Environmental pollutant identification method based on spectrum fusion

Abstract

The invention discloses an environmental pollutant identification method based on spectrum fusion, and the method comprises the steps: S1, carrying out the scanning of an environment through a spectrum analysis instrument, and obtaining the original data of a spectrum; s2, preprocessing the original spectral data to obtain preprocessed spectral data; s3, performing feature extraction on the preprocessed spectral data to obtain spectral feature data; s4, fusing the spectral feature data to obtain spectral fusion data; s5, constructing an environmental pollutant identification model according to the spectrum fusion data; and S6, identifying and analyzing the environmental pollutants in actual environmental monitoring by using the environmental pollutant identification model. According to the environmental pollutant identification method based on spectrum fusion provided by the invention, an environmental pollutant identification model is constructed by using fusion data after feature fusion, irrelevant variables are eliminated after feature extraction, the influence brought by noise data in original spectrum data is reduced, and the accuracy of environmental pollutant identification is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of environmental pollutant identification, and in particular relates to an environmental pollutant identification method based on spectral fusion. Background Art

[0002] In the field of environmental monitoring, spectral analysis instruments are usually used to collect the spectrum of the environment to be tested to analyze and identify whether there are pollutants. This can not only provide more accurate and real-time detection, but also avoid harm to operators caused by pollutants.

[0003] In the related art, when merging spectral data obtained by multiple spectral analysis instruments, the original data is usually pre-processed and directly spliced ​​together to obtain fused data, and then the fused data is identified. However, the data types, dimensions, and data values ​​of spectral data obtained by different spectral analysis instruments are different, which can easily lead to fusion failure, making it difficult to accurately identify environmental pollutants. Summary of the invention

[0004] The purpose of the present invention is to propose an environmental pollutant identification method based on spectral fusion, so as to solve the problem in the prior art that the original data are directly spliced ​​together after preprocessing to obtain fused data, and then the fused data is identified, which easily leads to fusion failure, making it difficult to accurately identify environmental pollutants.

[0005] To this end, the present invention provides an environmental pollutant identification method based on spectral fusion, comprising:

[0006] S1. Scan the environment through a spectrum analyzer to obtain raw spectrum data;

[0007] S2, preprocessing the raw spectrum data to obtain preprocessed spectrum data;

[0008] S3, performing feature extraction on the preprocessed spectral data to obtain spectral feature data;

[0009] S4, fusing the spectral feature data to obtain spectral fusion data;

[0010] S5, constructing an environmental pollutant identification model according to the spectral fusion data;

[0011] S6. Use the environmental pollutant identification model to identify and analyze environmental pollutants in actual environmental monitoring.

[0012] Preferably, in S1, a plurality of different spectral analysis instruments are used to scan the environment.

[0013] Preferably, the spectral analysis instrument is a combination of one or more of a visible spectrophotometer, an ultraviolet spectrophotometer, a fluorescence spectrophotometer, an atomic absorption spectrometer, an atomic fluorescence spectrometer, an infrared spectrometer, a near-infrared spectrometer, an X-ray fluorescence spectrometer, a photoelectric spectrometer, a laser Raman spectrometer, a plasma emission spectrometer, a grating spectrometer, and an optical fiber spectrometer.

[0014] Preferably, the preprocessing step in S2 includes cleaning, denoising and standardizing the raw spectral data.

[0015] Preferably, the feature extraction step in S3 is to reduce the dimensions of the pre-processed spectral data of different dimensions by a linear discriminant analysis algorithm.

[0016] Preferably, the steps of the linear discriminant analysis algorithm include calculating the intra-class scatter matrix S w and the inter-class scatter matrix S b , using the intra-class scatter matrix S w and the inter-class scatter matrix S b Construct the optimization function J(W).

