A gas identification method suitable for low-resolution high-noise spectral data

Abstract

The application discloses a gas identification method suitable for low-resolution high-noise spectral data, and is suitable for low-resolution high-noise spectral data obtained by a medium-long wave infrared spectrometer or a hyperspectral imager in an open space in a passive and passive measurement mode, and performs the following steps: a) extracting spectral radiation values of a measured gas at corresponding characteristic spectral position wave number coordinates by taking preset typical gas spectral characteristic points as templates, and filtering noise interference outside the characteristic spectral position wave number coordinates; b) judging a positive-negative matching number coefficient of multiple characteristic spectrums, judging a positive-negative matching direction coefficient of the multiple characteristic spectrums, and jointly judging a characteristic vector cosine similarity, so that the dimension of the judgment is increased, and the positive-negative temperature difference change of the gas and the background is adapted; and c) comprehensively judging by multiple gas similarity sorting methods. The application effectively matches the spectral gas identification accuracy, and is suitable for application scenes limited by conditions such as an open space environment, long-distance passive detection, and low sensor sensitivity.

Description

Technical Field

[0001] This invention belongs to the field of chemical gas detection and identification technology, and relates to a gas identification method suitable for low-resolution, high-noise spectral data. Background Technology

[0002] Infrared spectroscopy and infrared hyperspectral imaging technologies can directly identify the type and spatial distribution of gaseous targets in the atmosphere based on the infrared absorption or radiation spectra. This requires no artificial infrared light source and allows for long-distance, non-contact measurement. A common identification method involves matching the measured spectral curve with a preset standard gas absorption spectrum using a least-squares algorithm, with the matching result displayed as a pseudo-color image.

[0003] In patent (201110146379.3) "A Hyperspectral Imaging Method for Detecting and Identifying Chemical Gases," a spectral image is first captured when there is no chemical gas in the area to be detected, obtaining the pixel vector of the background. Then, the spectral vector of the chemical gas to be detected is obtained from a chemical gas spectral library. Finally, a spectral image is captured when there is chemical gas in the area to be detected. This method requires pre-capturing an infrared spectral image in the area to be detected when there is no chemical gas, thus limiting its usability.

[0004] Patent (201610347241.2) "A Gas Detection Method Based on Hyperspectral Infrared Image Processing" utilizes a gas hyperspectral imaging model and a gas infrared absorption spectrum database to invert the gas spectral curve. Finally, the calculated values ​​are compared with the actual measured values ​​to identify the gas type and estimate its concentration. This method requires calculation of background and air radiation, has a long spectral processing time, and demands high spectral resolution and signal-to-noise ratio.

[0005] In the patent (201710589248.X) "A Method for Gas Detection, Identification, and Concentration Representation Based on Hyperspectral Infrared Images," the absorbance-wavenumber curves of each pixel are used to determine the presence and wavenumber position of absorption peaks by differentiation. The curves are then fitted with the absorption peak shapes in a gas spectral library, and the difference is calculated. A nonlinear least squares method is then used to identify and invert the gas species. This method requires high spectral resolution and a high signal-to-noise ratio. Summary of the Invention

[0006] (I) Purpose of the Invention

[0007] The purpose of this invention is to address the technical problems of low target detection rate and high false alarm rate in existing chemical gas detection and identification methods based on spectral matching when analyzing spectral measurement data with low spectral resolution and low signal-to-noise ratio, and to provide a gas identification method suitable for low-resolution, high-noise spectral data.

[0008] (II) Technical Solution

[0009] To address the aforementioned technical problems, this invention provides a gas identification method suitable for low-resolution, high-noise spectral data, comprising the following steps:

[0010] Step 1: Preprocess the raw spectral curve of the gas being measured. Preprocessing includes: baseline removal, smoothing to remove high-frequency noise, and normalization.

[0011] Step 2: Based on the specific scenario of this measurement, select a group of possible gas types from the gas spectral feature library, and extract the spectral feature template of the i-th (i = 1, 2, 3, ..., M) gas from this group of suspicious gases.

