference pattern appears on the electronic imaging detector.

[0018] The interference fringes corresponding to the brightness of radiation with different concentrations are encoded line by line to generate a noise-free, field-of-view spatial heterodyne interference pattern with different concentration information in each line.

[0019] Furthermore, the specific steps of estimating the high-frequency noise intensity in S4 include:

[0020] Extract the first layer detail coefficients and sort them, removing the top 10% of the extreme values;

[0021] Assign exponentially decaying weights to the remaining coefficients.

[0022] Furthermore, the specific steps of baseline drift modeling in S4 include:

[0023] Perform LOESS smoothing on the highest layer approximation coefficients with a window width of 5%-20% of the signal length

[0024] Perform power spectrum analysis on the smoothed residuals to detect frequency components whose amplitude exceeds 3 times the standard deviation of the background noise;

[0025] Design an IIR notch filter for the detected interference frequency with a stopband width of ±(1-5Hz) (IV) Description of the accompanying drawings

[0026] Figure 1 Flowchart of the denoising method for spatial external differential field of view interference pattern based on wavelet transform (V) Specific implementation methods

[0027] The present invention will be further described below with reference to specific embodiments.

[0028] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is further described below with reference to the accompanying drawings. The specific process is as follows: Figure 1 shown.

[0029] S1. Establish a spectral forward model of the target to be measured and calculate the radiation brightness distribution of the target under different concentrations through the radiation transfer model;

[0030] Step S1 primarily utilizes the SCIATRAN radiative transfer model to simulate the radiance of different concentrations of atmospheric water vapor limb detection, which serves as the input light source for the spatial heterodyne spectrometer. Before performing the forward simulation, the calculation mode (radiance), trace gas (H2O), simulation band (1357-1363nm), and observation geometry (solar zenith angle 30°, relative azimuth angle 10°) must be set. The trace gas concentration is then varied to obtain the radiance at different concentrations.

[0031] S2. Using the different concentrations of radiation brightness obtained in S1 as an input light source, generating a noise-free simulated interference image of the measured target distributed row by row in a sub-field of view through an optical modulation process of a spatial heterodyne imaging spectrometer;

[0032] Furthermore, the specific process of step S2 includes the following steps:

[0033] The radiant brightness of the target under test with different concentrations is used as the input light source for spatial heterodyne modulation. The light source is incident on the beam splitter through a collimating lens. The beam splitter splits the incident light source into two coherent light beams of equal intensity: one beam is reflected on the grating through the beam splitter, and is reflected back to the beam splitter after being diffracted by the grating; the other beam is incident on the grating after passing through the beam splitter, and is returned to the beam splitter after being diffracted by the grating; the two coherent light beams are emitted at different angles, thus generating spatial interference fringes on the output wavefront, and finally an interference pattern appears on the electronic imaging detector.

[0034] By adding a cylindrical mirror to the spatial heterodyne front optical system, the scene within the total field of view is divided into multiple field of view slices. The interference information in each field of view slice is imaged onto several rows of the corresponding detector. Therefore, several rows on the detector surface correspond to the spectral information of a sub-field of view in the field of view.

[0035] Based on this principle, a simulation program was written for a spatial heterodyne spectrometer with a grating line density of 600 lines / mm and a grating width of 13.312 mm. The program detects the 1357-1363 nm band based on a Littow wavenumber of 1355 nm. The interference fringes corresponding to the radiation brightness of 1202 concentration values ​​were encoded line by line to obtain a noise-free, field-of-view spatial heterodyne interferogram with different concentration information for each row of the corresponding 1202×1202 pixels.

[0036] S3. Adding noise to the obtained sub-viewing field spatial heterodyne simulation interference image to simulate the sub-viewing field spatial heterodyne interference pattern in an experimental environment.

[0037] S4. Furthermore, the obtained noisy sub-field spatial heterodyne interferogram is subjected to row-by-row wavelet multi-layer decomposition to obtain the approximate coefficients and detail coefficients of each layer.

[0038] Furthermore, the specific process of S4 includes the following steps:

[0039] The interference pattern is decomposed row by row using the sym8 wavelet basis, and the frequency bands are divided as follows:

[0040] Layer 1 detail coefficient: [fs / 2, fs / 4] (suppresses high-frequency random noise)

[0041] Layer 3 detail coefficients: [fs / 8, fs / 16] (matching the fundamental frequency of the interference fringes)

[0042] Layer 6 detail coefficients: [0, fs / 128] (preserve density modulation information)

[0043] S5. Then, the noise intensity of the high-frequency detail coefficients is estimated and threshold processing is performed; the baseline drift of the low-frequency approximation coefficients is modeled and corrected.

[0044] Furthermore, the specific process of estimating the noise intensity of the high-frequency detail coefficient and performing threshold processing in S5 includes the following steps:

[0045] Extract the first layer detail coefficients and sort them, removing the top 10% of the extreme values;

[0046] Assign exponential decay weights to the remaining coefficients, and the weight formula is:

[0047] w i =e -5(i / n) ,i=1,2,...,n, where n is the number of coefficients after truncation;

[0048] The noise standard deviation is calculated as:

[0049]

[0050] Dynamically adjust the layer threshold according to the number of decomposition layers, the high-level threshold attenuation factor decreases linearly, and semi-soft threshold processing is performed on the detail coefficient of each layer;

[0051] The calculation formula of the hierarchical dynamic threshold is:

[0052]

[0053] k is the number of decomposition levels (k=1,2,...,L), N k is the length of the detail coefficient of the kth layer.

