Infrared spectrum chemometrics analysis method and system

Abstract

The invention discloses an infrared spectrum chemometrics analysis method and system, and relates to the technical field of infrared spectrum analysis. Comprising the following steps: acquiring a plurality of infrared spectrum images with known spectral characteristic parameters, and preprocessing the infrared spectrum images to obtain a training sample image set; based on the training sample image set, establishing a neural network prediction model, taking the infrared spectral image as input and the spectral characteristic parameters as labels, and training to obtain a spectral characteristic parameter extraction model; acquiring an infrared spectrum image of an object to be detected, preprocessing the infrared spectrum image, inputting the preprocessed infrared spectrum image into the trained model, outputting spectrum characteristic parameters, and acquiring environmental parameters; correcting the spectral characteristic parameters based on the environmental parameters, and further correcting the spectral characteristic parameters in combination with the physical characteristic parameters of the to-be-detected article to obtain an accurate value; and completing the stoichiometric analysis of the property and content of the to-be-detected article based on the accurate spectral characteristic parameters. The true values of the spectral characteristic parameters are accurately simulated, and the accuracy of an analysis result is remarkably improved.

Description

Technical Field

[0001] The invention relates to the technical field of infrared spectrum analysis, in particular to an infrared spectrum chemometric analysis method and system. Background Art

[0002] As a fast and non-destructive analytical tool, infrared spectroscopy is widely used in chemistry, materials science, biomedicine and other fields to study the molecular structure and chemical composition of substances. By analyzing the characteristic parameters of the spectrum (such as peak position, peak intensity and integrated area), researchers can obtain the chemical and physical properties of the sample. However, in practical applications, infrared spectroscopy is limited by many technical problems, which makes accurate analysis more difficult.

[0003] First, infrared spectral signals are more easily affected by experimental conditions and environmental factors, such as ambient temperature, humidity, light intensity and other factors during the acquisition process. These parameters will have a significant impact on the spectral signal. For example, changes in ambient temperature may cause the response of the infrared detector to be unstable, affecting the peak position and intensity of the spectrum; differences in light intensity may introduce additional background noise or baseline drift, interfering with the authenticity of the spectral signal. The interference of these external environmental factors makes it difficult for traditional spectral analysis methods to directly obtain accurate spectral characteristic parameters, and additional correction processing must be performed. However, most of the existing correction methods are simple linear models that fail to fully consider the complex nonlinear relationship between environmental parameters and spectral characteristics, resulting in low correction accuracy.

[0004] Secondly, the characteristic parameters of infrared spectroscopy are not only affected by environmental factors, but also significantly affected by the physical properties of the sample itself (such as particle size distribution, surface roughness and thickness). For example, the particle size of particulate matter will affect the scattering and transmission behavior of light, resulting in peak intensity deviation or peak position shift; changes in sample surface roughness will enhance the diffuse reflection of light and weaken the intensity of the spectral signal; differences in thickness directly affect the absorption intensity of light, which may lead to significant changes in the spectral integral area. Existing infrared spectroscopy analysis methods usually ignore the influence of these sample physical properties, or only correct them through simple empirical formulas, resulting in insufficient accuracy and robustness of the analysis results.

[0005] In summary, the existing infrared spectroscopy analysis methods still have obvious shortcomings in dealing with environmental interference, the influence of sample physical properties, and the extraction of complex spectral features. These problems limit the widespread application of infrared spectroscopy technology in the field of high-precision analysis. Therefore, there is an urgent need for an infrared spectroscopy chemometric analysis method that can comprehensively consider environmental parameters and sample physical properties to improve the accuracy of spectral characteristic parameters and the reliability of analysis results.

[0006] In the prior art, the publication number CN110909470B discloses an infrared spectroscopy chemometric analysis system and method, including: a model building module, which is used to establish a sample set according to the sample set file pre-stored in the system, and generate a quantitative model and a qualitative model corresponding to the sample properties for the samples in the sample set, and bind the sample and several quantitative models corresponding to the sample, and bind the sample set and several qualitative models and quantitative models corresponding to the sample set; a model transfer module, which is used to generate a spectral correction factor F for correcting the spectrum from the sub-spectrometer to the mother spectrometer spectrum, and evaluate the accuracy of F; an analysis and prediction module, which transfers the unknown spectral sample collected by the sub-spectrometer to the mother spectrometer through the spectral correction factor, and predicts the sample properties and corresponding sample property values ​​contained in the spectrum transferred to the mother spectrometer through the qualitative model and quantitative model. During the spectrum acquisition process, factors such as ambient temperature, humidity, and light intensity may cause significant interference to the spectral signal, but the scheme does not specifically deal with the influence of such environmental parameters. Ignoring environmental factors may lead to instability of prediction results, especially in complex practical application scenarios. As a result, the accuracy and effectiveness of the analysis system are reduced.

