A method and system for spectral line screening

Abstract

This invention provides a method and system for screening spectral analysis lines. The method includes: acquiring plasma spectra based on a laser probe, calculating the corresponding average spectrum using this as the original spectrum, and identifying spectral peaks in the average spectrum through wavelet transform; extracting predetermined features of the spectral peaks from the original spectrum, and classifying and scanning each spectral peak based on the predetermined features; forming a column vector matrix from the classification and scanning results, and calculating the cumulative correlation coefficients between the column phasors in the column vector matrix; arranging the cumulative correlation coefficients in ascending order, and selecting the spectral peaks corresponding to the highest-ranking cumulative correlation coefficients as the analysis line clusters; extracting predetermined features of the analysis line clusters from the original data, and constructing a multi-peak classification model based on the analysis line cluster features using a machine learning algorithm; and screening the spectral analysis lines to be identified using the multi-peak classification model. This approach can reduce the dimensionality of the spectral peak feature vectors, reduce the amount of model data, improve the efficiency of analysis line screening, and ensure classification accuracy.

Description

Technical Field

[0001] This invention belongs to the field of spectral analysis technology, and particularly relates to a method and system for screening spectral lines. Background Technology

[0002] Laser probe microscopy, also known as laser-induced breakdown spectroscopy (LIBS), is widely used in fields such as rock and mineral testing, deep-sea and deep-space exploration, environmental pollution detection, and industrial production. These applications include both qualitative and quantitative analysis. For qualitative analysis using LIBS, it is typically necessary to screen the analytical lines.

[0003] The selection of spectral lines involves first choosing a specific number of analytical lines from the spectrum, then extracting the spectral intensity of the selected lines from the original spectral data. Based on this, specific analytical algorithms such as principal component analysis and support vector machines are used to build clustering or classification models for selection. However, in portable and handheld LIBS systems with relatively limited computing power, to ensure the accuracy of the analysis results and avoid excessive computation and low efficiency in later applications of the built analytical model, the amount of data used for modeling and prediction needs to be as small as possible. Therefore, the selected analytical lines should ideally be few in number and possess strong characterization capabilities.

[0004] Currently, some research has been conducted by those skilled in the art in this area. For example, the published paper "An Automatic Peak Finding Method for LIBS Spectra Based on Continuous Wavelet Transform" (Spectroscopy and Spectral Analysis, Vol. 34, No.7, pp1969-1972) mentions a method for LIBS spectral peak finding using continuous wavelet transform. In this method, after many steps such as continuous wavelet transform optimized based on scale and translation parameters, ridge correction, and signal-to-noise ratio screening, 21 spectral peaks were identified in a segment of the LIBS spectrum of the Fe3O4 sample (382.274 – 385.347 nm, 300 pixels). However, the identification process of this method is relatively complex and parameter optimization is difficult. Since the identified spectral peaks are not further evaluated and utilized, their characterization ability and qualitative and quantitative analysis utility cannot be known. Moreover, when using signal-to-noise ratio for screening, background estimation is required for each wavelength. Manually specifying the background has efficiency problems, while automatic background identification will introduce evaluation errors.

[0005] Meanwhile, full-spectrum classification is also a feasible scheme for line selection and classification. Although the full-spectrum classification method is simple in steps, the large amount of full-spectrum data inevitably leads to low efficiency in the early modeling and later prediction applications of LIBS technology. Moreover, the introduction of too many approximate features in the full-spectrum data can easily cause the non-positive definite problem of the feature matrix, making modeling more difficult.

[0006] Therefore, how to accurately and efficiently select analytical lines with strong characterization capabilities from the original spectrum is a key problem that urgently needs to be solved. Summary of the Invention

[0007] In view of this, embodiments of the present invention provide a method and system for screening spectral analysis lines, which solves the problems of complex process, low accuracy and low efficiency in current analytical line screening.

[0008] In a first aspect of the present invention, a method for screening spectral analysis lines is provided, comprising:

[0009] Plasma spectra of a predetermined number of samples were collected using a laser probe.

[0010] The plasma spectrum is used as the original spectrum. The average spectrum corresponding to the original spectrum is calculated, and the peaks of the average spectrum are identified by wavelet transform.

[0011] Predetermined features of spectral peaks are extracted from the original spectrum, and each spectral peak is classified and scanned based on the predetermined features. The predetermined features are used to characterize the information features of the spectral peaks.

