Wavelength selection method and system for passive spectral calibration of on-orbit ultraviolet spectrometers

Abstract

The present disclosure relates to a wavelength selection method and system for passive spectral calibration of an ultraviolet spectral instrument in orbit, the method comprising: establishing a standard line list of preset wavelength bands, and preprocessing, performing single-line Gaussian fitting and local baseline modeling, using asymmetry and wavelength deviation for quality control, calculating the quality weight of each standard line, and calculating the similarity kernel of each standard line and other standard lines; based on the constraints of the wavelength range endpoints, the minimum distance constraints of two standard lines, and the constraints of abnormal standard line rejection, the standard lines in the set of standard lines after quality control are screened by using the determinant point process; further screening according to the joint evaluation of cross-day consistency and information amount; under the Bayesian framework, the standard lines are linearly regressed by a variable error model to obtain the finally selected standard lines. The present disclosure has high accuracy and high stability.

Description

Technical Field

[0001] This disclosure relates to the field of spectrometer technology, and in particular to a wavelength selection method and system for passive spectral calibration of an on-orbit ultraviolet spectrometer. Background Technology

[0002] The essence of spectral calibration is to establish a precise mapping between "pixel / step size and physical wavelength," ensuring that spectral line positions are traceable and quantifiable over long-term operation. On the ground, grating equations and discharge lamps such as mercury / neon / argon, cavity filters, or etalons provide a sparse yet reliable reference. However, during the instrument's operation in orbit, thermomechanical deformation, irradiation aging, gain and nonlinear drift, stray light, and contamination collectively alter the instrument's line spread function and effective dispersion, causing the ground mapping to quickly become invalid. Therefore, during the on-orbit phase, it is necessary to maintain the wavelength scale using external natural references (solar Fraunhofer standard), internal sources (mercury lamps / LEDs / laser diodes), and cross-calibration, and continuously correct drift through peak position estimation and low-order mappings (first-order or with weak second-order mapping).

[0003] In existing technologies, wavelength selection for on-orbit ultraviolet spectrometers largely relies on empirical line selection or single-line evaluation using a single index, resulting in low accuracy and stability. Therefore, there is a need to design a wavelength selection method and system for passive spectral calibration of on-orbit ultraviolet spectrometers. Summary of the Invention

[0004] Therefore, it is necessary to provide a wavelength selection method and system for passive spectral calibration of on-orbit ultraviolet spectrometers to address the above-mentioned problems.

[0005] To solve the above problems, the present disclosure adopts the following technical solution:

[0006] Firstly, a wavelength selection method for passive spectral calibration of on-orbit ultraviolet spectrometers is provided, comprising the following steps:

[0007] Establish a list of Fraunhofer standard lines for preset bands, wherein the preset bands belong to the ultraviolet band;

[0008] Preprocess the frames containing each Fraunhofer standard line in the Fraunhofer standard line list to obtain a clean one-dimensional spectrum and a per-pixel variance estimate.

[0009] Single-line Gaussian fitting and local baseline modeling are performed on the Fraunhofer standard line in a clean one-dimensional spectrum.

[0010] Quality control of Fraunhofer standard lines in a clean one-dimensional spectrum was performed using asymmetry and wavelength deviation, resulting in a set of quality-controlled Fraunhofer standard lines.

[0011] For the set of Fraunhofer standard lines after quality control, calculate the quality weight of each Fraunhofer standard line and calculate the similarity kernel between each Fraunhofer standard line and other Fraunhofer standard lines.

[0012] Based on the quality weight and the similarity kernel, and based on the constraints of wavelength range endpoints, the minimum spacing between two Fraunhofer standard lines, and the constraint of removing abnormal Fraunhofer standard lines, the Fraunhofer standard lines in the set of quality-controlled Fraunhofer standard lines are screened using a determinant point process to obtain the optimal subset of the set of quality-controlled Fraunhofer standard lines.

[0013] Based on the joint evaluation of cross-day consistency and information content, the Fraunhofer standard line is selected from the optimal subset to balance cross-day consistency and maximize information content.

[0014] Within the Bayesian framework, the Fraunhofer standard line, which balances cross-day consistency and maximizes information content, is calibrated by linear regression using a variable error model, resulting in the final selected Fraunhofer standard line.

[0015] In a preferred embodiment, the spectral mapping range is determined based on the finally selected Fraunhofer standard line, thereby determining the prediction range of the on-orbit ultraviolet spectrometer.

[0016] In a preferred embodiment, the finally selected Fraunhofer standard line is independently verified using mercury lamp spectroscopy.

[0017] In a preferred embodiment, the model for performing single-line Gaussian fitting and local baseline modeling on the Fraunhofer standard line in a clean one-dimensional spectrum is as follows:

[0018]

[0019] in, Indicates the wavelength of the reference standard spectral line. This represents the intercept of the coarse mapping. Indicates the slope of the coarse mapping. Indicates the step size. Indicates the step position One-dimensional spectral intensity at that location, The constant term representing the local baseline, The linear term representing the local baseline, For spectral line depth, Indicates the center wavelength. Indicates the equivalent linewidth.

[0020] In a preferred embodiment, the quality control of the Fraunhofer standard lines in a clean one-dimensional spectrum using asymmetry and wavelength deviation, to obtain a set of quality-controlled Fraunhofer standard lines, includes:

[0021] Perform baseline subtraction;

[0022] The wavelength is less than the center wavelength, calculated based on the spectral irradiance distribution curve of the Fraunhofer standard. Integral area on the side and wavelength greater than center wavelength Integral area on the side Define asymmetry , Calculate asymmetry ;

[0023] Preserve while satisfying asymmetry Meets the preset range The Fraunhofer standard line with a signal-to-noise ratio reaching the lower limit is used to obtain the set of quality-controlled Fraunhofer standard lines. This represents the deviation of the Fraunhofer standard line from the coarse mapping estimate. This represents the maximum estimated deviation of the single Fraunhofer standard line.