[0007] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0008] The object of the present invention is to provide an infrared spectroscopic chemometric analysis method and system to solve the problems raised in the above background technology.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] An infrared spectroscopic chemometric analysis method, the specific steps comprising:

[0011] Acquire a number of infrared spectral images with known spectral characteristic parameters, and preprocess the acquired infrared spectral images to obtain a training sample image set, wherein the preprocessing includes baseline correction preprocessing and noise removal preprocessing, and the spectral characteristic parameters include peak position, peak intensity and integrated area;

[0012] Based on the obtained training sample image set, a neural network prediction model is established, the infrared spectrum images in the training sample image set are used as inputs of the neural network prediction model, and the corresponding spectral feature parameters are used as labels to train the neural network prediction model to obtain a spectral feature parameter extraction model;

[0013] Acquire an infrared spectrum image of the object to be detected, pre-process the infrared spectrum image of the object to be detected, input the image into a trained spectrum feature parameter extraction model, output the spectrum feature parameters of the object to be detected, and simultaneously acquire environmental parameters when acquiring the infrared spectrum image of the object to be detected, wherein the environmental parameters include the ambient temperature of the acquisition location and the light intensity of the acquisition location;

[0014] Based on the collected environmental parameters, the output spectral characteristic parameters of the object to be detected are corrected to obtain environmental correction characteristic parameters, the physical property parameters of the object to be detected are analyzed, and the environmental correction characteristic parameters are corrected based on the obtained physical property parameters to obtain precise values ​​of the spectral characteristic parameters of the object to be detected. According to the obtained precise values ​​of the spectral characteristic parameters of the object to be detected, the chemometric analysis of the properties and contents of the object to be detected is completed, and the physical property parameters of the object to be detected include the mean particle size, surface roughness and thickness of the object to be detected.

[0015] Further, the collected infrared spectrum image is preprocessed to obtain a training sample image set, wherein the preprocessing includes baseline correction preprocessing and noise removal preprocessing, wherein a polynomial fitting method is selected for baseline correction, and the specific operation steps include: fitting a polynomial by a least squares method to obtain a baseline parameter, and then subtracting the baseline signal from the original spectrum data to obtain a corrected spectrum;

[0016] The noise removal preprocessing adopts the ensemble empirical mode decomposition method, and the specific steps include: adding noise of different amplitudes to the original spectrum, performing several empirical mode decompositions, averaging the IMF components obtained by several decompositions, retaining the low-frequency IMF components, and filtering out the high-frequency components as noise;

[0017] The method for generating the training sample image set is as follows: mapping the preprocessed infrared spectrum image with the spectrum characteristic parameters corresponding to the image one by one to form a corresponding grid, and recording the formed grid as the training sample image set.

[0018] Furthermore, based on the obtained training sample image set, a neural network prediction model is established, in which the long short-term memory network model LSTM model is selected as the base model, and the activation function and optimization algorithm are selected, in which the Tanh function is selected as the activation function, and Adam is selected as the optimization algorithm of the LSTM model; the formula of the Tanh function is:

[0019]

[0020] In the formula, Represents the Tanh function, independent variable Represents the weighted sum of the neuron's input, that is, the result of weighted summation of the input received by the neuron from the previous layer;

[0021] At the same time, the hyperparameters of the LSTM model are set, and the hyperparameters of the LSTM model include: the number of network layers, the number of iterations, the learning rate, the batch size, the number of training times, the batch processing number, and the number of hidden layer neurons;

[0022] The number of network layers is set to 3 layers, the number of iterations is set to 200, the learning rate is set to 0.001, the batch size is set to 32, the number of training times is set to 100, the batch size is set to 256, and the number of hidden layer neurons is 32;

[0023] The input of the trained spectral feature parameter extraction model is the preprocessed infrared spectral image, and the output is the spectral feature parameters in the infrared spectral image, including peak position, peak intensity and integrated area.

[0024] Furthermore, based on the collected environmental parameters, the output spectral characteristic parameters of the object to be detected are corrected to obtain environmental correction characteristic parameters, wherein the peak position is corrected by the ambient temperature to obtain the peak position environmental correction value, wherein the peak position environmental correction value is calculated based on the formula:

[0025]

[0026] In the formula, is the peak position environmental correction value of the infrared spectrum image of the object to be detected, is the predicted value of the peak position of the object to be detected output by the spectral feature parameter extraction model, is the ambient temperature of the collection site, The reference temperature for infrared spectrum image acquisition.

[0027] Furthermore, the peak intensity and the integrated area are corrected by the ambient temperature and the light intensity of the collection site to obtain the peak intensity environmental correction value and the integrated area environmental correction value, wherein the formula for calculating the peak intensity environmental correction value is:

[0028]

[0029] In the formula, is the peak intensity environmental correction value, is the peak intensity prediction value of the object to be detected output by the spectral feature parameter extraction model, is the light intensity at the collection site, Light intensity collected for infrared spectral images;

[0030] The formula for calculating the integrated area environmental correction value is:

[0031]

[0032] In the formula, is the integrated area environmental correction value, It is the predicted value of the integrated area of ​​the object to be detected output by the spectral feature parameter extraction model.