[0012] The classification scan results are arranged into a column vector matrix, and the cumulative correlation coefficient between any column vector in the column vector matrix and all other column phasors is calculated.

[0013] The cumulative correlation coefficients are arranged in ascending order, and the spectral peaks corresponding to the first preset number of cumulative correlation coefficients in the sort are selected as the analysis line clusters;

[0014] Predetermined features of analytical line clusters are extracted from the original spectrum, and a multi-peak classification model is constructed based on the predetermined features of the analytical line clusters using machine learning algorithms.

[0015] The spectral lines to be identified are screened using a multi-peak classification model.

[0016] In a second aspect of the present invention, a spectral analysis line screening system is provided, comprising:

[0017] The spectral acquisition module is used to acquire the plasma spectra of a preset number of samples based on a laser probe;

[0018] The peak identification module is used to take the plasma spectrum as the original spectrum, calculate the average spectrum corresponding to the original spectrum, and identify the peaks of the average spectrum through wavelet transform.

[0019] The spectral peak classification module is used to extract predetermined features of spectral peaks from the original spectrum and to classify and scan each spectral peak based on the predetermined features. The predetermined features are used to characterize the spectral peak information features.

[0020] The peak selection module is used to form a column vector matrix from the classification scan results, calculate the cumulative correlation coefficient between any column vector in the column vector matrix and all other column phasors, sort the cumulative correlation coefficients in ascending order, and select the spectral peaks corresponding to the first preset number of cumulative correlation coefficients as the analysis line clusters.

[0021] The model building module is used to extract predetermined features of analytical line clusters from the raw spectrum and to build a multi-peak classification model based on the predetermined features of the analytical line clusters using machine learning algorithms.

[0022] The analysis line screening module is used to screen the spectral analysis lines to be identified using a multi-peak classification model.

[0023] In a third aspect of the present invention, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor executes the computer program to implement the steps of the method as described in the first aspect of the present invention.

[0024] In a fourth aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method provided in the first aspect of the present invention.

[0025] In this embodiment of the invention, peak features are extracted from the original spectrum to classify and scan each peak. The cumulative correlation coefficient between the peak feature vectors is calculated, and peaks are selected as analytical line clusters based on the cumulative correlation coefficient. A multi-peak classification model is then constructed to screen analytical lines. This not only simplifies the peak feature vectors and reduces the amount of data in the modeling process, but also ensures the accuracy of spectral line screening and improves the computational efficiency of the model. Simultaneously, it avoids the problems of human interference, error introduction, modeling difficulties, and low efficiency in practical applications inherent in traditional spectral line screening. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a schematic flowchart of a spectral analysis line screening method provided in one embodiment of the present invention;

[0028] Figure 2 This is a schematic diagram of the peak identification results provided in one embodiment of the present invention;

[0029] Figure 3 This is a schematic diagram illustrating the calculation of the cumulative correlation coefficient according to an embodiment of the present invention;

[0030] Figure 4 A schematic diagram illustrating the relationship between the number of analysis lines and classification accuracy provided in one embodiment of the present invention;

[0031] Figure 5 This is a schematic diagram of the structure of a spectral analysis line screening system provided in one embodiment of the present invention;

[0032] Figure 6 This is a schematic diagram of the structure of an electronic device provided in one embodiment of the present invention. Detailed Implementation

[0033] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0034] It should be understood that the terms "comprising" and other similar expressions in the specification, claims, and accompanying drawings of this invention are intended to cover a non-exclusive inclusion, such as a process, method, system, or apparatus that includes a series of steps or units and is not limited to the listed steps or units. Furthermore, "first" and "second" are used to distinguish different objects and are not intended to describe a specific order.

[0035] Please see Figure 1 The present invention provides a schematic flowchart of a spectral analysis line screening method, comprising:

[0036] S101. Plasma spectra of a preset number of samples are collected using a laser probe;

[0037] The laser probe, also known as a laser-induced breakdown spectroscopy (LASPS) analyzer, is a material composition analyzer based on LASPS technology. LASPS utilizes a high-energy laser to excite a high-temperature, high-electron-density plasma on the sample surface, and analyzes the ionic or atomic spectra of the emitted light from the plasma. The plasma spectrum is an electromagnetic radiation spectrum emitted from within the plasma, ranging from infrared to vacuum ultraviolet.