[0024] In a preferred embodiment, the formula for calculating the quality weight is:

[0025]

[0026] in, This indicates the number of the Fraunhofer standard line that is currently retained. Indicates the first Scalar mass weights of the Fraunhofer standard lines Represents an exponential function. This represents the weighting hyperparameter, used to control the severity of the penalty for single-line deviation. This represents the weighting hyperparameter, used to control the degree of preference for the signal-to-noise ratio. Indicates the first The normalized result of the single-line deviation of the coarse mapping estimate of the Fraunhofer standard line. Indicates the first Normalized signal-to-noise ratio results for the Fraunhofer standard lines;

[0027] The formula for calculating the similarity kernel is:

[0028]

[0029] in, This indicates the number of the Fraunhofer standard line that is currently retained. Indicates the first The location of the Fraunhofer standard line, Indicates the first The location of the Fraunhofer standard line, Indicates the first The location of the Fraunhofer standard line and the first The similarity kernel between the positions of the Fraunhofer standard lines. Indicates the relevant scale.

[0030] In a preferred embodiment, the optimal subset is fixed as follows: , .

[0031] In a preferred embodiment, the step of selecting the Fraunhofer criterion line from the optimal subset that balances cross-day consistency and maximizes information content, based on a joint evaluation of cross-day consistency and information content, includes:

[0032] For the same Fraunhofer standard line observed over multiple days, the cross-day variance of the central estimate is summarized in the optimal subset of ultraviolet spectra. Deviation from average wavelength ;

[0033] The Fraunhofer standard line, which balances cross-day consistency and maximizes information content, is selected from the optimal subset by minimizing the objective function. The objective function is:

[0034]

[0035] in, Represents the objective function value. This represents the weighted Fisher information matrix. Indicates the first The Fraunhofer standard line belongs to the optimal subset. , This represents the stability penalty coefficient, used to adjust the weight of cross-day variance in the quality weighting. Indicates the first The estimated center wavelength of the Fraunhofer standard line. This indicates the strength of the control variance penalty. This indicates the intensity of the penalty for controlling the average deviation. Indicates the first The average wavelength deviation of the Fraunhofer standard lines.

[0036] In a preferred embodiment, the linear regression calibration includes: setting the noise level and wavelength to conform to a linear relationship; during the linear regression calibration process, if a systematic deviation is found at the endpoint of the preset band of the Fraunhofer standard line, the Fraunhofer standard line is judged by the information criterion to determine whether to retain it, and all retained Fraunhofer standard lines are used as the final selected Fraunhofer standard lines.

[0037] Secondly, a wavelength selection system for passive spectral calibration of on-orbit ultraviolet spectrometers is provided, including:

[0038] The list creation module is used to create a list of Fraunhofer standard lines for a preset band, wherein the preset band belongs to the ultraviolet band;

[0039] The preprocessing module is used to preprocess the frames containing each Fraunhofer standard line in the Fraunhofer standard line list to obtain a clean one-dimensional spectrum.

[0040] The modeling module is used to perform single-line Gaussian fitting and local baseline modeling on the Fraunhofer standard line in a clean one-dimensional spectrum.

[0041] The quality control module is used to perform quality control on Fraunhofer standard lines in a clean one-dimensional spectrum by utilizing asymmetry and wavelength deviation, and obtains a set of quality-controlled Fraunhofer standard lines.

[0042] The calculation module is used to calculate the quality weight of each Fraunhofer standard line in the set of quality-controlled Fraunhofer standard lines, and to calculate the similarity kernel of each Fraunhofer standard line with other Fraunhofer standard lines.

[0043] The first screening module is used to screen the Fraunhofer standard lines in the set of quality-controlled Fraunhofer standard lines based on the quality weight and the similarity kernel, and based on the constraints of wavelength range endpoints, the minimum spacing between two Fraunhofer standard lines, and the constraint of removing abnormal Fraunhofer standard lines, using a determinant point process to obtain the optimal subset of the set of quality-controlled Fraunhofer standard lines.

[0044] The second screening module is used to jointly evaluate the Fraunhofer standard line based on cross-day consistency and information content, and select the Fraunhofer standard line from the optimal subset that takes into account both cross-day consistency and information content maximization.

[0045] The linear regression calibration module is used to perform linear regression calibration on the Fraunhofer standard line, which balances cross-day consistency and maximizes information content, within a Bayesian framework, through a variable error model, to obtain the final selected Fraunhofer standard line.

[0046] The wavelength selection method and system for passive spectral calibration of the above-mentioned on-orbit ultraviolet spectrometer takes into account asymmetry, wavelength deviation, mass weight, similarity kernel, cross-sky consistency and information content, constraints of wavelength range endpoints, constraints of minimum spacing between two Fraunhofer standard lines, and constraints of removing abnormal Fraunhofer standard lines. It has carried out multi-angle screening, and the line selection accuracy and stability are high. Attached Figure Description

[0047] Figure 1 This is an example diagram of the internal optical path of an on-orbit ultraviolet spectrometer;

[0048] Figure 2 This is a flowchart illustrating a method in one embodiment of the present disclosure;

[0049] Figure 3 This is a schematic diagram illustrating the calculation of spectral irradiance symmetry in one embodiment of this disclosure. Detailed Implementation

[0050] The technical solutions of this disclosure will now be described in detail with reference to the accompanying drawings and preferred embodiments.