[0012] The template consists of two arrays. The first array contains the wavenumber (or wavelength) coordinates S of the characteristic spectral positions. p The second array is the wavenumber coordinates of the characteristic spectral position S. p The corresponding normalized spectral radiance value A s Both arrays have a length of N, where N is the number of characteristic spectra (N = 1, 2, 3...). For example... Figure 3 As shown, the continuous curve represents the standard spectral curve, and the x-coordinates of points A, B, ..., I represent the characteristic spectral positions and wavenumber coordinates S. p The vertical axis represents the normalized spectral radiance value A. s ;

[0013] The gas spectral feature library obtains spectral feature data of known gases through experimental measurements or by querying standard gas spectral feature databases in the industry; a group of suspected gases is selected, which includes M types (M = 1, 2, 3...); if the types of gases that may appear in the measurement scenario are uncertain, all gas types in the gas spectral feature library can be selected to participate in subsequent calculations.

[0014] Step 3: Find the wavenumber coordinates S of the characteristic spectral positions of the gas spectrum obtained in Step 2 from the spectral curve processed in Step 1. p The corresponding spectral radiation value at that location generates an array A of length N. d ;

[0015] Step 4: Array A s and array A d Each component performs a difference operation to generate a standard feature vector V. s and the characteristic vector V of the measured gas d The length of each vector is N-1;

[0016] Step 5: Calculate the coefficients for the number of positive and negative matches in the multi-feature spectrum and the direction coefficients for the positive and negative matches in the multi-feature spectrum, using the following formulas:

[0017] (1) K1 = max(n F nN )÷(n F +n N );

[0018] Where, n F The number of positive matches, i.e., the wavenumber coordinates of the characteristic spectral positions, S. p The corresponding gas feature vector V d Wavenumber coordinates S of the same characteristic spectral position p The corresponding standard feature vector V s The number of arrows pointing in the same direction, i.e., all pointing upwards or all pointing downwards; n N The number of negative matches, i.e., the wavenumber coordinates of the characteristic spectral position S. p The corresponding gas feature vector V d Wavenumber coordinates S of the same characteristic spectral position p The corresponding standard feature vector V s The number of arrows pointing in opposite directions, i.e., one arrow pointing upwards and the other downwards; where n F +n N =N-1, where N is the length of the eigenvector.

[0019] (2)n F >n N At that time, K2 = +1;

[0020] (3)n F ≤n N At that time, K2 = -1;

[0021] Wherein, K1 represents the coefficient of the number of positive and negative matches in the multi-feature spectrum, with a value ranging from 0 to 1; K2 represents the direction coefficient of the positive and negative matches in the multi-feature spectrum, with a value of +1 or -1;

[0022] If K1 ≤ threshold, it is determined as "no target gas" and jumps to step eight. If K1 > threshold, continue to step six.

[0023] It is worth noting that for semi-transparent media such as gases, opposite spectral radiation / absorption characteristics will be exhibited under high and low temperature backgrounds. Under high temperature backgrounds, absorption is observed, while under low temperature backgrounds, radiation is observed. The peaks and valleys of the spectral curves are in opposite directions. Figure 5 As shown. Therefore, in the process of calculating K1, the number of positive matches n was taken. F The number of negative matches n N The larger of the two, instead of just the number of forward matches n. F .

[0024] It is also worth noting that the characteristic spectral directions of random noise are random. Taking 10 characteristic spectra as an example, the probabilities of different numbers of positive and negative matches when matching with the standard spectrum are as follows:

[0025]

[0026]

[0027]

[0028]

[0029]

[0030]

[0031]

[0032]

[0033]

[0034]

[0035]

[0036] Therefore, it can be seen that the coefficient K1 of positive and negative matching of multi-feature spectrum under pure random noise has a 65.6% probability of not being greater than 0.6 and an 89% probability of not being greater than 0.7. Therefore, setting the threshold range to about 0.6 to 0.7 can eliminate most of the false alarms caused by noise interference.

[0037] Step Six: Calculate the cosine similarity of the feature vectors, keeping the sign of the result but not taking the absolute value. The specific formula is as follows:

[0038] K3 = V s ·V d

[0039] Step 7: Based on the results of Steps 5 and 6, calculate the similarity between the gas being tested and the template gas:

[0040] S = K1·K2·K3

[0041] Set threshold S th Gases with an S value greater than the threshold are retained and temporarily listed as potentially present; gases with an S value less than the threshold are discarded and considered not to exist.

[0042] It is worth noting that: threshold S th The value range is (0, K1). The value for calculating and identifying low-resolution, high-noise gas spectra is approximately 0.05 to 0.2, which needs to be adjusted according to different measuring instruments and usage scenarios.

