A raman spectroscopy liquid package decoupling method

Abstract

This invention relates to a Raman spectroscopy decoupling method for liquid packaging, belonging to the field of hazardous liquid safety inspection technology, and solves the problem of severe interference caused by existing liquid packaging to the Raman spectra of liquids. The decoupling method includes: collecting Raman spectral samples and performing standardized preprocessing on each sample; clustering all collected Raman spectral samples using a fuzzy clustering algorithm to obtain a predetermined number of clusters; filtering out Raman spectral samples within each cluster using a support vector machine algorithm to obtain pure Raman spectral samples for each cluster; determining the liquid type to which each cluster belongs based on the pure Raman spectral samples, thereby obtaining all pure Raman spectral samples corresponding to each liquid. This reduces the influence of liquid packaging on the Raman spectra of the liquid.

Description

Technical Field

[0001] This invention relates to the field of hazardous liquid security inspection technology, and in particular to a Raman spectroscopy-based liquid packaging decoupling method. Background Technology

[0002] Raman spectroscopy is a type of molecular vibrational spectroscopy that can reflect the molecular fingerprint characteristics of a substance. It can be used for substance detection and identification. Raman spectroscopy detection methods detect the spectrum formed by the Raman scattering effect produced when an object is irradiated with excitation light, thereby enabling the detection and identification of substances. It has been widely used in fields such as liquid security inspection, jewelry inspection, explosives detection, drug detection, pharmaceutical inspection, and pesticide residue detection.

[0003] In liquid security inspection scenarios, hazardous liquids, such as gasoline, alcohol, and acetone, are often contained in various packaging containers, such as plastic bottles, glass bottles, and metal cans. When Raman spectroscopy is used to detect liquids inside the packaging, the excitation light needs to penetrate the packaging material to illuminate the liquid sample. However, the packaging material itself also generates Raman scattering signals when illuminated by the excitation light. These signals from the packaging are superimposed on the Raman signals of the liquid, resulting in a significant difference between the collected spectrum and the spectrum of the liquid itself, thus causing inaccurate identification of the liquid components.

[0004] To address the problem of packaging interference, existing technologies have proposed several solutions. For example, adjusting the laser power or selecting a specific wavelength of excitation light (such as 1064nm near-infrared laser) can reduce fluorescence interference from packaging materials to some extent. Spatial shift Raman spectroscopy is used to collect Raman signals from the laser incident point and its shift position. By utilizing the differences in the contribution of different measurement positions to the signals of the container and its contents, the spectrum of the pure substance can be extracted through proportional subtraction or multivariate statistical methods.

[0005] However, the above solutions still have limitations in practical applications. Adjusting the laser power has limited effect on colored or dark-colored packaging. Although spatially shifted Raman spectroscopy can reduce packaging interference, it has high equipment requirements and is still not ideal for multi-layer composite packaging or strong fluorescent packaging.

[0006] Therefore, there is an urgent need for a technical solution to decouple the Raman spectrum of liquids from their packaging. Summary of the Invention

[0007] Based on the above analysis, the present invention aims to provide a Raman spectroscopy liquid packaging decoupling method to solve the problem that existing liquid packaging causes serious interference to the Raman spectrum of liquids.

[0008] This invention provides a Raman spectroscopy liquid packaging decoupling method, the decoupling method comprising: Raman spectral samples were collected, and each Raman spectral sample underwent standardized preprocessing. All collected Raman spectral samples are clustered using a fuzzy clustering algorithm to obtain a preset number of clusters; The Raman spectral samples within each cluster are filtered out using the support vector machine algorithm to obtain the pure Raman spectral samples for each cluster. Based on the pure Raman spectrum samples of each cluster, the liquid type to which each cluster belongs is determined, and then all pure Raman spectrum samples corresponding to each liquid are obtained.