[0033] Furthermore, the environmental correction characteristic parameters are corrected based on the obtained physical characteristic parameters to obtain the precise value of the spectral characteristic parameters of the object to be detected, wherein the formula for calculating the precise value of the peak position is:

[0034]

[0035] In the formula, is the exact value of the peak position, is the particle size distribution correction index, is the surface roughness correction index, and are the weight coefficients of particle size distribution correction index and surface roughness correction index, respectively, where and and All are greater than 0;

[0036] The formula for calculating the exact value of the peak intensity is:

[0037]

[0038] In the formula, is the exact value of the peak intensity, is the thickness correction index, is the weight coefficient of the sum of the particle size distribution correction index and the surface roughness correction index, is the weight coefficient of the thickness correction index, where and and All are greater than 0;

[0039] The formula for calculating the exact value of the integral area is:

[0040]

[0041] In the formula, is the exact value of the integrated area, is the weight coefficient of the sum of the particle size distribution correction index and the surface roughness correction index, is the weight coefficient of the thickness correction index, where and and Both are greater than 0.

[0042] Furthermore, the specific formulas for calculating the particle size distribution correction index, the surface roughness correction index and the thickness correction index are as follows:

[0043]

[0044]

[0045]

[0046] In the formula, is the average diameter of the particles of the objects to be detected, is the wavelength of the incident infrared light, is the surface roughness of the object to be tested, is the thickness of the object to be tested;

[0047] Among them, the surface roughness of the object to be tested The calculation is based on the formula:

[0048]

[0049]

[0050] In the formula, represents the sampling length of the stylus profilometer in the pth sampling area, p is the index of the sampling area, and , P is the number of sampling areas, is the height of the xth sampling point within the sampling length of the pth sampling area from the center line, x is the coordinate of the sampling point within the sampling length, is the surface roughness of the pth sampling area, is the surface roughness, center line and The data is obtained through the built-in software of the stylus profilometer.

[0051] The present invention also provides an infrared spectroscopy chemometrics analysis system, which is used to perform the above-mentioned infrared spectroscopy chemometrics analysis method, comprising:

[0052] A sample image processing module is used to obtain infrared spectrum images with known spectral characteristic parameters, and preprocess the collected infrared spectrum images to obtain a training sample image set, wherein the preprocessing includes baseline correction preprocessing and noise removal preprocessing, and the spectral characteristic parameters include peak position, peak intensity and integrated area;

[0053] The feature extraction model training module is used to establish a neural network prediction model based on the obtained training sample image set, use the infrared spectrum image in the training sample image set as the input of the neural network prediction model, and use the corresponding spectral feature parameters as labels to train the neural network prediction model to obtain a spectral feature parameter extraction model;

[0054] An environment correction module is used to obtain an infrared spectrum image of the object to be detected, pre-process the infrared spectrum image of the object to be detected, input the image into the trained spectrum feature parameter extraction model, output the spectrum feature parameters of the object to be detected, and simultaneously collect the environment parameters when the infrared spectrum image of the object to be detected is obtained, wherein the environment parameters include the ambient temperature of the collection location and the light intensity of the collection location;

[0055] The precise characteristic value analysis module is used to correct the output spectral characteristic parameters of the object to be detected based on the collected environmental parameters to obtain environmental correction characteristic parameters, analyze the physical characteristic parameters of the object to be detected, correct the environmental correction characteristic parameters based on the obtained physical characteristic parameters to obtain precise values ​​of the spectral characteristic parameters of the object to be detected, and complete the chemometric analysis of the properties and content of the object to be detected based on the obtained precise values ​​of the spectral characteristic parameters of the object to be detected, wherein the physical characteristic parameters of the object to be detected include the mean particle size, surface roughness and thickness of the object to be detected.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] First, this scheme significantly reduces the spectral distortion caused by spectral baseline drift and environmental noise by preprocessing the collected infrared spectral images, such as baseline correction and noise removal. This preprocessing process provides clearer and more accurate training data for subsequent neural network modeling, reduces the deviation in the model training process, and improves the extraction accuracy of spectral characteristic parameters including peak position, peak intensity and integral area. In addition, by establishing a neural network prediction model and training based on a large number of training sample spectral images and their known characteristic parameters, the complex nonlinear relationships and characteristic patterns in the spectral data are automatically captured, greatly improving the accuracy and robustness of characteristic parameter extraction. Secondly, this scheme innovatively incorporates environmental parameters including ambient temperature and light intensity of the collection site into the correction process of spectral characteristic parameters. Finally, combined with the physical properties of the sample to be tested (including particle size mean, surface roughness and thickness), the spectral characteristic parameters after environmental correction are secondary corrected. The influence of the physical properties of the sample on the spectral characteristic parameters is fully considered, including the nonlinear modulation of the particle size distribution on the peak intensity and peak position, the scattering effect of the surface roughness on the spectral signal, and the complex influence of thickness difference on the light absorption intensity. Through the correction process of comprehensive physical characteristic parameters, this scheme can more accurately simulate the true values ​​of spectral characteristic parameters and significantly improve the accuracy of analysis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0059] Figure 2 It is a schematic diagram of the overall system structure of the present invention. DETAILED DESCRIPTION

[0060] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0061] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0062] Example:

[0063] See also Figure 1 , the present invention provides a technical solution:

[0064] An infrared spectroscopic chemometric analysis method, the specific steps comprising:

[0065] Step 1: Acquire several infrared spectral images with known spectral characteristic parameters, and preprocess the acquired infrared spectral images to obtain a training sample image set, wherein the preprocessing includes baseline correction preprocessing and noise removal preprocessing, and the spectral characteristic parameters include peak position, peak intensity and integrated area.