[0043] It's also worth noting that since both K2 and K3 have signs, the calculation result S also carries a sign. A negative value indicates that the matching direction of the wavenumber coordinates of the characteristic spectral positions is inconsistent with the similarity direction. This is because the spectral radiance amplitude corresponding to a few characteristic spectral positions, or even a single characteristic spectral position, accounts for too large a proportion in the cosine similarity K3 calculation, thus skewing the entire calculation result. The direction coefficient K2 corrects the similarity S in a timely manner, avoiding false alarms caused by excessively large spectral radiance amplitudes corresponding to abnormal characteristic spectral coordinates.

[0044] It needs to be further explained that: such as Figure 6 and Figure 7 As shown, a standard characteristic spectrum curve and a measured curve are presented. The standard spectrum curve has one large peak and several smaller peaks. One large peak in the measured curve basically corresponds to one large peak in the standard spectrum curve, while the smaller peaks in the measured curve are not obvious. If only the cosine similarity of the eigenvectors K3 = V is used... s ·V d In this judgment, small peaks are ineffective and prone to misjudgment; relying solely on the coefficient K1 of positive and negative matching of multiple feature spectra is also susceptible to random noise interference; only by using S = K1·K2·K3, and combining three judgments, can the accuracy be improved.

[0045] Step 8: Change the suspected gas type (i+1) that is the focus of this measurement, repeat steps 2 to 7, obtain multiple similarity scores S, arrange them in order of size, and determine the gas type corresponding to the maximum similarity score.

[0046] It is also worth noting that the similarity S can characterize the relative concentration of gases, displayed through pseudo-color and numerical values.

[0047] Furthermore, in certain fixed locations where the types of suspected gases are clearly identified, only steps one through seven need to be calculated, and only the presence and relative concentration of the gas of interest need to be reported.

[0048] Ideally, spectral data for chemical gas detection exhibits high spectral resolution and signal-to-noise ratio, with prominent, accurately positioned characteristic absorption peaks and minimal noise. However, in practical applications involving passive gas measurement in open spaces, factors such as low gas concentration and small temperature difference between the gas and background result in broadened and blurred absorption peaks, positional shifts, and noticeable noise spikes or even spurious characteristic peaks. Figure 2 As shown.

[0049] (III) Beneficial Effects

[0050] The gas identification method for low-resolution, high-noise spectral data provided by the above technical solution has the following beneficial effects:

[0051] (1) The gas type is detected by using the spectral feature template of the gas of interest, and the noise interference outside the spectral radiation value corresponding to the wavenumber coordinate of the feature spectral position is filtered out. The accuracy of identification of low-resolution and high-noise spectral measurement data is higher and the false alarm rate is lower. In addition, the amount of calculation is small and the real-time performance is good.

[0052] (2) The method of judging the number of positive and negative matching coefficients of multi-feature spectrum, judging the direction coefficients of positive and negative matching of multi-feature spectrum, and judging the cosine similarity of feature vector is adopted. The three weak constraints are combined into one strong constraint, which reduces false alarms caused by a single weak constraint or missed detections caused by a single strong constraint, increases the dimension and accuracy of comprehensive judgment, and also adapts to the positive and negative temperature difference caused by the temperature change of the gas itself and the difference in the background temperature to which it diffuses.

[0053] (3) Multiple gas similarity ranking, adapting to the discrimination between gases whose characteristic spectral positions partially overlap or are close under low resolution and high noise conditions. Attached Figure Description

[0054] Figure 1 This is a flowchart of a gas identification method for low-resolution, high-noise spectral data according to the present invention.

[0055] Figure 2 This is a comparison chart of low-resolution, high-noise spectra and standard spectra for a gas identification method applicable to low-resolution, high-noise spectral data according to the present invention.

[0056] Figure 3 This is a schematic diagram of the standard spectral curve extraction feature spectral point positions and standard feature vector Vs for a gas identification method applicable to low-resolution, high-noise spectral data according to the present invention.

[0057] Figure 4 This is a schematic diagram of the extraction of feature spectral point positions and the feature vector Vd of the gas under test from the spectral curve of the gas under test corresponding to the standard spectral curve, which is a gas identification method applicable to low-resolution and high-noise spectral data according to the present invention.

[0058] Figure 5 These are the apparent spectral radiation characteristic curves of the gas being measured under three different background conditions: high temperature, low temperature, and room temperature.