[0009] Based on the further improvement of the above decoupling method, the clustering of all acquired Raman spectral samples based on the fuzzy clustering algorithm includes: The membership matrix is ​​updated iteratively multiple times based on all collected Raman spectral samples to obtain the final membership matrix; Based on the final membership matrix, the membership degree of each Raman spectral sample to each cluster is determined, and the Raman spectral sample is classified into the cluster with the highest membership degree, thus obtaining the Raman spectral samples included in each cluster.

[0010] Based on the further improvement of the above decoupling method, the membership matrix is ​​updated iteratively multiple times based on all the collected Raman spectral samples to obtain the final membership matrix. The following steps are performed in each iteration: The center of each cluster is calculated based on the membership matrix of the current iteration and all collected Raman spectral samples; The updated membership matrix is ​​calculated using the centers of each cluster and all collected Raman spectral samples. Based on the membership matrices before and after the update, determine whether the iteration termination condition is met. If the iteration termination condition is met, the iteration ends, and the updated membership matrix is ​​used as the final membership matrix. If the iteration termination condition is not met, the next iteration begins, and the updated membership matrix is ​​used as the membership matrix before the update in the next iteration.

[0011] Based on a further improvement to the above decoupling method, the center of each cluster is calculated using the following formula: ; in, Indicates the first The central light intensity vector of each cluster This represents the total number of all Raman spectral samples collected. Represents the fuzzy index. This represents the membership matrix before the update, specifically the first... The Raman spectrum sample belongs to the first The membership degree of each cluster, Indicates the first The light intensity vector corresponding to each Raman spectral sample.

[0012] Based on the further improvement of the above decoupling method, the updated membership matrix is ​​calculated using the following formula: ; in, In the updated membership matrix, the th The Raman spectrum sample belongs to the first The membership degree of each cluster, Indicates the preset quantity. , They represent the first Cluster, the first The central light intensity vector of each cluster Indicates the first The light intensity vector corresponding to the first Raman spectrum sample and the first Raman spectrum sample The Euclidean distance between the central light intensity vectors of each cluster Indicates the first The light intensity vector corresponding to the first Raman spectrum sample and the first Raman spectrum sample The Euclidean distance between the central light intensity vectors of each cluster.

[0013] Based on a further improvement to the above decoupling method, the step of determining whether the iteration termination condition is met based on the membership matrices before and after the update includes: The difference between the membership matrices before and after the update is calculated using the following formula: ; Determine if the difference value is less than or equal to the convergence threshold; if the difference value is less than or equal to the convergence threshold, the iteration termination condition is met; otherwise, the iteration termination condition is not met.

[0014] Based on a further improvement to the above decoupling method, the step of filtering out Raman spectral samples within each cluster using a support vector machine algorithm includes: All Raman spectral samples within each cluster are taken as positive samples of that cluster, while the remaining samples after removing the positive samples of that cluster from all collected Raman spectral samples are taken as negative samples of that cluster. The support vector machine model of the cluster is trained using positive and negative samples to obtain the trained classification model of the cluster. The trained cluster classification model is used to reclassify the positive samples of the cluster, and the reclassified positive samples are taken as the pure Raman spectra of the cluster.

[0015] Based on the further improvement of the above decoupling method, the kernel function of the support vector machine model adopts the radial basis kernel function; The penalty coefficients for positive and negative samples in the support vector machine model are as follows: ; in, , Let represent the penalty coefficients for positive and negative samples in the support vector machine model of this cluster, respectively. This represents the penalty coefficient of the support vector machine model. Indicates the first The number of positive samples in each cluster.

[0016] Based on the further improvement of the above decoupling method, the range of values ​​for the penalty coefficient of the support vector machine model is: ; The range of values ​​for the kernel parameters of the radial basis function is: ; in, The kernel parameters represent the radial basis function kernel.