[0066] The collected infrared spectrum images are preprocessed to obtain a training sample image set, wherein the preprocessing includes baseline correction preprocessing and noise removal preprocessing, wherein a polynomial fitting method is selected for baseline correction.

[0067] Polynomial fitting is a commonly used spectral baseline correction method, which is suitable for removing baseline drift in infrared spectral data. In infrared spectral analysis, due to instrument noise, environmental conditions or the physical properties of the sample itself, the baseline of the spectral signal may drift or change nonlinearly, resulting in reduced extraction accuracy of spectral characteristic parameters (such as peak position, peak intensity, etc.). Through the polynomial fitting method, the baseline can be effectively fitted and subtracted from the spectral data to restore the true spectral signal.

[0068] Get the original spectral data that needs to be corrected for the baseline, including the spectral wavelength and the corresponding spectral intensity. Spectral data is usually stored in the form of a two-dimensional array, where each data point consists of a wavelength and an intensity value. Select the order of the fitting polynomial; select the base point line, the baseline point is the area in the spectrum that is not strongly affected by the signal characteristic peak, and usually represents the true position of the baseline; at the selected baseline point, use the least squares method to fit the baseline polynomial; subtract the fitted baseline function from the original spectral intensity to obtain the corrected spectrum.

[0069] The specific method of the spectral characteristic parameters includes: using an algorithm (such as the first-order derivative method) to find the local maximum of the spectral curve, and the peak position is the wavelength or wave number corresponding to the maximum point;

[0070] Through spectral analysis software or programming tools, such as Origin, MATLAB or LabVIEW, directly take the intensity corresponding to the local maximum point in the spectrum curve, which is the peak intensity;

[0071] Common methods for numerical integration within a specific wavelength range or wavenumber range include: Trapezoidal method: divide the integration area into small trapezoids and add them one by one. The integration range is usually the full width half height of the peak or a specific wavelength / wavenumber window to obtain the integral area characteristic parameters.

[0072] Among them, the noise removal preprocessing is carried out using the ensemble empirical mode decomposition method. The specific steps include: adding noise of different amplitudes to the original spectrum, performing several empirical mode decompositions, averaging the IMF components obtained from several decompositions, retaining the low-frequency IMF components, and filtering out the high-frequency components as noise; the polynomial fitting method and the ensemble empirical mode decomposition method are both conventional technical means, and the detailed operation steps are not repeated here.

[0073] The method for generating the training sample image set is as follows: mapping the preprocessed infrared spectrum image with the spectrum characteristic parameters corresponding to the image one by one to form a corresponding grid, and recording the formed grid as the training sample image set.

[0074] Step 2: Based on the obtained training sample image set, a neural network prediction model is established. The infrared spectrum images in the training sample image set are used as the input of the neural network prediction model, and the corresponding spectral feature parameters are used as labels to train the neural network prediction model to obtain the spectral feature parameter extraction model.

[0075] Based on the obtained training sample image set, a neural network prediction model is established, in which the long short-term memory network model LSTM model is selected as the base model, and the activation function and optimization algorithm are selected. The Tanh function is selected as the activation function, and Adam is selected as the optimization algorithm of the LSTM model; the formula of the Tanh function is:

[0076]

[0077] In the formula, Represents the Tanh function, independent variable Represents the weighted sum of the neuron's input, that is, the result of weighted summation of the input received by the neuron from the previous layer;

[0078] At the same time, the hyperparameters of the LSTM model are set, and the hyperparameters of the LSTM model include: the number of network layers, the number of iterations, the learning rate, the batch size, the number of training times, the batch processing number, and the number of hidden layer neurons;

[0079] The number of network layers is set to 3 layers, the number of iterations is set to 200, the learning rate is set to 0.001, the batch size is set to 32, the number of training times is set to 100, the batch size is set to 256, and the number of hidden layer neurons is 32;

[0080] The input of the trained spectral feature parameter extraction model is the preprocessed infrared spectral image, and the output is the spectral feature parameters in the infrared spectral image, including peak position, peak intensity and integrated area.

[0081] Infrared spectral data often contain complex characteristic peaks and nonlinear changes such as baseline drift. LSTM can capture these complex features through its gating mechanism and accurately extract effective information from the spectrum. LSTM does not need to explicitly define artificial features such as characteristic peak position or width. Instead, it automatically learns the inherent laws of spectral features through training to reduce the impact of human intervention. The LSTM model can not only accurately extract the core features of the spectrum, but also has a certain robustness to noise or outliers through its dynamic memory mechanism, significantly improving the prediction accuracy of spectral feature parameters. Through the introduction of the LSTM model, the intelligent and automated spectral analysis can be realized, and the entire process from data input to result output can be completed without human intervention, which is suitable for efficient industrial production or rapid laboratory analysis.

[0082] Step 3: Obtain an infrared spectral image of the object to be detected, preprocess the infrared spectral image of the object to be detected, input it into the trained spectral feature parameter extraction model, output the spectral feature parameters of the object to be detected, and at the same time collect the environmental parameters when obtaining the infrared spectral image of the object to be detected, the environmental parameters include the ambient temperature of the collection site and the light intensity of the collection site.