[0059] Figure 6 This is a schematic diagram illustrating the positive matching and identification between a standard spectral curve with one large peak and multiple small peaks and the spectral curve under test.

[0060] Figure 7 This is a schematic diagram illustrating the negative matching and identification between a standard spectral curve with one large peak and multiple small peaks and the spectral curve under test. Detailed Implementation

[0061] To make the objectives, contents, and advantages of the present invention clearer, the specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples.

[0062] Reference Figure 1 As shown, the gas identification method for low-resolution, high-noise spectral data in this embodiment follows the following process:

[0063] Step 1: Preprocess the raw spectral curve of the gas being measured. Preprocessing includes: baseline removal, smoothing to remove high-frequency noise, and normalization.

[0064] Step 2: Based on the specific scenario of this measurement, select a group of possible gas types from the gas spectral feature library, and extract the spectral feature template of the i-th (i = 1, 2, 3, ..., M) gas from this group of suspicious gases.

[0065] The template consists of two arrays. The first array contains the wavenumber (or wavelength) coordinates S of the characteristic spectral positions. p The second array is the wavenumber coordinates of the characteristic spectral position S. p The corresponding normalized spectral radiance value A s Both arrays have a length of N, where N is the number of characteristic spectra (N = 1, 2, 3...). For example... Figure 3 As shown, the continuous curve represents the standard spectral curve, and the x-coordinates of points A, B, ..., I represent the characteristic spectral positions and wavenumber coordinates S. p The vertical axis represents the normalized spectral radiance value A. s ;

[0066] The gas spectral feature library obtains spectral feature data of known gases through experimental measurement or by querying industry standard gas spectral feature databases; a group of suspected gases is selected, which includes M types (M = 1, 2, 3...); if the types of gases that may appear in the measurement scenario are uncertain, all gas types in the gas spectral feature library can be selected to participate in subsequent calculations.

[0067] Step 3: Find the wavenumber coordinates S of the characteristic spectral positions of the gas spectrum obtained in Step 2 after the preprocessing in Step 1. p The corresponding spectral radiation value at that location generates an array A of length N. d ;

[0068] It is worth noting that: Figure 4 As shown, the continuous curve represents the low-resolution, high-noise spectral curve of the measured gas, and the x-coordinates of points A', B', ..., I' on the broken line represent the wavenumber coordinates S of the characteristic spectral position. p The vertical axis represents the corresponding normalized spectral radiance value A. d ;

[0069] Step 4: Array As and array A d Perform the front and back difference operations separately to generate the standard feature vector V. s and the characteristic vector V of the measured gas d The length of each vector is N-1;

[0070] It is worth noting that: Figure 3 As shown, the arrow represents the standard feature vector V. s Value, with the upward arrow indicating the eigenvector value V. s A positive value and a downward arrow indicate the eigenvector V. s The value is negative. For example... Figure 4 As shown, the arrow represents the characteristic vector V of the measured gas. d The upward arrow indicates the feature vector V. d A positive value and a downward arrow indicate the eigenvector V. d The value is negative;

[0071] Step 5: Calculate the coefficients for the number of positive and negative matches in the multi-feature spectrum and the direction coefficients for the positive and negative matches in the multi-feature spectrum, using the following formulas:

[0072] (1) K1 = max(n F n N )÷(n F +n N );

[0073] Among them, reference Figure 3 and Figure 4 The V shown s and V d n F The number of positive matches, i.e., the graph

[0074] 4 characteristic spectral position wavenumber coordinates S p The corresponding gas feature vector V d and Figure 3 The wavenumber coordinates S of the corresponding characteristic spectral position in the middle p The corresponding standard feature vector V s The number of arrows pointing in the same direction, i.e., all pointing upwards or all pointing downwards; n N This represents the number of negative matches, i.e. Figure 4 Wavenumber coordinates of characteristic spectral position S p The corresponding gas feature vector V d and Figure 3 The wavenumber coordinates S of the corresponding characteristic spectral position in the middle p The corresponding standard feature vector V s The number of arrows pointing in opposite directions, i.e., one arrow pointing upwards and the other downwards; where n F +n N =N-1, where N is the length of the eigenvector.

[0075] (2)n F >n N At that time, K2 = +1;

[0076] (3)n F ≤n N At that time, K2 = -1;

[0077] Wherein, K1 represents the coefficient of the number of positive and negative matches of the multi-feature spectrum, with a value ranging from 0 to 1; K2 represents the direction coefficient of the positive and negative matches of the multi-feature spectrum, with a value of +1 or -1.