[0017] Based on a further improvement to the above decoupling method, the standardization preprocessing for each Raman spectrum sample includes: The mean and standard deviation of light intensity in each Raman spectrum sample are calculated using the following formula: , ; in, This represents the average light intensity in each Raman spectrum sample. This represents the standard deviation of light intensity in each Raman spectrum sample. This represents the total number of light intensities in each Raman spectrum sample. Indicates the first Raman spectral sample. Individual light intensity; The light intensity in each Raman spectrum sample is processed using the following formula to obtain the standardized preprocessed light intensity: ; in, Indicates the first Raman spectral sample. The light intensity after standardization preprocessing.

[0018] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects: The collected Raman spectral samples are clustered using a fuzzy clustering algorithm to obtain a preset number of clusters containing Raman spectral samples. Then, a support vector machine algorithm is used to filter out the Raman spectral samples within each cluster, removing samples with the same characteristics from different clusters to obtain the pure Raman spectral samples for each cluster. Finally, based on the pure Raman spectral samples of each cluster, the liquid type of each cluster is determined, accurately extracting all pure Raman spectral samples corresponding to each liquid. This significantly reduces the interference of liquid packaging on the Raman spectra of each liquid, providing a foundation for further improving the accuracy and robustness of hazardous liquid identification.

[0019] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0020] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0021] Figure 1 This is a schematic flowchart of a Raman spectroscopy liquid packaging decoupling method provided in an embodiment of the present invention. Detailed Implementation

[0022] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0023] A specific embodiment of the present invention discloses a Raman spectroscopy liquid packaging decoupling method, such as... Figure 1 As shown, the decoupling method includes: Step S1: Collect Raman spectral samples and perform standardized preprocessing on each Raman spectral sample; Step S2: Cluster all the collected Raman spectral samples based on the fuzzy clustering algorithm to obtain a preset number of clusters; Step S3: Use the support vector machine algorithm to filter out the Raman spectral samples in each cluster to obtain the pure Raman spectral samples of each cluster; Step S4: Determine the liquid type of each cluster based on the pure Raman spectrum sample of each cluster, and then obtain all pure Raman spectrum samples corresponding to each liquid.

[0024] Specifically, such as Figure 1As shown, in step S1, a Raman spectrometer is used to collect Raman spectral data of different liquids and different packaging combinations under the same conditions. In the collected data, the same liquid is packaged in different types of liquid packaging, and sufficient Raman spectral samples are collected under the same conditions.

[0025] For example, suppose there are several types of liquids. Types, packaging types include species, together There are several combination methods, with any combination of packaging and liquid forming one category, totaling [number missing]. Classes, each class collects The Raman spectral data obtained totaled [number] samples. One, of which Any Raman spectrum sample Each element of the array represents the Raman shift in the Raman spectrum sample. The light intensity at that time, among which, .

[0026] Specifically, the range of Raman displacement is Exemplary One can choose 1400, meaning that each Raman spectrum sample includes 1400 light intensity data.

[0027] Specifically, in step S1, each Raman spectrum sample is subjected to standardized preprocessing to unify the range of Raman spectrum data to a similar scale.

[0028] Preferably, the standardization preprocessing for each Raman spectrum sample includes: The mean and standard deviation of light intensity in each Raman spectrum sample are calculated using the following formula: , ; in, This represents the average light intensity in each Raman spectrum sample. This represents the standard deviation of light intensity in each Raman spectrum sample. This represents the total number of light intensities in each Raman spectrum sample. Indicates the first Raman spectral sample. Individual light intensity; The light intensity in each Raman spectrum sample is processed using the following formula to obtain the standardized preprocessed light intensity: ; in, Indicates the first Raman spectral sample. The light intensity after standardization preprocessing.

[0029] Specifically, the Raman spectrum samples are processed using Z-score standardization to eliminate the influence of dimensions and intensity differences. The standardized data conforms to a normal distribution, meaning the mean is 0 and the standard deviation is 1.