[0051] Figure 1 This is an example diagram of the internal optical path of an on-orbit ultraviolet spectrometer: The target light is shaped by a lens group and then enters a dichroic filter for band separation; subsequently, it is refracted by a concave mirror and coupled to a holographic concave grating for dispersion and focusing. The light finally converges to a focal plane detector (not shown); the instrument is equipped with diffuser plates for uniform surface source injection during on-orbit / ground self-testing and calibration. Red light is used for tracing the main optical path, demonstrating the imaging and dispersion relationships between the various components.

[0052] See Figure 2 This disclosure provides a wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer, including:

[0053] Establish a list of Fraunhofer standard lines for preset bands, wherein the preset bands belong to the ultraviolet band;

[0054] Preprocess the frames containing each Fraunhofer standard line in the Fraunhofer standard line list to obtain a clean one-dimensional spectrum;

[0055] Single-line Gaussian fitting and local baseline modeling are performed on the Fraunhofer standard line in a clean one-dimensional spectrum.

[0056] Quality control of Fraunhofer standard lines in a clean one-dimensional spectrum was performed using asymmetry and wavelength deviation, resulting in a set of quality-controlled Fraunhofer standard lines.

[0057] For the set of Fraunhofer standard lines after quality control, calculate the quality weight of each Fraunhofer standard line and calculate the similarity kernel between each Fraunhofer standard line and other Fraunhofer standard lines.

[0058] Based on the quality weight and the similarity kernel, and based on the constraints of wavelength range endpoints, the minimum spacing between two Fraunhofer standard lines, and the constraint of removing abnormal Fraunhofer standard lines, the Fraunhofer standard lines in the set of quality-controlled Fraunhofer standard lines are screened using a determinant point process to obtain the optimal subset of the set of quality-controlled Fraunhofer standard lines.

[0059] Based on the joint evaluation of cross-day consistency and information content, the Fraunhofer standard line is selected from the optimal subset to balance cross-day consistency and maximize information content.

[0060] Within the Bayesian framework, the Fraunhofer standard line, which balances cross-day consistency and maximizes information content, is calibrated by linear regression using a variable error model, resulting in the final selected Fraunhofer standard line.

[0061] Preferably, the method further includes:

[0062] The final selected Fraunhofer standard line was independently verified using mercury lamp spectroscopy.

[0063] The wavelength selection method is described in detail below. In this embodiment, the method is specifically described for the 165–320 nm ultraviolet band, which is the most challenging area in the prior art. The method includes the following steps:

[0064] Step 1: Acquire spectral data from the on-orbit ultraviolet spectrometer and establish a list of Fraunhofer standard lines in the 165-320 nm range. The uniform medium is not limited to full vacuum or standard air. Then, optimize this list by integrating and filtering data based on NIST (National Institute of Standards and Technology) Fraunhofer standard lines and TSIS (Total and Spectral Solar Irradiance Sensor) mixed solar spectral reference data. The final list obtained in Step 1 serves only as a "candidate truth library" and "windowing initialization guide" (i.e., the spectrometer's windowing initialization operation) and is not used for fitting.

[0065] Step 1.1: Given the inconsistent spectral resolution of the on-orbit ultraviolet spectrometer, variational convolution (establishing consistency) is used on the TSIS spectral data. In other words, variational convolution is applied to the TSIS (Total Spectral Irradiance Sensor) spectral data to address the inconsistency in spectral resolution caused by environmental factors (such as temperature changes and mechanical vibrations), thus achieving consistency in the spectral data.

[0066] Step 1.2: Use the Fraunhofer standard line in the solar spectrum to calibrate the on-orbit ultraviolet spectrometer and determine the applicable bandwidth of the on-orbit spectrometer; use an off-orbit ultraviolet spectrometer to perform spectral calibration and correction on the on-orbit ultraviolet spectrometer.

[0067] Step 1.3: Select several Fraunhofer standard lines to form a preliminary Fraunhofer standard line list;

[0068] Step 1.4: Integrate and filter the Fraunhofer standard line list to obtain a new Fraunhofer standard line list. Specifically, based on the NIST Fraunhofer standard lines, TSIS, and the solar radiation spectrum (a standardized dataset used to describe the distribution of solar energy with wavelength, i.e., the solar spectral reference data mentioned above), the results of Step 1.3 are integrated and filtered to obtain a new Fraunhofer standard line list. Obviously, the new Fraunhofer standard line list will be used in subsequent steps.

[0069] Step 2: Preprocess the frame containing each Fraunhofer standard line in the Fraunhofer standard line list to obtain a clean one-dimensional spectrum. .

[0070] Preprocessing includes: dark field / background subtraction, nonlinear correction, bad pixel removal and saturation removal, and spike removal (MAD thresholding). Each Fraunhofer standard line in the list is preprocessed to output a clean one-dimensional spectrum. , This indicates the step index.

[0071] Furthermore, this also includes the cleaned one-dimensional spectrum. The above step involves estimating the measurement variance for each step size, i.e., calculating the per-pixel variance estimate. This variance can then be used as a weight in Gaussian fitting and Bayesian estimation to achieve weighted fitting and improve robustness. The per-pixel variance estimate is calculated based on a noise model or repeated measurement results.

[0072] Step 3, for the cleaned one-dimensional spectrum Single-line Gaussian fitting and local baseline modeling were performed on the Fraunhofer standard line.