[0078] If K1 ≤ threshold, it is determined as "no target gas" and jumps to step eight; if K1 > threshold, continue to step six.

[0079] It is worth noting that for semi-transparent media such as gases, opposite spectral radiation / absorption characteristics will be exhibited under high and low temperature backgrounds. Under high temperature backgrounds, absorption is observed, while under low temperature backgrounds, radiation is observed. The peaks and valleys of the spectral curves are in opposite directions. Figure 5 As shown. Therefore, in the process of calculating K1, the number of positive matches n was taken. F The number of negative matches n N The larger of the two, instead of just the number of forward matches n. F .

[0080] It is also worth noting that the characteristic spectral direction of random noise is random. Taking an eigenvector length of 10 as an example, the probabilities of matching the standard spectrum with different numbers of positive and negative signals are as follows:

[0081]

[0082]

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089]

[0090]

[0091]

[0092] Therefore, it can be seen that the coefficient K1 of positive and negative matching of multi-feature spectrum under pure random noise has a 65.6% probability of not being greater than 0.6 and an 89% probability of not being greater than 0.7. Therefore, setting the threshold range to about 0.6 to 0.7 can eliminate most of the false alarms caused by noise interference.

[0093] Step Six: Calculate the cosine similarity of the feature vectors, keeping the sign of the result but not taking the absolute value, as shown in the following formula:

[0094] K3 = V s ·V d

[0095] Step 7: Based on the results of Steps 5 and 6, calculate the similarity between the gas being tested and the template gas:

[0096] S = K1·K2·K3

[0097] Set threshold S th Gases with a value S greater than the threshold are retained and temporarily listed as potentially present; gases with a value S less than the threshold are discarded and considered not to exist. The template gas is selected from sulfur hexafluoride, sulfur dioxide, Freon R134a, and ammonia.

[0098] It is worth noting that: threshold S th The value range is (0, K1). The value for calculating and identifying low-resolution, high-noise gas spectra is approximately 0.05 to 0.2, which needs to be adjusted according to different measuring instruments and usage scenarios.

[0099] It's also worth noting that since both K2 and K3 have signs, the calculation result S also carries a sign. A negative value indicates that the matching direction of the wavenumber coordinates of the feature spectral position is inconsistent with the similarity direction. This is because the spectral radiance amplitude corresponding to a few feature spectral coordinates, or even just one feature spectral coordinate, accounts for too large a proportion in the calculation of the cosine similarity of the feature vector K3, thus skewing the entire calculation result. The similarity S is corrected in time by the direction coefficient K2, avoiding false alarms caused by excessively large spectral radiance amplitudes corresponding to abnormal feature spectral coordinates.

[0100] It needs to be further explained that: such as Figure 6 and Figure 7 As shown, a standard characteristic spectrum curve and a measured curve are presented. The standard spectrum curve has one large peak and several smaller peaks. One large peak in the measured curve basically corresponds to one large peak in the standard spectrum curve, while the smaller peaks in the measured curve are not obvious. If only the cosine similarity of the eigenvectors K3 = V is used... s ·V dIn this judgment, small peaks are ineffective and prone to misjudgment; relying solely on the coefficient K1 of positive and negative matching of multiple feature spectra is also susceptible to random noise interference; only by using S = K1·K2·K3, and combining three judgments, can the accuracy be improved.

[0101] Step 8: Change the suspected gas type (i+1) in this measurement, repeat steps 2 to 7, obtain multiple similarity scores S, arrange them in order of size, and determine the gas corresponding to the maximum similarity score as the type detected this time.

[0102] It is worth noting that the similarity S can characterize the relative concentration of the gas, displayed through pseudo-color and numerical values.

[0103] Furthermore, in certain fixed locations where the type of suspected gas is clearly identified, focusing only on whether there is a leak of that gas, steps one through seven can be calculated, and only the presence and relative concentration of the gas of concern can be reported.