[0030] Specifically, for any Raman spectrum sample Through formula Calculate the average light intensity in each Raman spectrum sample and use the formula Calculate the standard deviation of light intensity in each Raman spectrum sample, and then use the formula... For any Raman spectrum sample Perform Z-score standardization to obtain the standardized new samples. The sample has a mean of 0, a standard deviation of 1, and is dimensionless.

[0031] Specifically, all collected Raman spectral samples are subjected to standardized preprocessing to obtain standardized preprocessed Raman spectral samples, and subsequent steps are carried out based on the standardized preprocessed Raman spectral samples.

[0032] Specifically, such as Figure 1 As shown, in step S2, the standardized preprocessed Raman spectral samples obtained in step S1 are clustered based on the fuzzy clustering algorithm, and all the collected Raman spectral samples are divided into a preset number of clusters, each cluster including one or more Raman spectral samples.

[0033] It is worth noting that the preset number of clusters is determined based on the types of liquids and packaging involved, and the fuzziness index of fuzzy clustering... Method 2 can be used, where the clustering process is achieved by iteratively updating the membership matrix.

[0034] Preferably, the clustering of all acquired Raman spectral samples based on the fuzzy clustering algorithm includes: The membership matrix is ​​updated iteratively multiple times based on all collected Raman spectral samples to obtain the final membership matrix; Based on the final membership matrix, the membership degree of each Raman spectral sample to each cluster is determined, and the Raman spectral sample is classified into the cluster with the highest membership degree, thus obtaining the Raman spectral samples included in each cluster.

[0035] Specifically, a membership matrix is ​​randomly initialized, and the size of the membership matrix is... Each row represents the membership degree of any Raman spectral sample on a predetermined number of clusters, and satisfies the condition that the sum of the membership degrees of any Raman spectral sample on the predetermined number of clusters is 1. .

[0036] Specifically, the initial membership matrix is ​​iteratively updated multiple times based on all collected Raman spectral samples to obtain the final membership matrix.

[0037] Preferably, the membership matrix is ​​updated iteratively multiple times based on all collected Raman spectral samples to obtain the final membership matrix, and the following steps are performed in each iteration: The center of each cluster is calculated based on the membership matrix of the current iteration and all collected Raman spectral samples; The updated membership matrix is ​​calculated using the centers of each cluster and all collected Raman spectral samples. Based on the membership matrices before and after the update, determine whether the iteration termination condition is met. If the iteration termination condition is met, the iteration ends, and the updated membership matrix is ​​used as the final membership matrix. If the iteration termination condition is not met, the next iteration begins, and the updated membership matrix is ​​used as the membership matrix before the update in the next iteration.

[0038] Specifically, at the beginning of each iteration, the center of each cluster is calculated based on the membership matrix of the current iteration and all collected Raman spectral samples. The center of each cluster is calculated using the following formula: ; in, Indicates the first The central light intensity vector of each cluster This represents the total number of all Raman spectral samples collected. Represents the fuzzy index. This represents the membership matrix before the update, specifically the first... The Raman spectrum sample belongs to the first The membership degree of each cluster, Indicates the first The light intensity vector corresponding to each Raman spectral sample.

[0039] Specifically, based on the centers of each cluster in the current iteration and in conjunction with all collected Raman spectral samples, the updated membership matrix is ​​determined using the following formula: ; in, In the updated membership matrix, the th The Raman spectrum sample belongs to the first The membership degree of each cluster, Indicates the preset quantity. , They represent the first Cluster, the first The central light intensity vector of each cluster Indicates the first The light intensity vector corresponding to the first Raman spectrum sample and the first Raman spectrum sample The Euclidean distance between the central light intensity vectors of each cluster Indicates the first The light intensity vector corresponding to the first Raman spectrum sample and the first Raman spectrum sample The Euclidean distance between the central light intensity vectors of each cluster.

[0040] Specifically, after obtaining the membership matrix before and after each iteration, it is determined whether the iteration termination condition is met.