[0073] Around the candidate line (clean one-dimensional spectrum) Initial step size of the Fraunhofer standard line (in the image) A window is opened to simultaneously estimate the spectral line shape and the local continuous spectrum (baseline) within a narrow window. The model for single-line Gaussian fitting and local baseline modeling is as follows:

[0074]

[0075] in, This represents the initial step size. Indicates the wavelength (standard true value) of the reference standard spectral line. This represents the intercept of the coarse mapping. Indicates the slope of the coarse mapping. Indicates the step position One-dimensional spectral intensity at that location, Indicates the step size. Describe the local baseline, specifically. The constant term representing the local baseline, The linear term representing the local baseline, The spectral line depth (amplitude). , The center wavelength, The equivalent linewidth is the effective width obtained by convolving the inherent linewidth of the light source with the instrument response function (ILS / SRF). Furthermore, to improve robustness, the model's parameter estimation employs a weighted least squares method combined with Huber loss, with observation weights set according to the aforementioned measurement variance. To avoid incomplete understanding, a priori constraints are set on the equivalent linewidth, with the constraints being...

[0076]

[0077] and restrictions It must not deviate too far from the center of the window.

[0078] in, This represents the equivalent standard deviation calculated from the instrument's linearity function, corresponding to the baseline value of the spectral linewidth. The half-width at half-maximum (WHM) of the line spread function of an instrument reflects its intrinsic resolution.

[0079] Model After fitting, the parameter covariance matrix is ​​obtained, and then the standard deviation of the center position is given. and spectral line depth Along with the signal-to-noise ratio (SNR), these quantities are used for subsequent quality assessment and weighting. Indicates center wavelength standard deviation The parameter covariance matrix represents the relationship between... The corresponding variance term.

[0080] In some embodiments, if necessary, it is necessary to go back to the preprocessing in step 2 for correction.

[0081] Step 4: Utilize asymmetry and wavelength deviation to clean the one-dimensional spectrum. The Fraunhofer standard lines were subjected to double-threshold quality control to obtain a set of quality-controlled Fraunhofer standard lines.

[0082] Before quality control, baseline subtraction is performed, which means estimating the local background curve and then subtracting it so that only the shape of the standard line itself is retained for fitting.

[0083] After baseline subtraction, the Fraunhofer standard line was calculated using the center wavelength symmetric integration method. , The area of ​​the left and right integrals within the interval is used to define asymmetry. , ,

[0084]

[0085] in, Indicates less than wavelength The integral area of ​​the spectral irradiance curve, i.e. The integral area of ​​the interval. Indicates greater than wavelength The integral area of ​​the spectral irradiance curve, i.e. The integral area of ​​the interval. express or , Indicates the center wavelength. express and The ratio is calculated by integrating the area of ​​the curve based on the distribution of spectral irradiance with wavelength along the Fraunhofer standard line, with the horizontal axis representing wavelength. The vertical axis represents the radiant power density (SSI) per unit wavelength. and They represent the first on the curve, respectively. The point and the first One point, and They represent the first The point and the first Irradiance per unit wavelength at a point, and They represent the first The point and the first The wavelength at each point This represents the spectral response function. See also: Figure 3 , Figure 3 S1 and S2 in the text correspond to respectively and .

[0086] A clean one-dimensional spectrum that meets certain symmetry requirements is needed. Only the Fraunhofer standard line can be used for entry into the pool. Here, the asymmetry must meet a preset range, i.e. Only then can it be included in the pool. Use coarse mapping. Estimating the deviation of a single Fraunhofer standard line ,set up threshold It needs to meet the following requirements. (Recommended) (Or a fixed nanometer value), while requiring the core signal-to-noise ratio (SNR) to reach the lower limit. All those that pass form a "good line pool," which constitutes the set of Fraunhofer standard lines after quality control. The wavelength representing the predicted value can be changed by altering the letter. This represents the single-line bias of the coarse mapping estimate. This represents the (maximum) threshold of the estimated single-line deviation, where a single line refers to a single Fraunhofer standard line.

[0087] Step 5: Calculate the quality weight of each Fraunhofer standard line in the set of quality-controlled Fraunhofer standard lines; calculate the similarity kernel between each Fraunhofer standard line and other Fraunhofer standard lines in the set of quality-controlled Fraunhofer standard lines, which can be understood as information relevance, and use Gaussian kernel to measure positional similarity.

[0088] Before proceeding with diversity-based selection, the "availability and stability" of each candidate line is compressed into scalar quality weights. Furthermore, a kernel function is used to characterize "similarity / redundancy" on the position (step size / wavelength) axis, laying an algebraic foundation for the subsequent DPP's goal of "both being good and distributed":

[0089]

[0090] in, This indicates the number of the Fraunhofer standard line that is currently retained. Indicates the first Scalar mass weights of the Fraunhofer standard lines Represents an exponential function, where the exponent is... and These are weighted hyperparameters (which can be set according to task preferences, such as emphasizing signal-to-noise ratio). Specifically, This represents the weighting hyperparameter, used to control the severity of the penalty for single-line deviation. This represents the weighting hyperparameter, used to control the degree of preference for the signal-to-noise ratio. The tilde indicates normalization to the same dimension, i.e., Indicates the first The normalized result of the single-line deviation of the coarse mapping estimate of the Fraunhofer standard line. Indicates the first Normalized signal-to-noise ratio results for the Fraunhofer standard lines. Considering signal-to-noise ratio, line depth, symmetry, fit quality, and cross-sky stability, a higher numerical value indicates a more desirable line for inclusion in the calibration set. Positional similarity is measured using a Gaussian kernel on the step / wavelength axis.