[0083] Based on the collected environmental parameters, the output spectral characteristic parameters of the object to be detected are corrected to obtain environmental correction characteristic parameters, wherein the peak position is corrected by the ambient temperature to obtain the peak position environmental correction value, wherein the peak position environmental correction value is calculated based on the formula:

[0084]

[0085] In the formula, is the peak position environmental correction value of the infrared spectrum image of the object to be detected, is the predicted value of the peak position of the object to be detected output by the spectral feature parameter extraction model, is the ambient temperature of the collection site, The reference temperature for infrared spectrum image acquisition.

[0086] The peak position of the infrared spectrum is closely related to the molecular vibration characteristics of the object to be detected and the environmental conditions. The change of ambient temperature will affect the position of the spectrum peak through the following mechanisms: when the temperature rises, the molecular thermal motion intensifies, which may cause the peak position to drift to a lower frequency (in the direction of longer wavelength); when the temperature drops, the thermal motion weakens, and the peak may move to a higher frequency (in the direction of shorter wavelength), where Reflects the proportional relationship between the current ambient temperature and the reference temperature. , indicating that the ambient temperature is higher than the reference temperature, otherwise it indicates that the ambient temperature is lower than the reference temperature.

[0087] The correctness of the correction direction is ensured: when the temperature rises, the peak value may move to the low frequency and need to be corrected to increase; when the temperature decreases, the peak value may move to the high frequency and need to be corrected to decrease. Indicates the relative magnitude of temperature change and distinguishes the direction of effect of temperature increase or decrease.

[0088] The peak intensity and integrated area are corrected by the ambient temperature and the light intensity of the collection site to obtain the peak intensity environmental correction value and the integrated area environmental correction value. The formula for calculating the peak intensity environmental correction value is:

[0089]

[0090] In the formula, is the peak intensity environmental correction value, is the peak intensity prediction value of the object to be detected output by the spectral feature parameter extraction model, is the light intensity at the collection site, The light intensity collected for infrared spectral images.

[0091] Changes in ambient temperature will change the activity of molecular vibration and rotation. As the temperature rises, the thermal motion of the molecules increases, which may lead to a decrease in the peak intensity of the absorption spectrum; when the temperature decreases, the thermal motion decreases and the peak intensity may increase. The logarithmic function grows slowly and is suitable for describing the nonlinear effect of temperature change on peak intensity. That is, the greater the temperature change, the smaller the effect on peak intensity, which is consistent with actual physical phenomena. Since temperature increase usually leads to a decrease in peak intensity, Achieve a positive correction of the peak intensity as the temperature increases.

[0092] The light intensity at the collection site will affect the signal-to-noise ratio of the detection instrument. The background noise of the spectrum may increase due to the superposition of scattered light or reflected light, thereby affecting the accuracy of the peak intensity. Indicates the difference between the current collected light intensity and the reference light intensity, directly reflecting the impact of changes in light intensity on peak intensity. Exponential function It has the characteristic of rapid growth and is suitable for describing the nonlinear effect of light intensity changes on peak intensity. That is, when the light intensity changes slightly, the effect is relatively limited, while when the light intensity difference is large, the effect will be significantly amplified. Since the increase in light intensity usually increases the background noise or aggravates the nonlinear drift of the instrument response, Achieve negative correction of peak intensity when light intensity increases.

[0093] The formula for calculating the integrated area environmental correction value is:

[0094]

[0095] In the formula, is the integrated area environmental correction value, It is the predicted value of the integrated area of ​​the object to be detected output by the spectral feature parameter extraction model.

[0096] As the temperature rises, the thermal vibration of molecules or atoms inside the substance increases, which may cause changes in the peak intensity, position and width of certain absorption peaks in the spectrum. In most cases, the increase in temperature will lead to an increase in the width of the absorption peak, because the thermal vibration of molecules or lattices makes the distribution of different vibrational energy levels or transition energy levels wider. In the absorption spectrum, the increase in temperature may lead to the generation of more excited states, thereby increasing the absorption intensity and thus increasing the integrated area, so it should be calculated by the ratio of temperature. To achieve a negative correction of the integral area, the formula is subtracted This item shows that the higher the temperature, the smaller the corrected integral area The smaller it is, the closer it is to the actual feature.

[0097] The effect of light intensity on the integrated area is similar to that of peak intensity. An increase in light intensity usually causes the integrated area data to be too high. The square of the intensity difference is nonlinearly scaled, so by subtracting To perform correction and compensate for the error introduced by changes in light intensity.

[0098] Step 4: Based on the collected environmental parameters, the output spectral characteristic parameters of the object to be detected are corrected to obtain environmental correction characteristic parameters, the physical property parameters of the object to be detected are analyzed, and the environmental correction characteristic parameters are corrected based on the obtained physical property parameters to obtain precise values ​​of the spectral characteristic parameters of the object to be detected. According to the obtained precise values ​​of the spectral characteristic parameters of the object to be detected, the chemometric analysis of the properties and content of the object to be detected is completed, and the physical property parameters of the object to be detected include the mean particle size, surface roughness and thickness of the object to be detected.