[0038] The plasma spectra of a predetermined number of samples are obtained by using a laser probe. For a predetermined number of samples, multiple collection points can be set at different locations on the samples, and multiple spectra can be collected at each collection point to obtain multiple plasma spectra corresponding to each sample.

[0039] For example, spectral acquisition was performed on 12 igneous rock samples, with one sample of each type. Ten points were collected at different locations on each sample, and ten spectra were collected at each point. Each spectrum had 4096 pixels. Therefore, the total number of plasma spectra for the samples was 12 × 10 × 10, and the original spectral data matrix can be represented as N. 4096ⅹ1200 .

[0040] S102. Using the plasma spectrum as the original spectrum, calculate the average spectrum corresponding to the original spectrum, and identify the spectral peaks of the average spectrum through wavelet transform.

[0041] The average spectrum is the result obtained by averaging the spectral data of multiple samples. By averaging the spectral values ​​of multiple samples at various wavelengths, an average spectrum can be obtained, thereby reducing noise, minimizing differences between samples, and extracting common information. Wavelet transform is a mathematical analysis method that decomposes a spectral signal into a series of wavelet functions, thereby analyzing the spectral signal at different scales.

[0042] The average spectrum is obtained by averaging the original spectrum, and then the spectral peak information is obtained by decomposing the average spectrum through wavelet transform.

[0043] For example, such as Figure 2 As shown, the Matlab function mean(N) is used. 4096，1200, 2) Average the original spectral data to obtain the average spectrum, and then use the wavelet transform function cwt(y, scales, 'mexh') to perform wavelet-based peak identification. This identifies 282 peaks in the average spectrum, denoted as P. 282ⅹ2 .

[0044] S103. Extract predetermined features of spectral peaks from the original spectrum, and classify and scan each spectral peak based on the predetermined features of the spectral peaks. The predetermined features are used to characterize the information features of the spectral peaks.

[0045] The classification scan refers to classifying spectral peaks using a classification model and outputting classification labels.

[0046] Specifically, predetermined features of individual wavelength spectral peaks are extracted from the original spectrum to obtain one-dimensional spectral predetermined features corresponding to each spectral peak. After the one-dimensional spectral predetermined features are divided into training set and test set according to the proportion, the single spectral peak classification model is trained and tested. The single spectral peak classification model is a linear discriminant analysis model.

[0047] Based on the trained single-peak classification models, K-fold cross-validation is performed on the spectral data in the test set to obtain the peak classification results of each single-peak classification model for the original spectrum. Each peak category corresponds to a predicted label value.

[0048] The predetermined features include spectral intensity, peak area, peak intensity ratio, or internal standard intensity value of the peak. For example, the spectral intensity and peak intensity ratio of different wavelength peaks can be extracted from the original spectrum.

[0049] For spectral peaks of different wavelengths, corresponding single-peak classification models are established and trained. For example, if there are 282 spectral peaks, 282 single-peak classification models need to be established. Based on the trained single-peak classification models, K-fold cross-validation is performed on the spectral data in the test set to obtain the spectral peak classification results of each single-peak classification model for the original spectrum. Each spectral peak category corresponds to a predicted label value, and the number of these values ​​is the number of spectra of each category in the test set.

[0050] For example, suppose there are N in total. s There are N samples, and the number of sample categories is N. type Then the number of spectra for each type of sample is n / N type At this point, assuming the correct predicted label value for the first type of sample is denoted as 1, the incorrect predicted label values ​​could range from 2 to N. type The numbers in the set are arbitrary, but the total number of predicted label values ​​for each class of samples is a fixed value: .

[0051] The spectral intensities of the 282 obtained wavelengths were extracted sequentially, a single-peak classification model was constructed, and cross-validation was performed. The sample quantity and category N were also considered. s and N type Both are 12, with n / N being the number of spectra per class. type The value is 100, and the ratio of the training set to the test set data of the model is N. train :N test If the ratio is 4:1, then the total number of predicted label values ​​for each type of sample is... .

[0052] S104. Form a column vector matrix from the classification scan results, and calculate the cumulative correlation coefficient between any column vector in the column vector matrix and all other column phasors.