[0041] Preferably, the step of determining whether the iteration termination condition is met based on the membership matrix before and after the update includes: The difference between the membership matrices before and after the update is calculated using the following formula: ; Determine if the difference value is less than or equal to the convergence threshold; if the difference value is less than or equal to the convergence threshold, the iteration termination condition is met; otherwise, the iteration termination condition is not met.

[0042] Specifically, through the formula We measure the difference between the membership matrices before and after the update, and then determine whether the difference is less than or equal to a convergence threshold. If the difference is less than or equal to the convergence threshold, the iteration termination condition is met; if the difference is greater than the convergence threshold, the iteration termination condition is not met.

[0043] Preferably, the convergence threshold is selected. .

[0044] Specifically, if the iteration termination condition is met, the iteration ends and the updated membership matrix is ​​used as the final membership matrix; if the iteration termination condition is not met, the next iteration begins and the updated membership matrix is ​​used as the membership matrix before the update in the next iteration.

[0045] Specifically, the membership matrix is ​​updated multiple times through multiple iterations to obtain the final membership matrix.

[0046] Specifically, after obtaining the final membership matrix, the membership degree of each Raman spectral sample to each cluster is determined based on the final membership matrix. The cluster to which the highest membership degree belongs is selected as the cluster to which each Raman spectral sample belongs, thereby classifying all collected Raman spectral samples into a preset number of clusters, with each cluster including one or more Raman spectral samples.

[0047] Specifically, such as Figure 1As shown, in step S3, the support vector machine algorithm is used to filter out one or more Raman spectral samples included in each cluster obtained in step S2, removing Raman spectral samples in each cluster that are easily confused with other clusters, and retaining pure Raman spectral samples, thereby obtaining pure Raman spectral samples for each cluster.

[0048] Preferably, the step of filtering out Raman spectral samples within each cluster using a support vector machine algorithm includes: All Raman spectral samples within each cluster are taken as positive samples of that cluster, while the remaining samples after removing the positive samples of that cluster from all collected Raman spectral samples are taken as negative samples of that cluster. The support vector machine model of the cluster is trained using positive and negative samples to obtain the trained classification model of the cluster. The trained cluster classification model is used to reclassify the positive samples of the cluster, and the reclassified positive samples are taken as the pure Raman spectra of the cluster.

[0049] Specifically, a training dataset is constructed for each cluster. For each cluster, a binary classification training set is constructed. All Raman spectral samples within each cluster are used as positive samples of that cluster and labeled as +1. At the same time, the remaining samples after removing the positive samples of that cluster from all collected Raman spectral samples are used as negative samples of that cluster and labeled as -1.

[0050] Specifically, a support vector machine model is trained for each cluster to obtain the trained classification model for that cluster.

[0051] Preferably, the kernel function of the support vector machine model is a radial basis function. The penalty coefficients for positive and negative samples in the support vector machine model are as follows: ; in, , Let represent the penalty coefficients for positive and negative samples in the support vector machine model of this cluster, respectively. This represents the penalty coefficient of the support vector machine model. Indicates the first The number of positive samples in each cluster.

[0052] Specifically, to further increase the classification accuracy of the cluster classification model, since the number of positive samples in each cluster is much smaller than the number of negative samples, resulting in a class imbalance problem, this invention sets the penalty coefficient for positive samples in the support vector machine model to be [value missing]. The penalty coefficient for counterexamples is .

[0053] Specifically, considering the nonlinear characteristics of Raman spectral samples, this invention uses the radial basis kernel function as the kernel function of the support vector machine model.

[0054] Preferably, the range of values ​​for the penalty coefficient of the support vector machine model is: ; The range of values ​​for the kernel parameters of the radial basis function is: ; in, The kernel parameters represent the radial basis function kernel.