[0099] Based on the obtained physical characteristic parameters, the environmental correction characteristic parameters are corrected to obtain the precise value of the spectral characteristic parameters of the object to be detected, where the formula for calculating the precise value of the peak position is:

[0100]

[0101] In the formula, is the exact value of the peak position, is the particle size distribution correction index, is the surface roughness correction index, and are the weight coefficients of particle size distribution correction index and surface roughness correction index, respectively, where and and All are greater than 0;

[0102] When the particle size distribution becomes larger (particle size increases), the scattering behavior changes from Rayleigh scattering (particle size is much smaller than the wavelength) to Mie scattering (particle size is close to the wavelength). That is, the increase in particle size will cause the peak position in the spectrum to redshift (move toward the lower frequency direction). Therefore, the peak position accuracy should be positively compensated by indicates positive compensation. As the particle size distribution correction index increases, The increase of The amplitude of the precise value of the peak position gradually increases.

[0103] When the surface roughness increases, the diffuse reflection component of light increases and the specular reflection component decreases. This change has little effect on the peak position of the spectrum, but in some cases, it may cause a slight blue shift of the peak position (moving toward higher frequencies). Therefore, negative compensation should be performed on the peak position accuracy value. represents positive compensation. The square root function makes the correction term When it is larger, the change is faster, and When it is smaller, the change is slower.

[0104] The particle size directly affects the scattering behavior of light. Especially in the spectral measurement dominated by light scattering and absorption, the change of particle size will significantly change the scattering mode (such as Rayleigh scattering or Mie scattering), resulting in a significant shift in the spectral peak. As the spectral peak position increases, it may change significantly, so a higher weight needs to be given. and and Both are greater than 0.

[0105] The formula for calculating the exact value of the peak intensity is:

[0106]

[0107] In the formula, is the exact value of the peak intensity, is the thickness correction index, is the weight coefficient of the sum of the particle size distribution correction index and the surface roughness correction index, is the weight coefficient of the thickness correction index, where and and All are greater than 0;

[0108] The larger the particle size, the stronger the scattering, and more light is scattered and cannot enter the detector, causing the peak intensity to gradually decrease. Therefore, a positive correction is required to make it closer to the true value.

[0109] The greater the surface roughness, the more significant the effect of the surface microstructure on the scattering and reflection of light, causing part of the light to be scattered or reflected and not enter the detector, thereby reducing the peak intensity. Therefore, the particle size distribution correction index and the surface roughness correction index have a synergistic effect on the peak intensity. It represents the correction of peak intensity due to particle size distribution and surface roughness. Their influence is usually the scattering, reflection and absorption of light, which leads to a decrease in peak intensity, so positive compensation correction is needed.

[0110] The thicker the sample, the longer the light propagation path in the sample, the more significant the absorption effect, resulting in a decrease in the intensity of the transmitted light, and thus a decrease in the peak intensity of the spectrum. The effect of sample thickness is usually significant, especially in spectral measurements, where changes in thickness can cause exponential attenuation of light intensity. Peak intensity accurate value Complete positive correction.

[0111] The change of thickness directly affects the propagation path length of light, thus having a significant impact on light absorption. Although the changes of particle size distribution and surface roughness can affect light scattering and reflection, their influence is usually local. and and Both are greater than 0.

[0112] The formula for calculating the exact value of the integral area is:

[0113]

[0114] In the formula, is the exact value of the integrated area, is the weight coefficient of the sum of the particle size distribution correction index and the surface roughness correction index, is the weight coefficient of the thickness correction index, where and and Both are greater than 0.

[0115] The thicker the sample, the longer the propagation path of light inside the sample, the stronger the light absorption, resulting in a decrease in the integrated area of ​​the transmitted light in the spectrum. Need to be positively corrected, the effect of thickness on light transmittance is usually nonlinear, showing that as the thickness increases, the rate of absorption increase gradually slows down, so through the logarithm Reflects nonlinearity.

[0116] The larger the particle size, the stronger the light scattering. More light is scattered outside the sample or away from the detector, resulting in a decrease in the light intensity received by the detector, thereby reducing the integral area. The greater the surface roughness, the stronger the scattering and reflection effects, and part of the light is not measured, resulting in a decrease in the integral area. Therefore, the square root function is used. It reflects the nonlinear forward compensation effect of light scattering and surface roughness on the integral area.

[0117] The change of thickness directly affects the length of the light propagation path, thus significantly changing the light absorption intensity. The change of particle size distribution and surface roughness only affects the scattering and reflection of light. Although it will also change the integral area, its effect is usually secondary. and and Both are greater than 0.

[0118] The specific formulas for calculating the particle size distribution correction index, surface roughness correction index and thickness correction index are as follows:

[0119]

[0120]

[0121]

[0122] In the formula, is the average diameter of the particles of the objects to be detected, is the wavelength of the incident infrared light, is the surface roughness of the object to be tested, is the thickness of the object to be tested;

[0123] The object to be tested is divided into multiple sampling areas, and the surface roughness of the object to be tested is represented by the average surface roughness of the multiple sampling areas. The calculation is based on the formula:

[0124]

[0125]

[0126] In the formula, represents the sampling length of the stylus profilometer in the pth sampling area, p is the index of the sampling area, and , P is the number of sampling areas, is the height of the xth sampling point within the sampling length of the pth sampling area from the center line, x is the coordinate of the sampling point within the sampling length, is the surface roughness of the pth sampling area, is the surface roughness, center line and The data is obtained through the built-in software of the stylus profilometer.