[0054] Furthermore, the specific process of performing baseline drift modeling and correction on the low-frequency approximation coefficient in step S5 includes the following steps:

[0055] The highest layer approximation coefficients are smoothed using LOESS, with a window width of 5%-20% of the signal length;

[0056] Perform power spectrum analysis on the smoothed residuals to detect frequency components whose amplitude exceeds 3 times the standard deviation of the background noise;

[0057] According to the detected interference frequency, an IIR notch filter is designed with a stopband width of ±(1-5Hz).

[0058] S6. Finally, the high-frequency coefficients processed row by row and the low-frequency coefficients corrected are reconstructed by wavelet, and a noise-reduced field-of-view spatial heterodyne interference image is output.

[0059] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the scope of protection of the present invention.

Claims

1. A method for denoising a spatial heterodyne interferogram of a sub-field of view based on wavelet transform, comprising the following steps: S1. First, a complete spectral forward model is established to obtain the radiation brightness of different concentrations of the measured target. S2. Next, the radiation brightness of the target under test with different concentrations is used as the input light source to obtain the sub-viewing field spatial heterodyne simulation interference image of the target under test with different concentrations line by line. S3. Then, noise is added to the obtained sub-viewing field spatial heterodyne simulation interference image to simulate the sub-viewing field spatial heterodyne interference pattern in the experimental environment. S4. Furthermore, the obtained noisy sub-field spatial heterodyne interferogram is subjected to row-by-row wavelet multi-layer decomposition to obtain the approximate coefficients and detail coefficients of each layer. S5. Then, the noise intensity of the high-frequency detail coefficients is estimated and threshold processing is performed; the low-frequency approximation coefficients are baseline modeled and corrected. S6. Finally, the high-frequency coefficients processed row by row and the low-frequency coefficients corrected are reconstructed by wavelet, and a noise-reduced field-of-view spatial heterodyne interference image is output.

2. The noise reduction method based on the spatial heterodyne interference pattern of the field of view according to claim 1, characterized in that: The process of generating a noise-free simulated interferogram includes: By setting reasonable fixed values ​​of atmospheric parameters of the radiation model, the spectral radiance of the target with different concentrations can be measured by using concentration changes; The radiance of the target under test at different concentrations is used as the input light source for spatial heterodyne modulation. The light source passes through a collimating lens and is incident on a beam splitter. The beam splitter splits the incident light source into two coherent beams of equal intensity: one beam is reflected by the beam splitter onto a grating, diffracted by the grating, and reflected back to the beam splitter; the other beam passes through the beam splitter and is incident on the grating, diffracted by the grating, and returned to the beam splitter. The two coherent lights are emitted at different angles, thus generating spatial interference fringes on the output wavefront, and finally an interference pattern appears on the electronic imaging detector. The interference fringes corresponding to the brightness of radiation with different concentrations are encoded line by line to generate a noise-free, field-of-view spatial heterodyne interference pattern with different concentration information in each line.

3. The noise reduction method based on the spatial heterodyne interference pattern of the field of view according to claim 1, characterized in that: By adding a cylindrical mirror to the spatial heterodyne front optical system, the scene within the total field of view is divided into multiple field of view slices. The interference information in each field of view slice is imaged onto several rows of the corresponding detector. Therefore, several rows on the detector surface correspond to the spectral information of a sub-field of view in the field of view.

4. The noise reduction method based on the spatial heterodyne interference pattern of the field of view according to claim 1, characterized in that: During the actual operation of the spatial heterodyne imaging spectrometer, the data obtained will be affected by factors such as the experimental environment and human operation, including high-frequency noise interference from dark current noise and readout noise and low-frequency background noise formed by stray light interference. Therefore, the noise injected in step S3 includes high-frequency random noise and low-frequency background noise, in order to simulate the spatial heterodyne interferogram of the divided field of view in the experimental environment to the greatest extent.

5. The noise reduction method based on the spatial heterodyne interference pattern of the field of view according to claim 1, characterized in that: The specific steps of estimating the high-frequency noise intensity in the step include: S51, extracting and sorting the first layer detail coefficients, and removing the extreme values ​​in the top 10% of the absolute values; S52. Assign exponential decay weights to the remaining coefficients. The weight formula is: w i =e -5(i / n) ,i=1,2,...,n; Where n is the number of coefficients after truncation; S53, noise standard deviation is calculated as:

6. The noise reduction method according to claim 1, wherein: The specific steps of baseline modeling described in the steps include: S61, perform LOESS smoothing on the highest layer approximation coefficients, with a window width of 5%-20% of the signal length; S62, performing power spectrum analysis on the smoothed residual to detect frequency components with amplitudes exceeding three standard deviations of the background noise; S63. Design an IIR notch filter for the detected interference frequency with a stopband width of ±(1-5Hz).

Images