[0091]

[0092] in, and These all represent the numbers of the Fraunhofer standard lines that are currently retained. For the first The position (pixel or wavelength) of the Fraunhofer standard line. For the first The position (pixel or wavelength) of the Fraunhofer standard line. Indicates the first The location of the Fraunhofer standard line and the first The similarity kernel between the positions of the Fraunhofer standard lines. For relevant scales, when the two Fraunhofer standard lines are too close together hour, This indicates that the two Fraunhofer standard lines are "highly similar / redundant"; when the spacing is large enough, This indicates that the two Fraunhofer standard lines are "information complementary".

[0093] Step 6: Based on the quality weights and similarity kernels obtained in Step 5, and constrained by the wavelength range endpoints, the minimum spacing between two Fraunhofer standard lines, and the elimination of abnormal Fraunhofer standard lines, the Fraunhofer standard lines in the set of quality-controlled Fraunhofer standard lines are filtered using DPP (Determinantal Point Process) to obtain the optimal subset of the set of quality-controlled Fraunhofer standard lines.

[0094] Construct an L-kernel function to combine the quality weights and similarity kernel into a positive semi-definite kernel: .

[0095] in, This represents the core function used to handle the diagonal of a matrix. Embed single-point mass into the core. = This indicates location-dependent / redundant; location-dependent refers to the situation described above. Position redundancy, as mentioned above , Therefore, both "line quality" and "dispersion degree (slightly correlated)" are encoded simultaneously. Under the premise of satisfying engineering constraints, a fixed size is calculated. The optimal subset:

[0096]

[0097] in, This represents the optimal subset selected from the set of Fraunhofer standard lines after quality control. This represents a function that selects the optimal subset by maximizing the determinant. This represents the optimal subset of candidates. Represents the optimal subset of candidates Submatrices corresponding to rows and columns, The optimal subset is fixed as , This indicates the total number of Fraunhofer standard lines. This indicates that the total number of entries is 3 or 4.

[0098] In this disclosure, the engineering constraints are as follows: the selected Fraunhofer standard lines can cover the endpoints of the wavelength range, and the spacing between any two selected Fraunhofer standard lines is not less than the minimum spacing. The selected Fraunhofer standard lines contain no anomalous Fraunhofer standard lines. Understandably, the selected spectral line set must include lines near the start of the band (shortest wavelength), so the constraint covers the endpoints of the wavelength range; it must also include lines near the end of the band (longest wavelength). Anomalous Fraunhofer standard lines may have low signal-to-noise ratios, large fitting biases, or incomplete dark field subtraction, etc.

[0099] Given a positive semidefinite kernel, k-DPP applies to all sets. Upper-defined probability: The geometrical equivalent of its MAP solution is to select the set of lines with the largest "volume," balancing high quality with dissimilarity. The k-DPP MAP algorithm can be used, offering low complexity, stable implementation, and ease of integration with rule constraints (spacing thresholds). Represents a set The probability of being selected. This function represents the function for calculating the determinant of a matrix. It means proportional to.

[0100] If a greater emphasis is needed on "parameter estimation information content," Fisher information can be used. With coefficient Integration with Goals:

[0101]

[0102] in, express coefficient, The function that calculates the logarithm of the determinant of a matrix. This is the Fisher information matrix constructed from the selected lines using the calibration observation equations, thus allowing for an adjustable trade-off between "diversity" and "parameter accuracy." Among these... Represents the optimal subset Submatrices corresponding to rows and columns.

[0103] Step 7: Evaluate the Fraunhofer standard line based on cross-day consistency and information content, and select the Fraunhofer standard line that balances cross-day consistency and information content from the optimal subset.

[0104] Specifically, when observing ultraviolet spectra over multiple days, the cross-day variance of the central estimate is summarized for the same Fraunhofer standard line in the optimal subset of these days. Deviation from average wavelength Incorporate stability penalties into quality weights (e.g., in...) Add to Alternatively, minimize the stability penalty in the objective function, specifically:

[0105]

[0106] in, This represents the objective function value, used to comprehensively evaluate the effectiveness of selecting the Fraunhofer standard line. Represents the variance function across days. Indicates the center wavelength. Indicates the first The Fraunhofer standard line belongs to the optimal subset. , This represents the stability penalty coefficient, used to adjust the weight of cross-day variance in the quality weighting. Indicates the first The estimated center wavelength of the Fraunhofer standard line. Represents normalization , Right now , Indicates the first The average wavelength deviation of the Fraunhofer standard lines. This indicates the strength of the control variance penalty. Right now , This indicates the intensity of the penalty for controlling the average deviation. Additionally, among other things, , representing the weighted Fisher information matrix, , Indicates the first The weighting factors of the Fraunhofer standard line. Indicates the first The standard deviation of the Fraunhofer standard lines is selected in this way to maximize information content while choosing cross-day stable standard lines from the optimal subset. The function for calculating the determinant of the square matrix is... (The determinant of the submatrix) represents the optimal subset. The larger the "volume," the more dispersed and informative the selected Fraunhofer standard lines are. Optimizing the objective function value means minimizing... That is, equivalent to maximizing .

[0107] Step 8: Under the Bayesian framework, the Fraunhofer standard line obtained in Step 7 is calibrated by linear regression using the EIV model (variable error model) to obtain the final selected Fraunhofer standard line.

