Atmospheric monitoring method
By combining blackbody spectral calibration and principal component analysis with spectral similarity matching, the problems of accuracy in identifying atmospheric pollutants and multi-component detection under moving backgrounds were solved, enabling automatic and rapid pollutant identification and quantitative analysis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HANGZHOU PUYU TECH DEV CO LTD
- Filing Date
- 2026-04-01
- Publication Date
- 2026-06-12
AI Technical Summary
Existing technologies for identifying atmospheric pollutants in moving environments are easily affected by subjective factors, are cumbersome to operate, and have difficulty detecting multiple components simultaneously, especially when characteristic bands overlap, making them prone to missed detections.
Brightness-temperature spectra are obtained by blackbody spectral calibration, background is removed by principal component eigenvector fitting, and automatic identification and quantitative analysis are achieved by characteristic band detection and spectral similarity matching combined with non-negative least squares fitting.
It reduces human intervention, improves the accuracy and speed of identification, can automatically extract feature bands in moving backgrounds, achieves simultaneous identification of multiple components, and reduces the false alarm rate.
Smart Images
Figure CN122193130A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to absorption technology, and more particularly to atmospheric monitoring methods. Background Technology
[0002] Fourier transform infrared (FTIR) telemetry, as a passive remote sensing method for detecting and identifying atmospheric pollutants, is increasingly widely used in various fields such as industry and fire protection. However, the background changes brought about by mobile detection and the complexity of the atmospheric environment itself pose significant challenges to the identification of pollutants. The most crucial step is to extract characteristic spectral information from complex systems for pollutant identification.
[0003] Currently, several methods have been developed for detecting pollutant gases under measurement conditions with a moving background, such as: 1. To determine the background to be measured, a background energy spectrum is collected beforehand. During automatic measurement, the absorption spectrum is obtained by subtracting the background spectrum from the energy spectrum, and the composition is directly determined from the absorption spectrum.
[0004] This method is susceptible to subjective influences under human intervention, resulting in unstable results and cumbersome operation. Furthermore, if the background energy spectrum is not properly selected, false alarms are likely to occur.
[0005] 2. Brightness-temperature spectroscopy. This method uses blackbody spectra at different temperatures for calibration to obtain brightness-temperature spectra, and then identifies and quantifies the components from these spectra.
[0006] The brightness temperature spectrum obtained by this method under outdoor background conditions (such as leaves, walls, or strong sunlight) is very complex. Usually, the target absorption peak is submerged, meaning that the absorption produced by the target component is very small relative to the complex background. Therefore, many components are missed. In addition, current methods have difficulty detecting multiple components simultaneously, especially when multiple components have certain overlapping characteristic bands. There is a lack of effective characteristic band detection and analysis methods. Summary of the Invention
[0007] To address the shortcomings of the existing technical solutions, the present invention provides an atmospheric monitoring method.
[0008] The objective of this invention is achieved through the following technical solution: An atmospheric monitoring method includes the following steps: A1. Use the blackbody spectrum to calibrate the collected energy spectrum and obtain the brightness temperature spectrum; A2. Fit the brightness temperature spectrum using principal component eigenvectors, remove the background, and obtain the absorption spectrum; A3. Perform characteristic band detection on the absorption spectrum; A4. Perform spectral similarity matching using characteristic bands and spectral databases to obtain a list of candidate components; A5. Fit the absorption spectrum using the candidate component list to obtain the concentration of each component in the atmosphere.