[0053] The classification scan results can be classified by a single-peak classification model and then subjected to K-fold cross-validation to obtain predicted label values. The predicted label values ​​of each round of K-fold cross-validation form a one-dimensional column vector, and the one-dimensional column vectors of multiple rounds of scanning can form a column vector matrix.

[0054] Specifically, the spectral peaks of all wavelengths are sequentially classified and scanned, and the predicted label values ​​of each round of K-fold cross-validation are vertically spliced ​​to obtain a one-dimensional column vector composed of the predicted label values. The column vector matrix is ​​then formed based on the classification and scanning results of all wavelengths.

[0055] Substitute all one-dimensional column vectors into the Pearson correlation coefficient calculation formula in turn to obtain the Pearson correlation coefficient between any column vector and all other column vectors. Then, sum the obtained Pearson correlation coefficients to obtain the cumulative correlation coefficient corresponding to each spectral peak.

[0056] By iteratively substituting all column vectors into the Pearson correlation coefficient formula, the Pearson correlation coefficient R between any given column vector and all other vectors can be obtained. ij (L i , L j The obtained correlation coefficients are then summed to obtain the cumulative correlation coefficient. Ultimately, for each spectral peak, there exists a cumulative correlation coefficient for its corresponding wavelength, which can be represented by the cumulative correlation coefficient column vector C. wⅹ1 For details, please refer to Figure 3 .

[0057] For example, in each round of cross-validation after modeling and classifying for each wavelength, each class of samples can obtain a column vector L of predicted label values. Nsampleⅹ1 = L 20ⅹ1 The L is obtained by vertically concatenating the column vectors of predicted label values ​​for all samples. Nsample*Ntypeⅹ1 = L 240ⅹ1 After further 5-fold cross-validation, the column vector L of all predicted label values ​​corresponding to the wavelength of this spectral peak can be obtained. K*Nsample*Ntypeⅹ1 =L 1200ⅹ1 Finally, the correlation coefficient C between any predicted label value column vector and all other column vectors is calculated using the Pearson correlation coefficient formula. i The cumulative correlation coefficient column vector C is obtained by summing the results. 282ⅹ1 ,like Figure 3 As shown in the figure, for spectral peaks at different wavelengths, the Pearson correlation coefficient between any column vector and all other vectors is calculated. and accumulate. .

[0058] S105. Arrange the cumulative correlation coefficients in ascending order, and select the spectral peaks corresponding to the first preset number of cumulative correlation coefficients in the sort as the analysis line clusters;

[0059] After sorting the cumulative correlation coefficients in ascending order, the earlier the cumulative correlation coefficient appears, the smaller it is, indicating a weaker correlation between the spectral peaks; conversely, the later the cumulative correlation coefficient appears, the larger it is, indicating a stronger correlation between the spectral peaks. Therefore, the spectral peaks corresponding to the highest-ranking cumulative correlation coefficients can be extracted as analytical line clusters.

[0060] Preferably, a relationship curve between the number of spectral peak wavelengths and the classification scanning accuracy is constructed, and the number of wavelengths corresponding to the maximum or stable value of the relationship curve is obtained, and the number of wavelengths is used as the preset number.

[0061] For example, the accuracy of classification for each scan is denoted as CA. i After the scan is completed, the classification accuracy column vector CA can be obtained. 282ⅹ1 Establish the number of wavelengths and CA i The change curve, the relationship curve is as follows Figure 4 As shown, the horizontal axis represents the number of wavelengths, and the vertical axis represents the classification accuracy. In this example, when N max = 91 CA i The value reached its maximum of 94.8%. Therefore, in C... 282ⅹ1 The wavelengths corresponding to the first 91 cumulative correlation coefficients are selected to form the final analysis line cluster.

[0062] S105. Extract the predetermined features of the analytical line clusters from the original spectrum, and construct a multi-peak classification model based on the predetermined features of the analytical line clusters using a machine learning algorithm.

[0063] The analytical line cluster refers to a cluster composed of multiple analytical lines, which are spectral lines selected for qualitative and quantitative analysis in atomic emission spectroscopy. By extracting predetermined features from the analytical line cluster and training a multi-peak classification model using machine learning algorithms, a multi-peak classification model for peak identification is obtained.

[0064] Among them, the machine learning algorithm is the linear discriminant analysis algorithm, the support vector machine algorithm, the neural network algorithm, or the K-nearest neighbor algorithm.