[0108] Specifically, during the linear regression calibration process, if a systematic deviation is found at the endpoint of the preset band of the Fraunhofer standard line, the information criterion is used to determine whether the Fraunhofer standard line should be retained. All retained Fraunhofer standard lines are used as the final selected Fraunhofer standard lines.

[0109] Linear regression calibration includes setting the noise level and wavelength to have a linear relationship.

[0110] Both the reference wavelength and the center wavelength of the on-orbit ultraviolet spectrometer are noisy, and the magnitude of the noise can vary with the line (heteroscedasticity). The noise magnitude and wavelength (reference wavelength and center wavelength of the on-orbit ultraviolet spectrometer) are assumed to have a linear relationship. Within a Bayesian framework, the intercept and slope of the linear relationship between noise magnitude and wavelength are treated as parameters to be estimated, and a relaxed prior is given to them. Simultaneously, a heavy-tailed t-distribution (equivalent to adaptive weighting) is used to describe the residuals to improve robustness to outliers and model mismatch. By jointly utilizing information from all Fraunhofer standard lines through posterior inference, estimates of the intercept and slope, along with their uncertainties (covariance / confidence intervals), are obtained. Adaptive weights for each Fraunhofer standard line can be output to identify and suppress outlier observations. If a systematic bias is found at the endpoints of the preset band, a quadratic correction term can be added to the EIV model. The information criterion (BIC / WAIC) is then used to determine whether the Fraunhofer standard line obtained in step 7, where a systematic bias exists at the endpoints of the preset band, should be retained. WAIC represents the Watanabe information criterion, and BIC represents the Bayesian information criterion. It is understood that the information criterion is not limited to BIC / WAIC. Overall, this method is more stable, traceable, and provides a more complete uncertainty assessment than ordinary least squares regression when both "error is in the independent variable" and "outliers exist."

[0111] Assume the latent truth values ​​satisfy a linear relationship:

[0112]

[0113] in, This represents the observed noisy measurement value. This indicates the corresponding known reference wavelength. The slope of the linear relationship satisfied by the latent truth value. It represents the intercept of the linear relationship satisfied by the potential truth value.

[0114] The actual observations contain noise (heteroscedasticity) at both ends. A weighted robust line representing the "error in the variable" is plotted on the selected optimal subset: ; , and give , Relaxed Gaussian Prior Scale mixing (adaptive weighting) through t-distribution Obtain an anomaly-resistant posterior estimate and output. and If the endpoint residuals exhibit systematic bias, a quadratic term is then introduced. And determine whether to retain it based on the Information Criteria (BIC / WAIC). Indicates the first The estimated center wavelength of the Fraunhofer standard line. Indicates the first The reference wavelength of the Fraunhofer standard line. Indicates the first The noise term for the wavelength offset of the Fraunhofer standard line. Indicates the first The measurement wavelength of the Fraunhofer standard line. Indicates a normal distribution. Describing the degrees of freedom as The t-distribution, Intercept The prior mean, This represents the corresponding prior variance. Indicates slope The prior mean, This represents the corresponding prior variance. This represents the estimated value of the intercept obtained through EIV estimation. This represents the estimated slope obtained through the EIV model. express covariance, The systematic bias term represents the residual.

[0115] Step 8 yields the final selected Fraunhofer standard line. Following step 8, the following steps may be performed; it is understood that steps 10 and 9 are not required to coexist, nor is it required that step 10 follow step 9.

[0116] Step 9: Following the steps above, calculate other step size / wavelength prediction values ​​(any step size / wavelength prediction value) and the 95% prediction interval.

[0117] In this embodiment, an arbitrary step size / wavelength prediction value and a 95% prediction range are given to facilitate the subsequent transmission of product uncertainty.

[0118] In this embodiment, based on the selected Fraunhofer standard line and the estimated model parameters, the final wavelength mapping function and its uncertainty (95% prediction interval) are given, which facilitates subsequent product error propagation and quality control, and enables prediction across the entire wavelength band and quantification of the confidence interval.

[0119] Establish (or ), Indicates according to step size The predicted wavelength, for any step size, gives a 95% prediction interval, satisfying:

[0120]

[0121] in, This represents a representative value (or is negligible) of the uncertainty originating from the pixel center. This represents the prediction interval with a confidence level of 95%. This represents the secondary correction coefficient.

[0122] Step 10: Verify the finally selected Fraunhofer standard line using mercury lamp spectroscopy.

[0123] Without using training samples, the accuracy of the mapping is independently tested using the mercury lamp feature line (converted to either air or vacuum) to obtain the center. and , Indicates the center wavelength of the mercury lamp. This represents the center wavelength width of the mercury lamp, mapped to... , This represents the wavelength of the mercury lamp estimated using the calibration model, which is the wavelength value predicted after regression calibration. The residual is then calculated. and standardized residuals ,

[0124] ,

[0125] Summary of RMSE, maximum absolute error, endpoint error and 95% coverage ( The proportion of ( ) is used as a "completely independent" acceptance indicator; if there is an abnormal standard line, that is or The performance of downgrading or removing anomalies and reporting "containing anomalies / removing anomalies" respectively is evaluated.

[0126] In addition, regarding long-term on-orbit operation and maintenance and version expansion, this route selection will be used to collect... , Archive the data along with covariance, threshold and hyperparameter, applicable temperature zone / date range, and mercury lamp validation report; multi-day / multi-batch data can be processed using hierarchical Bayes or Kalman filtering. and Perform time-series updates to track temperature drift and aging.