[0127] See also Figure 2 The present invention also provides an infrared spectroscopy chemometrics analysis system, which is used to perform the above-mentioned infrared spectroscopy chemometrics analysis method, comprising:

[0128] A sample image processing module is used to obtain infrared spectrum images with known spectral characteristic parameters, and preprocess the collected infrared spectrum images to obtain a training sample image set, wherein the preprocessing includes baseline correction preprocessing and noise removal preprocessing, and the spectral characteristic parameters include peak position, peak intensity and integrated area;

[0129] The feature extraction model training module is used to establish a neural network prediction model based on the obtained training sample image set, use the infrared spectrum image in the training sample image set as the input of the neural network prediction model, and use the corresponding spectral feature parameters as labels to train the neural network prediction model to obtain a spectral feature parameter extraction model;

[0130] An environment correction module is used to obtain an infrared spectrum image of the object to be detected, pre-process the infrared spectrum image of the object to be detected, input the image into the trained spectrum feature parameter extraction model, output the spectrum feature parameters of the object to be detected, and simultaneously collect the environment parameters when the infrared spectrum image of the object to be detected is obtained, wherein the environment parameters include the ambient temperature of the collection location and the light intensity of the collection location;

[0131] The precise characteristic value analysis module is used to correct the output spectral characteristic parameters of the object to be detected based on the collected environmental parameters to obtain environmental correction characteristic parameters, analyze the physical characteristic parameters of the object to be detected, correct the environmental correction characteristic parameters based on the obtained physical characteristic parameters to obtain precise values ​​of the spectral characteristic parameters of the object to be detected, and complete the chemometric analysis of the properties and content of the object to be detected based on the obtained precise values ​​of the spectral characteristic parameters of the object to be detected, wherein the physical characteristic parameters of the object to be detected include the mean particle size, surface roughness and thickness of the object to be detected.

[0132] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0133] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0134] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0135] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. An infrared spectroscopic chemometric analysis method, characterized in that: The specific steps include: Acquire a number of infrared spectral images with known spectral characteristic parameters, and preprocess the acquired infrared spectral images to obtain a training sample image set, wherein the preprocessing includes baseline correction preprocessing and noise removal preprocessing, and the spectral characteristic parameters include peak position, peak intensity and integrated area; Based on the obtained training sample image set, a neural network prediction model is established, the infrared spectrum images in the training sample image set are used as inputs of the neural network prediction model, and the corresponding spectral feature parameters are used as labels to train the neural network prediction model to obtain a spectral feature parameter extraction model; Acquire an infrared spectrum image of the object to be detected, pre-process the infrared spectrum image of the object to be detected, input the image into a trained spectrum feature parameter extraction model, output the spectrum feature parameters of the object to be detected, and simultaneously acquire environmental parameters when acquiring the infrared spectrum image of the object to be detected, wherein the environmental parameters include the ambient temperature of the acquisition location and the light intensity of the acquisition location; Based on the collected environmental parameters, the output spectral characteristic parameters of the object to be detected are corrected to obtain environmental correction characteristic parameters, the physical property parameters of the object to be detected are analyzed, and the environmental correction characteristic parameters are corrected based on the obtained physical property parameters to obtain precise values ​​of the spectral characteristic parameters of the object to be detected. According to the obtained precise values ​​of the spectral characteristic parameters of the object to be detected, the chemometric analysis of the properties and contents of the object to be detected is completed, and the physical property parameters of the object to be detected include the mean particle size, surface roughness and thickness of the object to be detected.

2. The infrared spectroscopic chemometric analysis method according to claim 1, characterized in that: The collected infrared spectrum images are preprocessed to obtain a training sample image set, wherein the preprocessing includes baseline correction preprocessing and noise removal preprocessing, wherein a polynomial fitting method is selected for baseline correction, and the specific operation steps include: fitting a polynomial by a least square method to obtain a baseline parameter, and then subtracting the baseline signal from the original spectrum data to obtain a corrected spectrum; The noise removal preprocessing adopts the ensemble empirical mode decomposition method, and the specific steps include: adding noise of different amplitudes to the original spectrum, performing several empirical mode decompositions, averaging the IMF components obtained by several decompositions, retaining the low-frequency IMF components, and filtering out the high-frequency components as noise; The method for generating the training sample image set is as follows: mapping the preprocessed infrared spectrum image with the spectrum characteristic parameters corresponding to the image one by one to form a corresponding grid, and recording the formed grid as the training sample image set.

3. The infrared spectroscopic chemometric analysis method according to claim 2, characterized in that: Based on the obtained training sample image set, a neural network prediction model is established, in which the long short-term memory network model LSTM model is selected as the base model, and the activation function and optimization algorithm are selected. The Tanh function is selected as the activation function, and Adam is selected as the optimization algorithm of the LSTM model; the formula of the Tanh function is: In the formula, Represents the Tanh function, independent variable Represents the weighted sum of the neuron's input, that is, the result of weighted summation of the input received by the neuron from the previous layer; At the same time, the hyperparameters of the LSTM model are set, and the hyperparameters of the LSTM model include: the number of network layers, the number of iterations, the learning rate, the batch size, the number of training times, the batch processing number, and the number of hidden layer neurons; The number of network layers is set to 3 layers, the number of iterations is set to 200, the learning rate is set to 0.001, the batch size is set to 32, the number of training times is set to 100, the batch size is set to 256, and the number of hidden layer neurons is 32; The input of the trained spectral feature parameter extraction model is the preprocessed infrared spectral image, and the output is the spectral feature parameters in the infrared spectral image, including peak position, peak intensity and integrated area.