[0017] Preferably, the optimization function J(W) is

[0018]

[0019] Among them, S b is the inter-class scatter matrix, which calculates the weighted sum of the vector from each class center point to the overall mean and the transpose of the vector; S w is the intra-class scatter matrix, which calculates the variance of sample points within each class; w i is the eigenvector; d is the maximum number of eigenvectors.

[0020] Preferably, the fusion step in S4 is to splice the spectral feature data together.

[0021] Preferably, constructing the environmental pollutant identification model in S5 includes building an RNN recurrent neural network, using spectral fusion data to train the RNN recurrent neural network, and obtaining the environmental pollutant identification model.

[0022] Preferably, the spectral fusion data is divided into a training set and a validation set. The training set is used to train the environmental pollutant recognition model in the RNN recurrent neural network, and the validation set is used to verify the training results to evaluate the performance of the model.

[0023] Beneficial effects:

[0024] 1. The present invention provides an environmental pollutant identification method based on spectral fusion. The original spectral data from different sources are first preprocessed, and then feature extraction is performed, and then feature fusion is performed. The fused data after feature fusion is used to build an environmental pollutant identification model. After feature extraction, irrelevant variables are eliminated, which reduces the impact of noise data in the original spectral data and improves the accuracy of environmental pollutant identification.

[0025] 2. In the present invention, the linear discriminant analysis algorithm is used for feature extraction. The linear discriminant analysis algorithm can not only effectively reduce the dimension of the pre-processed spectral data and simplify the calculation, but also keep the category output after the pre-processed spectral data is reduced in dimension, which helps the environmental pollutant identification model to further classify and identify it. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Figure 1 This is a schematic diagram of the system structure of Example 1 of an environmental pollutant identification method based on spectral fusion in the present invention; DETAILED DESCRIPTION

[0028] The content of the present invention can be more easily understood by selecting the following detailed description of the preferred implementation method of the present invention and the embodiments included. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those of ordinary skill in the art to which the present invention belongs. When there is a conflict, the definition in this specification shall prevail.

[0029] Embodiment 1:

[0030] like Figure 1 A method for identifying environmental pollutants based on spectral fusion is shown, comprising:

[0031] S1. Scan the environment through a spectrum analyzer to obtain raw spectrum data;

[0032] A plurality of different spectral analysis instruments are used to scan the environment. In this embodiment, a visible spectrophotometer, an ultraviolet spectrophotometer and an infrared spectrometer are used to detect the spectrum in the environment.

[0033] S2, preprocessing the original spectrum data to obtain preprocessed spectrum data;

[0034] The preprocessing steps include cleaning, denoising and standardizing the raw spectral data. Cleaning and denoising can improve data quality, remove noise in the data, and make the data more complete. Standardization can eliminate the impact of dimensions on data analysis results and ensure that all features have the same importance in data analysis.

[0035] S3, extracting features from the preprocessed spectral data to obtain spectral feature data;

[0036] Among them, the feature extraction step is to reduce the dimensions of pre-processed spectral data of different dimensions through a linear discriminant analysis algorithm.

[0037] The steps of the Linear Discriminant Analysis algorithm include:

[0038] S31. Calculate the intra-class scatter matrix S w and the inter-class scatter matrix S b ,

[0039] in,

[0040]

[0041] k is the number of categories, x is the chance of a sample of category j, and μ is the population mean vector of all samples.

[0042] S32, using the intra-class scatter matrix S w and the inter-class scatter matrix S b Construct the optimization function J(W).

[0043] The optimization function J(W) is

[0044]

[0045] Among them, S b is the inter-class scatter matrix, which calculates the weighted sum of the vector from each class center point to the overall mean and the transpose of the vector; S w is the intra-class scatter matrix, which calculates the variance of sample points within each class; w i is the eigenvector; d is the maximum number of eigenvectors.

[0046] S33, projecting the preprocessed spectral data into a new feature space to obtain spectral feature data.

[0047] Z=XW

[0048] Among them, Z is the spectral feature data, X is the preprocessed spectral data, and W is the matrix composed of the eigenvectors selected by the optimization function.