[0104] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A gas identification method suitable for low-resolution, high-noise spectral data, characterized in that, Includes the following steps: Step 1: Preprocess the original spectral curve of the gas to be measured; preprocessing includes: baseline removal, smoothing to remove high-frequency noise, and normalization; Step 2: Based on the specific scenario of this measurement, select a group of M possible gases from the gas spectral feature library as suspected gases, and extract the spectral feature template of the i-th gas from this group of suspected gases; i = 1, 2, 3, ..., M; the spectral features are two arrays of length N, the first array being the wavenumber coordinates S of the characteristic spectral position. p The second array is the wavenumber coordinates of the characteristic spectral position S. p The corresponding normalized spectral radiance value A s N is the number of characteristic spectra, N = 1, 2, 3...; Step 3: Find the wavenumber coordinates S of the characteristic spectral positions of the gas spectrum obtained in Step 2 from the spectral curve processed in Step 1. p The corresponding spectral radiation value at that location generates an array A of length N. d ; Step 4: Array A s and array A d Each component performs a difference operation to generate a standard feature vector V. s and the characteristic vector V of the measured gas d The length of each vector is N-1; Step 5: Calculate the number coefficients of positive and negative matches in the multi-feature spectrum and the direction coefficients of positive and negative matches in the multi-feature spectrum; if the number coefficients of positive and negative matches in the multi-feature spectrum are less than or equal to the threshold, determine "no target gas" and skip to step 8; if the number coefficients of positive and negative matches in the multi-feature spectrum are greater than the threshold, continue to step 6; Step 6: Calculate the cosine similarity of the feature vectors, keeping the sign of the result but not taking the absolute value; Step 7: Based on the results of Step 5 and Step 6, calculate the similarity S between the gas being tested and the template gas; the template gas is selected from sulfur hexafluoride, sulfur dioxide, Freon R134a, and ammonia. Step 8: Change the suspected gas type that is the focus of this measurement, repeat steps 2 to 7, obtain multiple similarity scores S, arrange them in order of magnitude, and determine the gas type corresponding to the maximum similarity score.

2. The gas identification method for low-resolution, high-noise spectral data as described in claim 1, characterized in that, In step two, the gas spectral feature library obtains spectral feature data of known gases through experimental measurements or by querying standard gas spectral feature databases in the industry.

3. The gas identification method for low-resolution, high-noise spectral data as described in claim 2, characterized in that, In step two, if the types of gases that may appear in the measurement scenario are uncertain, all gas types in the gas spectral feature library are selected to participate in subsequent calculations.

4. The gas identification method for low-resolution, high-noise spectral data as described in claim 3, characterized in that, In step five, the coefficient for the number of positive and negative matches in the multi-feature spectrum is: K1 = max(n) F n N )÷(n F +n N ); Where, n F The number of positive matches, i.e., the wavenumber coordinates of the characteristic spectral positions, S. p The corresponding gas feature vector V d Wavenumber coordinates S of the same characteristic spectral position p The corresponding standard feature vector V s The number of arrows pointing in the same direction, i.e., all pointing upwards or all pointing downwards; n N The number of negative matches, i.e., the wavenumber coordinates of the characteristic spectral positions, S. p The corresponding gas feature vector V d Wavenumber coordinates S of the same characteristic spectral position p The corresponding standard feature vector V s The number of arrows pointing in opposite directions, i.e., one arrow pointing upwards and the other downwards; where n F +n N =N-1, where K1 is the length of the feature vector; K1 takes values ​​from 0 to 1.

5. The gas identification method for low-resolution, high-noise spectral data as described in claim 4, characterized in that, In step five, the positive and negative matching direction coefficient K2 of the multi-feature spectrum takes the value of +1 or -1, when n F >n N When n, K2 = +1; when n F ≤n N At that time, K2 = -1.

6. The gas identification method for low-resolution, high-noise spectral data as described in claim 5, characterized in that, In step five, the threshold is set to 0.6 to 0.

7.

7. The gas identification method for low-resolution, high-noise spectral data as described in claim 6, characterized in that, In step six, the cosine similarity of the feature vectors is: K3=V s ·V d 。 8. The gas identification method for low-resolution, high-noise spectral data as described in claim 7, characterized in that, In step seven, the similarity between the gas being tested and the template gas is: S = K1·K2·K3 Set threshold S th S is greater than the threshold S th The presence of this gas is temporarily listed as possible; S is less than the threshold S. th The omission of this term indicates that the gas does not exist.

9. The gas identification method for low-resolution, high-noise spectral data as described in claim 8, characterized in that, Threshold S th The range of values ​​for is (0, K1).

10. The application of a gas identification method based on any one of claims 1-9 for low-resolution, high-noise spectral data in the field of chemical gas detection and identification technology.

Images