[0065] S106. The spectral lines to be identified are screened using a multi-peak classification model.

[0066] The multi-peak classification model is continuously trained by analyzing the predetermined features of the line clusters of a preset number of samples, and the spectral lines to be identified are then screened based on the trained multi-peak classification model.

[0067] It is understandable that the laser probe classification method using cumulative correlation coefficient scanning for line selection (classification correlation coefficient scanning line selection method) is compared with the traditional classification method (wavelet transform line selection direct classification method): The traditional line selection method, such as direct classification using all wavelengths after wavelet transform peak finding, can achieve a classification accuracy of 93.8% for 12 types of igneous rock samples, but requires 282 analysis lines, and the training and prediction data dimensions are 282 x 960 and 282 x 240, respectively. However, using the method provided in this embodiment, while maintaining the original classification accuracy, only 91 analysis lines are needed, and the training and prediction data dimensions can be reduced to 91 x 960 and 91 x 240, respectively, significantly reducing the data volume and improving computational efficiency.

[0068] In this embodiment, based on the cumulative correlation coefficient, the corresponding spectral peaks are selected as the analysis line clusters, which further simplifies the traditional spectral feature vector, reduces the dimension of the feature vector, significantly reduces the amount of data, and improves classification efficiency. The actual computational efficiency is significantly improved.

[0069] Meanwhile, there is no need to perform tedious optimization of many parameters in the wavelet transform process, such as scaling parameters and translation parameters. It is only necessary to use the transform to roughly identify the more obvious spectral peaks. It can accurately and quantitatively evaluate the target wavelength, reduce the manual intervention process, and thus ensure the reliability and uniqueness of the line selection process and results. It also reduces the amount of feature data and lowers the difficulty of modeling.

[0070] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0071] Figure 5 This is a schematic diagram of a spectral analysis line screening system provided in an embodiment of the present invention. The system includes:

[0072] The spectral acquisition module 510 is used to acquire the plasma spectra of a preset number of samples based on a laser probe;

[0073] The peak identification module 520 is used to take the plasma spectrum as the original spectrum, calculate the average spectrum corresponding to the original spectrum, and identify the peaks of the average spectrum through wavelet transform.

[0074] The spectral peak classification module 530 is used to extract predetermined features of spectral peaks from the original spectrum and to classify and scan each spectral peak based on the predetermined features. The predetermined features are used to characterize the spectral peak information features.

[0075] Specifically, predetermined features of individual wavelength spectral peaks are extracted from the original spectrum to obtain one-dimensional spectral predetermined features corresponding to each spectral peak. After the one-dimensional spectral predetermined features are divided into training set and test set according to the proportion, the single spectral peak classification model is trained and tested. The single spectral peak classification model is a linear discriminant analysis model.

[0076] Based on the trained single-peak classification models, K-fold cross-validation is performed on the spectral data in the test set to obtain the peak classification results of each single-peak classification model for the original spectrum. Each peak category corresponds to a predicted label value.

[0077] The predetermined features are spectral intensity, peak area, peak intensity ratio, or internal standard intensity value of the peak.

[0078] The peak selection module 540 is used to form a column vector matrix from the classification scanning results, calculate the cumulative correlation coefficient between any column vector in the column vector matrix and all other column phasors, arrange the cumulative correlation coefficients in ascending order, and select the spectral peaks corresponding to the first preset number of cumulative correlation coefficients as the analysis line clusters.

[0079] Specifically, the step of constructing a column vector matrix using the classification scan results and calculating the cumulative correlation coefficient between any column vector and all other column phasors includes:

[0080] The spectral peaks of all wavelengths are sequentially classified and scanned. The predicted label values ​​of each round of K-fold cross-validation are vertically concatenated to obtain a one-dimensional column vector composed of the predicted label values. The column vector matrix is ​​formed based on the classification and scanning results of all wavelengths.

[0081] Substitute all one-dimensional column vectors into the Pearson correlation coefficient calculation formula in turn to obtain the Pearson correlation coefficient between any column vector and all other column vectors. Then, sum the obtained Pearson correlation coefficients to obtain the cumulative correlation coefficient corresponding to each spectral peak.