[0055] Specifically, the training of penalty parameters and kernel parameters employs a combination of grid search and K-Fold CrossValidation for parameter optimization, including: Determine the value ranges for the penalty parameter and kernel parameter. The value range for the penalty coefficient is as follows: The range of values ​​for the kernel parameters is: ; The penalty parameters and kernel parameters of each group are trained and validated K times using positive and negative samples of each cluster to obtain the score corresponding to each group of parameters. The penalty parameters and kernel parameters with the highest scores are selected as the optimal parameters. Under the optimal penalty parameter and kernel parameter, the support vector machine model is retrained using positive and negative samples from each cluster until completion, resulting in a cluster classification model for each cluster.

[0056] Specifically, the trained cluster classification model is used to reclassify the positive samples of the cluster. The reclassified positive samples are used as the pure Raman spectral samples of the cluster, while the reclassified negative samples are filtered out to obtain the pure Raman spectral samples of each cluster.

[0057] Specifically, such as Figure 1 As shown, after obtaining one or more pure Raman spectral samples corresponding to each cluster in step S3, the pure Raman spectral samples of each cluster are counted in step S4 to determine the liquid type to which each pure Raman spectral sample belongs, the proportion of each liquid type is counted, and the liquid type with the largest proportion is taken as the liquid type of the cluster, thereby obtaining the liquid type to which each cluster belongs.

[0058] Specifically, in step S4, since the liquid type and packaging type corresponding to each Raman spectral sample are known during the initial data acquisition, the liquid type of each cluster can be determined by majority voting or statistical methods based on the original labels of the samples retained within the cluster. For example, for each cluster, the frequency of occurrence of each type of liquid in its pure Raman spectral samples is counted, and the liquid type with the highest frequency is taken as the liquid belonging to that cluster. Finally, all clusters are merged according to their assigned liquid type to obtain all pure Raman spectral samples corresponding to each liquid. These samples have effectively eliminated packaging interference and can accurately reflect the intrinsic molecular structure information of each liquid, which can be used for subsequent hazardous liquid identification or classification model training.

[0059] Compared with existing technologies, the Raman spectroscopy liquid packaging decoupling method provided in this embodiment of the invention is based on a fuzzy clustering algorithm to cluster all collected Raman spectral samples to obtain a preset number of clusters containing Raman spectral samples. Then, a support vector machine algorithm is used to filter out the Raman spectral samples within each cluster, removing samples with the same characteristics from different clusters to obtain the pure Raman spectral samples of each cluster. Finally, based on the pure Raman spectral samples of each cluster, the liquid type of each cluster is determined, accurately extracting all pure Raman spectral samples corresponding to each liquid. This greatly reduces the interference of liquid packaging on the Raman spectra of each liquid, providing a foundation for further improving the accuracy and robustness of hazardous liquid identification.

[0060] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0061] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for decoupling liquid packaging using Raman spectroscopy, characterized in that, The decoupling method includes: Raman spectral samples were collected, and each Raman spectral sample underwent standardized preprocessing. All collected Raman spectral samples are clustered using a fuzzy clustering algorithm to obtain a preset number of clusters; The Raman spectral samples within each cluster are filtered out using the support vector machine algorithm to obtain the pure Raman spectral samples for each cluster. Based on the pure Raman spectrum samples of each cluster, the liquid type to which each cluster belongs is determined, and then all pure Raman spectrum samples corresponding to each liquid are obtained.

2. The decoupling method according to claim 1, characterized in that, The clustering of all collected Raman spectral samples based on the fuzzy clustering algorithm includes: The membership matrix is ​​updated iteratively multiple times based on all collected Raman spectral samples to obtain the final membership matrix; Based on the final membership matrix, the membership degree of each Raman spectral sample to each cluster is determined, and the Raman spectral sample is classified into the cluster with the highest membership degree, thus obtaining the Raman spectral samples included in each cluster.