[0127] This disclosure provides a wavelength selection system for passive spectral calibration of an on-orbit ultraviolet spectrometer, characterized by comprising:

[0128] The list creation module is used to create a list of Fraunhofer standard lines for a preset band, wherein the preset band belongs to the ultraviolet band;

[0129] The preprocessing module is used to preprocess the frames containing each Fraunhofer standard line in the Fraunhofer standard line list to obtain a clean one-dimensional spectrum.

[0130] The modeling module is used to perform single-line Gaussian fitting and local baseline modeling on the Fraunhofer standard line in a clean one-dimensional spectrum.

[0131] The quality control module is used to perform quality control on Fraunhofer standard lines in a clean one-dimensional spectrum by utilizing asymmetry and wavelength deviation, and obtains a set of quality-controlled Fraunhofer standard lines.

[0132] The calculation module is used to calculate the quality weight of each Fraunhofer standard line in the set of quality-controlled Fraunhofer standard lines, and to calculate the similarity kernel of each Fraunhofer standard line with other Fraunhofer standard lines.

[0133] The first screening module is used to screen the Fraunhofer standard lines in the set of quality-controlled Fraunhofer standard lines based on the quality weight and the similarity kernel, and based on the constraints of wavelength range endpoints, the minimum spacing between two Fraunhofer standard lines, and the constraint of removing abnormal Fraunhofer standard lines, using a determinant point process to obtain the optimal subset of the set of quality-controlled Fraunhofer standard lines.

[0134] The second screening module is used to jointly evaluate the Fraunhofer standard line based on cross-day consistency and information content, and select the Fraunhofer standard line from the optimal subset that takes into account both cross-day consistency and information content maximization.

[0135] The linear regression calibration module is used to perform linear regression calibration on the Fraunhofer standard line, which balances cross-day consistency and maximizes information content, within a Bayesian framework, through a variable error model, to obtain the final selected Fraunhofer standard line.

[0136] In specific implementation, the wavelength selection system for passive spectral calibration of an on-orbit ultraviolet spectrometer can be implemented by referring to the wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer in any of the above embodiments. The specific implementation steps will not be repeated.

[0137] This disclosure also provides an electronic device comprising: a memory; one or more processors; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the wavelength selection methods for passive spectral calibration of an on-orbit ultraviolet spectrometer.

[0138] This disclosure also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer.

[0139] This disclosure presents a systematic method and system for line selection, taking into account asymmetry, wavelength deviation, quality weight, similarity kernel, cross-day consistency and information content, constraints on wavelength range endpoints, constraints on the minimum spacing between two Fraunhofer standard lines, and constraints on the removal of abnormal Fraunhofer standard lines. It performs multi-angle screening, resulting in high accuracy and stability in line selection.

[0140] Specifically, this disclosure improves upon on-orbit measurement data by employing an improved ultraviolet on-orbit spectral calibration method using the Fraunhofer standard line. It analyzes wavelength variations since the instrument began operating in orbit and uses the TSIS-1 mixed solar spectrum (TSIS-1HSRS) as a reference spectrum to simulate spectral broadening caused by spectral asymmetry at different resolutions. Furthermore, it investigates the broadening effect caused by spectral asymmetry and quantifies its correlation with peak position shift, using spectral asymmetry and... Thresholding to remove defective lines incorporates peak position uncertainty into the likelihood using an error-in-variable / Bayesian robust straight line, and provides RMSE, endpoint error, and coverage using independent mercury lamp verification. This allows for more stable and traceable spectral calibration results in on-orbit environments with limited resolution and line count.

[0141] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0142] The embodiments described above are merely illustrative of several implementations of this disclosure, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this disclosure, and these all fall within the protection scope of this disclosure. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer, characterized in that, Includes the following steps: Establish a list of Fraunhofer standard lines for preset bands, wherein the preset bands belong to the ultraviolet band; Preprocess the frames containing each Fraunhofer standard line in the Fraunhofer standard line list to obtain a clean one-dimensional spectrum; Single-line Gaussian fitting and local baseline modeling are performed on the Fraunhofer standard line in a clean one-dimensional spectrum. Quality control of Fraunhofer standard lines in a clean one-dimensional spectrum was performed using asymmetry and wavelength deviation, resulting in a set of quality-controlled Fraunhofer standard lines. For the set of Fraunhofer standard lines after quality control, calculate the quality weight of each Fraunhofer standard line and calculate the similarity kernel between each Fraunhofer standard line and other Fraunhofer standard lines. Based on the quality weight and the similarity kernel, and based on the constraints of wavelength range endpoints, the minimum spacing between two Fraunhofer standard lines, and the constraint of removing abnormal Fraunhofer standard lines, the Fraunhofer standard lines in the set of quality-controlled Fraunhofer standard lines are screened using a determinant point process to obtain the optimal subset of the set of quality-controlled Fraunhofer standard lines. Based on the joint evaluation of cross-day consistency and information content, the Fraunhofer standard line is selected from the optimal subset to balance cross-day consistency and maximize information content. Within the Bayesian framework, the Fraunhofer standard line, which balances cross-day consistency and maximizes information content, is calibrated by linear regression using a variable error model, resulting in the final selected Fraunhofer standard line.

2. The wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer according to claim 1, characterized in that, Based on the finally selected Fraunhofer standard line, the spectral mapping range is determined, and the prediction range of the on-orbit ultraviolet spectrometer is determined.

3. The wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer according to claim 1, characterized in that, The final selected Fraunhofer standard line was independently verified using mercury lamp spectroscopy.