4. The infrared spectroscopic chemometric analysis method according to claim 3, characterized in that: Based on the collected environmental parameters, the output spectral characteristic parameters of the object to be detected are corrected to obtain environmental correction characteristic parameters, wherein the peak position is corrected by the ambient temperature to obtain the peak position environmental correction value, wherein the peak position environmental correction value is calculated based on the formula: In the formula, is the peak position environmental correction value of the infrared spectrum image of the object to be detected, is the predicted value of the peak position of the object to be detected output by the spectral feature parameter extraction model, is the ambient temperature of the collection site, The reference temperature for infrared spectrum image acquisition.

5. The infrared spectroscopic chemometric analysis method according to claim 4, characterized in that: The peak intensity and integrated area are corrected by the ambient temperature and the light intensity of the collection site to obtain the peak intensity environmental correction value and the integrated area environmental correction value. The formula for calculating the peak intensity environmental correction value is: In the formula, is the peak intensity environmental correction value, is the peak intensity prediction value of the object to be detected output by the spectral feature parameter extraction model, is the light intensity at the collection site, Light intensity collected for infrared spectral images; The formula for calculating the integrated area environmental correction value is: In the formula, is the integrated area environmental correction value, It is the predicted value of the integrated area of ​​the object to be detected output by the spectral feature parameter extraction model.

6. The infrared spectroscopic chemometric analysis method according to claim 5, characterized in that: Based on the obtained physical characteristic parameters, the environmental correction characteristic parameters are corrected to obtain the precise value of the spectral characteristic parameters of the object to be detected, where the formula for calculating the precise value of the peak position is: In the formula, is the exact value of the peak position, is the particle size distribution correction index, is the surface roughness correction index, and are the weight coefficients of particle size distribution correction index and surface roughness correction index, respectively, where and and All are greater than 0; The formula for calculating the exact value of the peak intensity is: In the formula, is the exact value of the peak intensity, is the thickness correction index, is the weight coefficient of the sum of the particle size distribution correction index and the surface roughness correction index, is the weight coefficient of the thickness correction index, where and and All are greater than 0; The formula for calculating the exact value of the integral area is: In the formula, is the exact value of the integrated area, is the weight coefficient of the sum of the particle size distribution correction index and the surface roughness correction index, is the weight coefficient of the thickness correction index, where and and Both are greater than 0.

7. The infrared spectroscopic chemometric analysis method according to claim 6, characterized in that: The specific formulas for calculating the particle size distribution correction index, surface roughness correction index and thickness correction index are as follows: In the formula, is the average diameter of the particles of the objects to be detected, is the wavelength of the incident infrared light, is the surface roughness of the object to be tested, is the thickness of the object to be tested; Among them, the surface roughness of the object to be tested The calculation is based on the formula: In the formula, represents the sampling length of the stylus profilometer in the pth sampling area, p is the index of the sampling area, and , P is the number of sampling areas, is the height of the xth sampling point within the sampling length of the pth sampling area from the center line, x is the coordinate of the sampling point within the sampling length, is the surface roughness of the pth sampling area, is the surface roughness, center line and The data is obtained through the built-in software of the stylus profilometer.

8. An infrared spectroscopic chemometric analysis system, characterized in that: The infrared spectroscopy chemometrics analysis system is used to perform the infrared spectroscopy chemometrics analysis method according to any one of claims 1 to 7, comprising: A sample image processing module is used to obtain infrared spectrum images with known spectral characteristic parameters, and preprocess the collected infrared spectrum images to obtain a training sample image set, wherein the preprocessing includes baseline correction preprocessing and noise removal preprocessing, and the spectral characteristic parameters include peak position, peak intensity and integrated area; The feature extraction model training module is used to establish a neural network prediction model based on the obtained training sample image set, use the infrared spectrum image in the training sample image set as the input of the neural network prediction model, and use the corresponding spectral feature parameters as labels to train the neural network prediction model to obtain a spectral feature parameter extraction model; An environment correction module is used to obtain an infrared spectrum image of the object to be detected, pre-process the infrared spectrum image of the object to be detected, input the image into the trained spectrum feature parameter extraction model, output the spectrum feature parameters of the object to be detected, and simultaneously collect the environment parameters when the infrared spectrum image of the object to be detected is obtained, wherein the environment parameters include the ambient temperature of the collection location and the light intensity of the collection location; The precise characteristic value analysis module is used to correct the output spectral characteristic parameters of the object to be detected based on the collected environmental parameters to obtain environmental correction characteristic parameters, analyze the physical characteristic parameters of the object to be detected, correct the environmental correction characteristic parameters based on the obtained physical characteristic parameters to obtain precise values ​​of the spectral characteristic parameters of the object to be detected, and complete the chemometric analysis of the properties and content of the object to be detected based on the obtained precise values ​​of the spectral characteristic parameters of the object to be detected, wherein the physical characteristic parameters of the object to be detected include the mean particle size, surface roughness and thickness of the object to be detected.

Images