[0009] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The algorithm for automatically extracting characteristic bands and automatically identifying atmospheric pollutants reduces manual intervention and avoids the input of prior background information, making FTIR telemetry operations simpler and faster; 2. The adjacent location difference method is used to automatically determine whether a concentration difference exists between adjacent locations; 3. Iterative principal component analysis is used to update the background components, and least squares fitting is used to improve the accuracy of the absorption spectrum; 4. Use secondary filtering for feature band detection to reduce the impact of background, noise, and interference components on target component detection; 5. A multi-band similarity matching algorithm is used to search the standard library to obtain candidate target spectra. At the same time, the non-negative least squares fitting method is used to fit and analyze the absorption spectrum to obtain the fitting coefficients. Qualitative and quantitative analysis is achieved through the coefficients. 6. Feature spectrum extraction under movable background (up to 5 background categories), enabling simultaneous identification of multiple components for overlapping signals. Attached Figure Description
[0010] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are merely illustrative of the technical solutions of this invention and are not intended to limit the scope of protection of this invention. In the drawings: Figure 1 This is a flowchart illustrating the atmospheric monitoring method according to the present invention; Figure 2 This is the brightness temperature spectrum according to the present invention. Detailed Implementation
[0011] Figures 1-2 The following description illustrates optional embodiments of the invention to teach those skilled in the art how to implement and reproduce the invention. Some conventional aspects have been simplified or omitted to teach the technical solutions of the invention. Those skilled in the art should understand that variations or substitutions derived from these embodiments will be within the scope of the invention. Those skilled in the art should understand that the following features can be combined in various ways to form multiple variations of the invention. Therefore, the invention is not limited to the optional embodiments described below, but is defined only by the claims and their equivalents.
[0012] Example 1
[0013] This embodiment provides an atmospheric monitoring method, such as... Figure 1 As shown, it includes the following steps: A1. Use the blackbody spectrum to calibrate the collected energy spectrum and obtain the brightness temperature spectrum; A2. Fit the brightness temperature spectrum using principal component eigenvectors, remove the background, and obtain the absorption spectrum; A3. Perform characteristic band detection on the absorption spectrum; A4. Perform spectral similarity matching using characteristic bands and spectral databases to obtain a list of candidate components; A5. Fit the absorption spectrum using the candidate component list to obtain the concentration of each component in the atmosphere.
[0014] To obtain accurate matching results, further step A1 involves establishing a database: Collect standard absorption spectra of all analytes, and establish 600–1500 cm⁻¹ spectra for each standard absorption spectrum. -1 Model band M i ={m1,m2···m k}, where i represents the index of the different standard, and k represents the number of characteristic bands; All standard absorption spectra are standardized to ensure that the absorption spectrum intensity corresponds to 1 ppm·m.
[0015] To obtain an accurate absorption spectrum, further, in step A2, the difference in brightness temperature spectra at adjacent time points is first obtained; If the difference exceeds the threshold, the component concentration changes, and the brightness temperature spectrum is fitted.
[0016] To obtain an accurate brightness temperature spectrum, the calibration method is further as follows: Collect multiple data points related to temperature t i The corresponding blackbody spectrum h at the j-th wavelength ij i = 1, 2, ..., K, j = 1, 2, ..., N, where K is an integer not less than 3 and N is an integer not less than 2; Using t i =k j ·h ij +b j Obtain the coefficient k i b i ; Based on the collected energy spectrum e ij and x i =k j ·e ij +b j The brightness temperature spectrum x was obtained. i .
[0017] To obtain an accurate absorption spectrum, the background matrix is further decomposed into principal components, where T is the score matrix, V is the load matrix, and W is the transpose. B=T·V W; Obtain the first 5 columns \(T\) of the score matrix \(T\). 1-5 Perform least squares fitting on the brightness temperature spectrum; , \(y = x - T\) 1-5 ·s opt ; \(s\) is the coefficient matrix.
[0018] s opt is the optimal coefficient matrix, and the absorption spectrum \(y\) is obtained.