4. The wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer according to claim 1, characterized in that, The model for single-line Gaussian fitting and local baseline modeling of the Fraunhofer standard line in a clean one-dimensional spectrum is as follows: ； in, Indicates the wavelength of the reference standard spectral line. This represents the intercept of the coarse mapping. Indicates the slope of the coarse mapping. Indicates the step size. Indicates the step position One-dimensional spectral intensity at that location, The constant term representing the local baseline, The linear term representing the local baseline, For spectral line depth, Indicates the center wavelength. Indicates the equivalent linewidth.

5. The wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer according to claim 1, characterized in that, The method of using asymmetry and wavelength deviation to perform quality control on Fraunhofer standard lines in a clean one-dimensional spectrum yields a set of quality-controlled Fraunhofer standard lines, including: Perform baseline subtraction; The wavelength is less than the center wavelength, calculated based on the spectral irradiance distribution curve of the Fraunhofer standard. Integral area on the side and wavelength greater than center wavelength Integral area on the side Define asymmetry , Calculate asymmetry ; Preserve while satisfying asymmetry Meets the preset range The Fraunhofer standard line with a signal-to-noise ratio reaching the lower limit is used to obtain the set of quality-controlled Fraunhofer standard lines. This represents the deviation of the Fraunhofer standard line from the coarse mapping estimate. This represents the maximum estimated deviation of the single Fraunhofer standard line.

6. The wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer according to claim 1, characterized in that, The formula for calculating the quality weight is: ； in, This indicates the number of the Fraunhofer standard line that is currently retained. Indicates the first Scalar mass weights of the Fraunhofer standard lines Represents an exponential function. This represents the weighting hyperparameter, used to control the severity of the penalty for single-line deviation. This represents the weighting hyperparameter, used to control the degree of preference for the signal-to-noise ratio. Indicates the first The normalized result of the single-line deviation of the coarse mapping estimate of the Fraunhofer standard line. Indicates the first Normalized signal-to-noise ratio results for the Fraunhofer standard lines; The formula for calculating the similarity kernel is: ； in, This indicates the number of the Fraunhofer standard line that is currently retained. Indicates the first The location of the Fraunhofer standard line, Indicates the first The location of the Fraunhofer standard line, Indicates the first The location of the Fraunhofer standard line and the first The similarity kernel between the positions of the Fraunhofer standard lines. Indicates the relevant scale.

7. The wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer according to claim 1, characterized in that, The optimal subset is fixed as , .

8. The wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer according to claim 1, characterized in that, The steps for selecting the Fraunhofer criterion line that balances cross-day consistency and information maximization from the optimal subset based on the joint evaluation of cross-day consistency and information maximization include: For the same Fraunhofer standard line observed over multiple days, the cross-day variance of the central estimate is summarized in the optimal subset of ultraviolet spectra. Deviation from average wavelength ; The Fraunhofer standard line, which balances cross-day consistency and maximizes information content, is selected from the optimal subset by minimizing the objective function. The objective function is: ； in, Represents the objective function value. This represents the weighted Fisher information matrix. Indicates the first The Fraunhofer standard line belongs to the optimal subset. , This represents the stability penalty coefficient, used to adjust the weight of cross-day variance in the quality weighting. Indicates the first The estimated center wavelength of the Fraunhofer standard line. This indicates the strength of the control variance penalty. This indicates the intensity of the penalty for controlling the average deviation. Indicates the first The average wavelength deviation of the Fraunhofer standard lines.

9. The wavelength selection method for passive spectral calibration of an on-orbit ultraviolet spectrometer according to claim 1, characterized in that, The linear regression calibration includes: setting the noise level and wavelength to conform to a linear relationship; during the linear regression calibration process, if a systematic deviation is found at the endpoint of the preset band of the Fraunhofer standard line, the information criterion is used to determine whether the Fraunhofer standard line should be retained, and all retained Fraunhofer standard lines are used as the final selected Fraunhofer standard lines.

10. A wavelength selection system for passive spectral calibration of an on-orbit ultraviolet spectrometer, characterized in that, include: The list creation module is used to create a list of Fraunhofer standard lines for a preset band, wherein the preset band belongs to the ultraviolet band; The preprocessing module is used to preprocess the frames containing each Fraunhofer standard line in the Fraunhofer standard line list to obtain a clean one-dimensional spectrum. The modeling module is used to perform single-line Gaussian fitting and local baseline modeling on the Fraunhofer standard line in a clean one-dimensional spectrum. The quality control module is used to perform quality control on Fraunhofer standard lines in a clean one-dimensional spectrum by utilizing asymmetry and wavelength deviation, and obtains a set of quality-controlled Fraunhofer standard lines. The calculation module is used to calculate the quality weight of each Fraunhofer standard line in the set of quality-controlled Fraunhofer standard lines, and to calculate the similarity kernel of each Fraunhofer standard line with other Fraunhofer standard lines. The first screening module is used to screen the Fraunhofer standard lines in the set of quality-controlled Fraunhofer standard lines based on the quality weight and the similarity kernel, and based on the constraints of wavelength range endpoints, the minimum spacing between two Fraunhofer standard lines, and the constraint of removing abnormal Fraunhofer standard lines, using a determinant point process to obtain the optimal subset of the set of quality-controlled Fraunhofer standard lines. The second screening module is used to jointly evaluate the Fraunhofer standard line based on cross-day consistency and information content, and select the Fraunhofer standard line from the optimal subset that takes into account both cross-day consistency and information content maximization. The linear regression calibration module is used to perform linear regression calibration on the Fraunhofer standard line, which balances cross-day consistency and maximizes information content, within a Bayesian framework, through a variable error model, to obtain the final selected Fraunhofer standard line.

Images