[0082] Preferably, the step of selecting the spectral peaks corresponding to the cumulative correlation coefficients of the top-ranked preset number as the analysis line clusters further includes:

[0083] Construct a curve showing the relationship between the number of spectral peak wavelengths and the classification scanning accuracy, obtain the number of wavelengths corresponding to the maximum or stable value of the curve, and use the number of wavelengths as the preset number.

[0084] The model building module 550 is used to extract predetermined features of analytical line clusters from the original spectrum and to build a multi-peak classification model based on the predetermined features of analytical line clusters using machine learning algorithms.

[0085] The analysis line screening module 560 is used to screen the spectral analysis lines to be identified using a multi-peak classification model.

[0086] The machine learning algorithm is a linear discriminant analysis algorithm, a support vector machine algorithm, a neural network algorithm, or a K-nearest neighbor algorithm.

[0087] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system and modules described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0088] Figure 6 This is a schematic diagram of an electronic device according to an embodiment of the present invention. The electronic device is used for spectral analysis line screening. Figure 6 As shown, the electronic device 6 of this embodiment includes: a memory 610, a processor 620, and a system bus 630. The memory 610 includes an executable program 6101 stored thereon. As those skilled in the art will understand, Figure 6 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0089] The following is combined with Figure 6 A detailed introduction to each component of the electronic device:

[0090] The memory 610 can be used to store software programs and modules. The processor 620 executes various functional applications and data processing of the electronic device by running the software programs and modules stored in the memory 610. The memory 610 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device (such as cached data), etc. In addition, the memory 610 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0091] The memory 610 contains an executable program 6101 with a network request method. This executable program 6101 can be divided into one or more modules / units, which are stored in the memory 610 and executed by the processor 620 to perform tasks such as spectral line screening. Each module / unit can be a series of computer program instruction segments capable of performing a specific function, describing the execution process of the computer program 6101 in the electronic device 6. For example, the computer program 6101 can be divided into functional modules such as a spectral acquisition module, a peak identification module, a peak classification module, a peak selection module, a model construction module, and an analytical line screening module.

[0092] The processor 620 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines. By running or executing software programs and / or modules stored in the memory 610, and by calling data stored in the memory 610, it performs various functions and processes data, thereby monitoring the overall status of the electronic device. Optionally, the processor 620 may include one or more processing units; preferably, the processor 620 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, application programs, etc., and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 620.

[0093] The system bus 630 is used to connect various functional components within the computer, transmitting data, address, and control information. Its type can be, for example, a PCI bus, an ISA bus, or a CAN bus. Instructions from the processor 620 are transmitted to the memory 610 via the bus, and the memory 610 sends data back to the processor 620. The system bus 630 handles the data and instruction exchange between the processor 620 and the memory 610. Of course, the system bus 630 can also connect to other devices, such as network interfaces and display devices.

[0094] In this embodiment of the invention, the executable program executed by the processing 620 of the electronic device includes:

[0095] Plasma spectra of a predetermined number of samples were collected using a laser probe.

[0096] The plasma spectrum is used as the original spectrum. The average spectrum corresponding to the original spectrum is calculated, and the peaks of the average spectrum are identified by wavelet transform.

[0097] Predetermined features of spectral peaks are extracted from the original spectrum, and each spectral peak is classified and scanned based on the predetermined features. The predetermined features are used to characterize the information features of the spectral peaks.

[0098] The classification scan results are arranged into a column vector matrix, and the cumulative correlation coefficient between any column vector in the column vector matrix and all other column phasors is calculated.

[0099] The cumulative correlation coefficients are arranged in ascending order, and the spectral peaks corresponding to the first preset number of cumulative correlation coefficients in the sort are selected as the analysis line clusters;

[0100] Predetermined features of analytical line clusters are extracted from the raw data, and a multi-peak classification model is constructed based on the predetermined features of analytical line clusters using machine learning algorithms.

[0101] The spectral lines to be identified are screened using a multi-peak classification model.