[0019] In order to obtain accurate characteristic bands, further, use the quadratic filtering algorithm to detect the characteristic bands of the absorption spectrum \(y\). The quadratic filtering algorithm is as follows: Use two filters to take the derivative of the absorption spectrum \(y\); y i ˊ = (-2y i-2 - y i-1 + y i+1 + 2y i+2 ) / 2, y i ˊˊ = (2y i-2 - y i-1 - 2y i - y i+1 + 2y i+2 ) / 7; The maximum \(F0\) of the amplitude difference between adjacent points of the absorption spectrum \(y\); the maximum \(F1\) of the first derivative \(y\) i ˊ, and the maximum \(F2\) of the second derivative \(y\) i ˊˊ; Among the data below the amplitude difference, \(y\) i ˊ and \(y\) i ˊˊ, respectively set the thresholds (such as 5% of the maximum value), and the median \(TF0\) of the amplitude difference, the median \(TF1\) of the first derivative, and the median \(TF2\) of the second derivative are used as the boundary thresholds for feature peak detection When \(F0 > TF0\) and \(F1 > TF1\), it is the left boundary; When \(F0 < TF0\) and \(F1 < TF1\), it is the right boundary; When the sign of the first derivative changes and \(F2 < TF2\), it is the peak vertex; Obtain the characteristic band interval \(G = \{g1, g2 ··· g k \}.
[0020] In order to obtain accurate similarity matching, further, the similarity matching is as follows: Calculate the overlap degree of the characteristic band \(G\) of the input absorption spectrum and the modeling band \(M\). By comparing the overlap degree of individual regions and retaining the intervals with an overlap degree greater than 0.7, divide the total number of overlapping points by the total number of points to obtain the overlap similarity; Calculate the Pearson similarity of intensity shape by piecing together points from all overlapping regions in the input spectrum and the library spectrum into a new vector, denoted as I and L respectively, and calculate the Pearson correlation coefficient. Calculate the weighted average of overlap similarity and shape similarity, sim. ; i is the index of overlapping regions greater than 0.7, PI is the number of overlapping region points in the statistical absorption spectrum, and PL is the number of overlapping region points in the standard spectrum of the statistical library. This is the mean of the corresponding vector; Based on the set similarity threshold, the weighted average similarity is filtered to obtain the candidate spectrum set S={s1,s2···s k}
[0021] To obtain accurate quantitative results, the absorption spectrum x is further analyzed using the acquired candidate spectrum set S. d Perform nonnegative least squares fitting, with the objective function being: c≥0; c is the fitting coefficient, and its length is equal to the number of candidate mass spectra. The absolute intensity and explanatory power of each component are calculated using the fitting coefficients. The components are then filtered using intensity and explanatory power thresholds, and the remaining components are the qualitatively identified components. The fitting coefficients are the concentrations.
[0022] Example 2
[0023] An application example of the atmospheric monitoring method in Example 1 of this embodiment.
[0024] In this application example, such as Figure 1 As shown, it includes the following steps: A1. Establish the database: Collect standard absorption spectra of all analytes, and establish 600–1500 cm⁻¹ spectra for each standard absorption spectrum. -1 Model band M i ={m1,m2···m k}, where i represents the index of different standards and k represents the number of characteristic bands.
[0025] All standard absorption spectra are standardized to ensure that the absorption spectrum intensity corresponds to 1 ppm·m.
[0026] The collected energy spectrum was calibrated using a blackbody spectrum to obtain the brightness temperature spectrum. The calibration method was as follows: Collect multiple data points related to temperature t i The corresponding blackbody spectrum h at the j-th wavelength ij i = 1, 2, ..., K, j = 1, 2, ..., N, where K is an integer not less than 3 and N is an integer not less than 2.
[0027] Using t i =k j ·h ij +b j Obtain the coefficient k i b i .
[0028] Based on the collected energy spectrum e ij and x i =k j ·e ij +b j The brightness temperature spectrum x was obtained. i .
[0029] A2. Obtain the difference x in the brightness temperature spectrum at adjacent time points. d .
[0030] If the difference exceeds the threshold, the component concentration changes, and the brightness temperature spectrum is fitted.
[0031] Brightness temperature spectra are fitted using principal component eigenvectors to remove background and obtain absorption spectra.
[0032] Principal component decomposition is performed on the background matrix, where T is the score matrix, V is the load matrix, and W is the transpose.
[0033] B=T·V W .
[0034] Obtain the first 5 columns T of the fractional matrix T 1-5 Least squares fitting was performed on the brightness temperature spectrum.
[0035] y=xT 1-5 ·s opt ; s is the coefficient matrix.