[0049] S4, fusing the spectral feature data to obtain spectral fusion data;

[0050] The fusion step is to stitch the spectral feature data together.

[0051] S5. Construct an environmental pollutant identification model based on spectral fusion data;

[0052] Among them, building an environmental pollutant identification model includes building an RNN recurrent neural network, using spectral fusion data to train the RNN recurrent neural network, and obtaining an environmental pollutant identification model.

[0053] The spectral fusion data is divided into a training set and a validation set. The training set is used to train the environmental pollutant recognition model in the RNN recurrent neural network, and the validation set is used to verify the training results to evaluate the performance of the model.

[0054] S6. Use the environmental pollutant identification model to identify and analyze environmental pollutants in actual environmental monitoring.

Claims

1. A method for identifying environmental pollutants based on spectral fusion, characterized in that: include: S1. Scan the environment through a spectrum analyzer to obtain raw spectrum data; S2, preprocessing the raw spectrum data to obtain preprocessed spectrum data; S3, performing feature extraction on the preprocessed spectral data to obtain spectral feature data; S4, fusing the spectral feature data to obtain spectral fusion data; S5, constructing an environmental pollutant identification model according to the spectral fusion data; S6. Use the environmental pollutant identification model to identify and analyze environmental pollutants in actual environmental monitoring.

2. The method for identifying environmental pollutants based on spectral fusion according to claim 1, characterized in that: In S1, several different spectral analysis instruments are used to scan the environment.

3. The method for identifying environmental pollutants based on spectral fusion according to claim 2 is characterized in that: The spectral analysis instrument is a combination of one or more of a visible spectrophotometer, an ultraviolet spectrophotometer, a fluorescence spectrophotometer, an atomic absorption spectrometer, an atomic fluorescence spectrometer, an infrared spectrometer, a near-infrared spectrometer, an X-ray fluorescence spectrometer, a photoelectric spectrometer, a laser Raman spectrometer, a plasma emission spectrometer, a grating spectrometer, and an optical fiber spectrometer.

4. The method for identifying environmental pollutants based on spectral fusion according to claim 1, characterized in that: The preprocessing steps described in S2 include cleaning, denoising and standardization of the raw spectral data.

5. The method for identifying environmental pollutants based on spectral fusion according to claim 1, characterized in that: The feature extraction step described in S3 is to reduce the dimensions of the pre-processed spectral data of different dimensions through a linear discriminant analysis algorithm.

6. The method for identifying environmental pollutants based on spectral fusion according to claim 5, characterized in that: The steps of the linear discriminant analysis algorithm include calculating the intra-class scatter matrix S w and the inter-class scatter matrix S b , using the intra-class scatter matrix s w and the inter-class scatter matrix s b Construct the optimization function J(W).

7. The method for identifying environmental pollutants based on spectral fusion according to claim 6, characterized in that: The optimization function J(W) is in, b is the inter-class scatter matrix, which calculates the weighted sum of the vector from each class center point to the overall mean and the transpose of the vector; S w is the intra-class scatter matrix, which calculates the variance of sample points within each class; w i is the eigenvector; d is the maximum number of eigenvectors.

8. The method for identifying environmental pollutants based on spectral fusion according to claim 7, characterized in that: The fusion step described in S4 is to stitch the spectral feature data together.

9. The method for identifying environmental pollutants based on spectral fusion according to claim 8, characterized in that: The construction of the environmental pollutant identification model described in S5 includes building an RNN recurrent neural network, using spectral fusion data to train the RNN recurrent neural network, and obtaining the environmental pollutant identification model.

10. The method for identifying environmental pollutants based on spectral fusion according to claim 9, characterized in that: The spectral fusion data is divided into a training set and a validation set. The training set is used to train the environmental pollutant recognition model in the RNN recurrent neural network, and the validation set is used to verify the training results and evaluate the performance of the model.

Images