3. The decoupling method according to claim 2, characterized in that, The membership matrix is ​​updated iteratively multiple times based on all collected Raman spectral samples to obtain the final membership matrix. The following steps are performed in each iteration: The center of each cluster is calculated based on the membership matrix of the current iteration and all collected Raman spectral samples; The updated membership matrix is ​​calculated using the centers of each cluster and all collected Raman spectral samples. Determine whether the iteration termination condition is met based on the membership matrix before and after the update; If the iteration termination condition is met, the iteration ends, and the updated membership matrix is ​​used as the final membership matrix. If the iteration termination condition is not met, the next iteration begins, and the updated membership matrix is ​​used as the membership matrix before the update in the next iteration.

4. The decoupling method according to claim 3, characterized in that, The center of each cluster is calculated using the following formula: ； in, Indicates the first The central light intensity vector of each cluster This represents the total number of all Raman spectral samples collected. Represents the fuzzy index. This represents the membership matrix before the update, specifically the first... The Raman spectrum sample belongs to the first The membership degree of each cluster, Indicates the first The light intensity vector corresponding to each Raman spectral sample.

5. The decoupling method according to claim 4, characterized in that, The updated membership matrix is ​​calculated using the following formula: ； in, In the updated membership matrix, the th The Raman spectrum sample belongs to the first The membership degree of each cluster, Indicates the preset quantity. , They represent the first Cluster, the first The central light intensity vector of each cluster Indicates the first The light intensity vector corresponding to the first Raman spectrum sample and the first Raman spectrum sample The Euclidean distance between the central light intensity vectors of each cluster Indicates the first The light intensity vector corresponding to the first Raman spectrum sample and the first Raman spectrum sample The Euclidean distance between the central light intensity vectors of each cluster.

6. The decoupling method according to claim 3, characterized in that, The step of determining whether the iteration termination condition is met based on the membership matrices before and after the update includes: The difference between the membership matrices before and after the update is calculated using the following formula: ； Determine if the difference value is less than or equal to the convergence threshold; if the difference value is less than or equal to the convergence threshold, the iteration termination condition is met; otherwise, the iteration termination condition is not met.

7. The decoupling method according to any one of claims 1-6, characterized in that, The step of filtering out Raman spectral samples within each cluster using a support vector machine algorithm includes: All Raman spectral samples within each cluster are taken as positive samples of that cluster, while the remaining samples after removing the positive samples of that cluster from all collected Raman spectral samples are taken as negative samples of that cluster. The support vector machine model of the cluster is trained using positive and negative samples to obtain the trained classification model of the cluster. The trained cluster classification model is used to reclassify the positive samples of the cluster, and the reclassified positive samples are taken as the pure Raman spectra of the cluster.

8. The decoupling method according to claim 7, characterized in that, The kernel function of the support vector machine model is the radial basis kernel function; The penalty coefficients for positive and negative samples in the support vector machine model are as follows: ； in, , Let represent the penalty coefficients for positive and negative samples in the support vector machine model of this cluster, respectively. This represents the penalty coefficient of the support vector machine model. Indicates the first The number of positive samples in each cluster.

9. The decoupling method according to claim 8, characterized in that, The range of values ​​for the penalty coefficient in the support vector machine model is: ; The range of values ​​for the kernel parameters of the radial basis function is: ; in, The kernel parameters represent the radial basis function kernel.

10. The decoupling method according to claim 1, characterized in that, The standardization preprocessing for each Raman spectrum sample includes: The mean and standard deviation of light intensity in each Raman spectrum sample are calculated using the following formula: ， ； in, This represents the average light intensity in each Raman spectrum sample. This represents the standard deviation of light intensity in each Raman spectrum sample. This represents the total number of light intensities in each Raman spectrum sample. Indicates the first Raman spectral sample. Individual light intensity; The light intensity in each Raman spectrum sample is processed using the following formula to obtain the standardized preprocessed light intensity: ； in, Indicates the first Raman spectral sample. The light intensity after standardization preprocessing.

Images