[0036] s opt It is the optimal coefficient matrix, from which the absorption spectrum y is obtained.
[0037] A3. Perform characteristic band detection on the absorption spectrum, specifically: The absorption spectrum y is subjected to feature band detection using a quadratic filtering algorithm, which is as follows: The absorption spectrum y is differentiated using two filters.
[0038] y i ˊ=(-2y i-2 -y i-1 +y i+1 +2y i+2 ) / 2, y i ˊˊ=(2y i-2 -y i-1 -2y i -y i+1+2y i+2 ) / 7。
[0039] The maximum F0 of the amplitude difference between two adjacent points of the absorption spectrum y; the first derivative y i ˊs maximum value F1, and the second derivative y i ˊˊs maximum value F2.
[0040] Lower than the amplitude difference, y i ˊ and y i ˊˊs respective set thresholds (such as 5% of the maximum value), the median TF0 of the amplitude difference, the median TF1 of the first derivative, and the median TF2 of the second derivative are used as the boundary thresholds for feature peak detection respectively.
[0041] When F0 > TF0 and F1 > TF1, it is the left boundary.
[0042] When F0 < TF0 and F1 < TF1, it is the right boundary.
[0043] When the sign of the first derivative changes and F2 < TF2, it is the peak vertex.
[0044] Obtain the characteristic band interval G = {g1, g2 ··· g k}.
[0045] A4. Use the characteristic band and the spectral database to perform spectral similarity matching to obtain a list of candidate components, specifically: Calculate the overlap degree of the input absorption spectrum characteristic band G and the modeling band M. By comparing the overlap degree of individual regions and retaining the intervals with an overlap degree greater than 0.7, divide the total number of overlapping points by the total number of points to obtain the overlap similarity.
[0046] Calculate the Pearson similarity of the intensity shape. Piece together the points in all overlapping regions of the input spectrum and the library spectrum into new vectors, denoted as I and L respectively, and calculate the Pearson correlation coefficient.
[0047] Calculate the weighted average sim of the overlap similarity and the shape similarity.
[0048] .
[0049] i is the index of the overlapping region greater than 0.7, PI is the number of overlapping region points counted in the absorption spectrum, PL is the number of overlapping region points counted in the library standard spectrum, is the mean of the corresponding vector.
[0050] According to the set similarity threshold, screen the weighted average similarity to obtain the candidate spectrum set S = {s1, s2 ··· s k}.
[0051] A5. The absorption spectrum was fitted using a candidate component list to obtain the concentrations of each component in the atmosphere, specifically: Using the obtained candidate spectrum set S to analyze the absorption spectrum x d Perform nonnegative least squares fitting, with the objective function being: , c≥0.
[0052] c is the fitting coefficient, and its length is equal to the number of candidate mass spectra.
[0053] The absolute intensity and explanatory power of each component are calculated using the fitting coefficients. The components are then filtered using intensity and explanatory power thresholds, and the remaining components are the qualitatively identified components. The fitting coefficients are the concentrations.
[0054] like Figure 2 As shown, in this embodiment, the weighted similarity of methanol is 0.965, and the concentration is 127.02 ppm·m; the weighted similarity of acetone is 0.985, and the concentration is 15.79 ppm·m.
Claims
1. An atmospheric monitoring method, characterized in that, The monitoring method includes the steps: A1. Calibrate the acquired energy spectrum using a blackbody spectrum to obtain a brightness temperature spectrum; A2. Fit the brightness temperature spectrum using the principal component eigenvector to remove the background and obtain an absorption spectrum; A3. Detect the characteristic bands of the absorption spectrum; A4. Perform spectral similarity matching using the characteristic bands and a spectral database to obtain a list of candidate components; A5. Fit the absorption spectrum using the list of candidate components to obtain the concentrations of various components in the atmosphere.
2. The analytical method according to claim 1, characterized in that, In step A1, establish a database: Collect standard absorption spectra of all analytes, and establish 600–1500 cm⁻¹ spectra for each standard absorption spectrum. -1 Model band M i ={m1,m2···m k }, where i represents the index of the different standard, and k represents the number of characteristic bands; Standardize all standard absorption spectra to ensure that the absorption spectrum intensity corresponds to 1 ppm·m.