[0102] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0103] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0104] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for screening spectral analysis lines, characterized in that, include: Plasma spectra of a predetermined number of samples were collected using a laser probe. The plasma spectrum is used as the original spectrum. The average spectrum corresponding to the original spectrum is calculated, and the peaks of the average spectrum are identified by wavelet transform. Predetermined features of spectral peaks are extracted from the original spectrum, and each spectral peak is classified and scanned based on the predetermined features. The predetermined features are used to characterize the information features of the spectral peaks. The classification scan results are arranged into a column vector matrix, and the cumulative correlation coefficient between any column vector in the column vector matrix and all other column phasors is calculated. The cumulative correlation coefficients are arranged in ascending order, and the spectral peaks corresponding to the first preset number of cumulative correlation coefficients in the sort are selected as the analysis line clusters; Predetermined features of analytical line clusters are extracted from the original spectrum, and a multi-peak classification model is constructed based on the predetermined features of the analytical line clusters using machine learning algorithms. The spectral lines to be identified are screened using a multi-peak classification model.

2. The method according to claim 1, characterized in that, The step of extracting predetermined features of spectral peaks from the original spectrum and classifying and scanning each spectral peak based on the predetermined features includes: Predetermined features of individual wavelength spectral peaks are extracted from the original spectrum to obtain one-dimensional spectral predetermined features corresponding to each spectral peak. After dividing the one-dimensional spectral predetermined features into training and testing sets according to a certain ratio, the single-peak classification model is trained and tested. The single-peak classification model is a linear discriminant analysis model. Based on the trained single-peak classification models, K-fold cross-validation is performed on the spectral data in the test set to obtain the peak classification results of each single-peak classification model for the original spectrum. Each peak category corresponds to a predicted label value.

3. The method according to claim 2, characterized in that, The predetermined features are spectral intensity, peak area, peak intensity ratio, or peak internal standard intensity value.

4. The method according to claim 2, characterized in that, The step of assembling the classification scan results into a column vector matrix and calculating the cumulative correlation coefficient between any column vector in the column vector matrix and all other column phasors includes: The spectral peaks of all wavelengths are sequentially classified and scanned. The predicted label values ​​of each round of K-fold cross-validation are vertically concatenated to obtain a one-dimensional column vector composed of the predicted label values. The column vector matrix is ​​formed based on the classification and scanning results of all wavelengths. Substitute all one-dimensional column vectors into the Pearson correlation coefficient calculation formula in turn to obtain the Pearson correlation coefficient between any column vector and all other column vectors. Then, sum the obtained Pearson correlation coefficients to obtain the cumulative correlation coefficient corresponding to each spectral peak.

5. The method according to claim 1, characterized in that, The spectral peaks corresponding to the cumulative correlation coefficients of the selected, pre-defined number of peaks in the ranking, as the analytical line clusters, also include: Construct a curve showing the relationship between the number of spectral peak wavelengths and the classification scanning accuracy, obtain the number of wavelengths corresponding to the maximum or stable value of the curve, and use the number of wavelengths as the preset number.

6. The method according to claim 1, characterized in that, The machine learning algorithm is a linear discriminant analysis algorithm, a support vector machine algorithm, a neural network algorithm, or a K-nearest neighbor algorithm.

7. A spectral analysis line screening system, characterized in that, include: The spectral acquisition module is used to acquire the plasma spectra of a preset number of samples based on a laser probe; The peak identification module is used to take the plasma spectrum as the original spectrum, calculate the average spectrum corresponding to the original spectrum, and identify the peaks of the average spectrum through wavelet transform. The spectral peak classification module is used to extract predetermined features of spectral peaks from the original spectrum and to classify and scan each spectral peak based on the predetermined features. The predetermined features are used to characterize the spectral peak information features. The peak selection module is used to form a column vector matrix from the classification scan results, calculate the cumulative correlation coefficient between any column vector in the column vector matrix and all other column phasors, sort the cumulative correlation coefficients in ascending order, and select the spectral peaks corresponding to the first preset number of cumulative correlation coefficients as the analysis line clusters. The model building module is used to extract predetermined features of analytical line clusters from the raw spectrum and to build a multi-peak classification model based on the predetermined features of the analytical line clusters using machine learning algorithms. The analysis line screening module is used to screen the spectral analysis lines to be identified using a multi-peak classification model.

8. The system according to claim 7, characterized in that, The machine learning algorithm is a linear discriminant analysis algorithm, a support vector machine algorithm, a neural network algorithm, or a K-nearest neighbor algorithm.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of a spectral analysis line screening method as described in any one of claims 1 to 6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed, it implements the steps of a spectral analysis line screening method as described in any one of claims 1 to 6.

Images