3. The analytical method according to claim 1, characterized in that, In step A2, firstly, the difference x between the brightness temperature spectra at adjacent time points is obtained. d ; If the difference exceeds the threshold, the component concentration changes, and fit the brightness temperature spectrum.
4. The analytical method according to claim 1 or 3, characterized in that, The calibration method is: Collect multiple data points related to temperature t i The corresponding blackbody spectrum h at the j-th wavelength ij i = 1, 2, ..., K, j = 1, 2, ..., N, where K is an integer not less than 3 and N is an integer not less than 2; Using t i =k j ·h ij +b j Obtain the coefficient k i b i ; Based on the collected energy spectrum e ij and x i =k j ·e ij +b j The brightness temperature spectrum x was obtained. i .
5. The analysis method according to claim 4, wherein Perform principal component decomposition on the background matrix, T is the score matrix, V is the loading matrix, and W is the transpose; B=T·V W ; Obtain the first 5 columns T of the fractional matrix T 1-5 Least-squares fitting was performed on the brightness temperature spectrum; y=xT 1-5 ·s opt ; s is the coefficient matrix; s opt It is the optimal coefficient matrix, from which the absorption spectrum y is obtained.
6. The analytical method according to claim 1, characterized in that, Use a quadratic filtering algorithm to detect the characteristic bands of the absorption spectrum y, and the quadratic filtering algorithm is: Use two filters to take the derivative of the absorption spectrum y; y i ˊ=(-2y i-2 -y i-1 +y i+1 +2y i+2 ) / 2,y i ˊˊ=(2y i-2 -y i-1 -2y i -y i+1 +2y i+2 ) / 7; The maximum amplitude difference F0 between two adjacent points in the absorption spectrum y; the first derivative y i The maximum value of ˊ is F1, and the second derivative is y. i The maximum value of ˊˊ is F2; Below the amplitude difference, y i ˊ and y i In the data with each set threshold, the median of the amplitude difference TF0, the median of the first derivative TF1, and the median of the second derivative TF2 are used as the boundary thresholds for feature peak detection. When F0 > TF0 and F1 > TF1, it is the left boundary; When F0 < TF0 and F1 < TF1, it is the right boundary; When the sign of the first derivative changes and F2 < TF2, it is the peak vertex; Obtain the characteristic band interval G={g1,g2···g k } 7. The analytical method according to claim 6, characterized in that, The set threshold is 5% of its maximum value.
8. The analytical method according to claim 2, characterized in that, The similarity matching is: Calculate the overlap degree between the characteristic bands G of the input absorption spectrum and the modeling bands M, compare the overlap degree of individual regions, and retain the intervals with an overlap degree greater than 0.
7. Divide the total number of overlapping points by the total number of points to obtain the overlap similarity; Calculate the Pearson similarity of the intensity shape. Piece together the points in all overlapping regions of the input spectrum and the library spectrum into new vectors, denoted as I and L respectively, and calculate the Pearson correlation coefficient; Calculate the weighted average sim of the overlap similarity and the shape similarity; ； i is the index of overlapping regions greater than 0.7, PI is the number of overlapping region points in the statistical absorption spectrum, and PL is the number of overlapping region points in the standard spectrum of the statistical library. , This is the mean of the corresponding vector; Based on the set similarity threshold, the weighted average similarity is filtered to obtain the candidate spectrum set S={s1,s2···s k } 9. The analysis method according to claim 1, wherein Using the obtained candidate spectrum set S to analyze the absorption spectrum x d Perform nonnegative least squares fitting, with the objective function being: c≥0; c is the fitting coefficient, and its length is equal to the number of candidate mass spectra; Calculate the absolute intensity and interpretability of each component using the fitting coefficient, filter using the intensity threshold and the interpretability threshold, retain the components as the qualitatively identified components, and the fitting coefficient is the concentration.