A method and device for predicting tobacco chemical components based on spectral conversion

Abstract

The present invention provides a method and device for predicting tobacco chemical components based on spectral conversion, and belongs to the technical field of tobacco chemical component detection. The present invention establishes a corresponding relationship between different moisture contents and correction parameters, obtains the real moisture content and the first near-infrared spectrum of the tobacco to be tested in an actual scene, and extracts the corresponding correction parameters from the pre-acquired corresponding relationship according to the real moisture content. The first near-infrared spectrum is converted into a second near-infrared spectrum under a standard moisture content by the correction parameters. The second near-infrared spectrum is input into a pre-established near-infrared prediction model, and the content of the chemical components in the tobacco to be tested is output. In this way, rapid measurement can be achieved in an actual scene, thereby improving work efficiency; at the same time, there is no need to establish multiple near-infrared prediction models corresponding to different moisture contents in advance, which reduces the amount of modeling tasks and reduces costs.

Description

Technical Field

[0001] The invention relates to a tobacco chemical component prediction method and device based on spectrum conversion, belonging to the technical field of tobacco chemical component detection. Background Art

[0002] The chemical composition of tobacco is an intrinsic factor that determines the quality of tobacco. For example, the conventional components that are essential in tobacco quality evaluation and formula design: total alkaloids, reducing sugars, total sugars, total nitrogen, total potassium, etc., and the chemical components used to reflect the style and quality of tobacco and cigarette style and quality: tobacco polyphenols, amino acids, non-volatile organic acids, anions and cations, Amadori compounds, solanesol, neophytadiene, etc. Chemical analysis methods are usually used to test these chemical components, such as continuous flow method, ion chromatography, ICP-MS method, amino acid analyzer method, liquid chromatography tandem mass spectrometry, gravimetric method, high performance liquid chromatography and gas chromatography, etc.

[0003] As a fast and non-destructive detection technology for multiple chemical component indicators, near infrared spectroscopy mainly uses the vibration of chemical bonds such as CH, NH, OH, CC, etc. contained in organic matter to detect a variety of tobacco chemical components, such as total alkaloids, water-soluble reducing sugars, total sugars, total nitrogen, potassium, chlorine, some polyphenols, some organic acids, etc. Different near infrared prediction models need to be established for different chemical components. At present, there is no near infrared spectroscopy prediction model that can simultaneously determine 69 chemical components in tobacco.

[0004] In near-infrared spectroscopy analysis, due to the stretching vibration of OH in water, its harmonic peak and combined frequency peak have strong absorption peaks in the near-infrared spectrum (about 6897cm -1 and 5181cm -1 Therefore, it is necessary to adjust the moisture content of the modeled samples in the laboratory to be in the same moisture range as the predicted sample moisture content, thereby reducing the impact of moisture on the accuracy of the prediction results.

[0005] However, in actual work scenarios, such as tobacco leaf purchase and delivery, redrying tobacco strips, and tobacco production, the number of tobacco leaves is huge, and the moisture content of tobacco leaves varies greatly and has a wide range. If the chemical composition of tobacco is determined by the modeling prediction method during the experiment, it is not only necessary to establish a near-infrared prediction model corresponding to different moisture ranges, but the modeling task is large and the cost is high. In addition, it is difficult to predict large quantities of tobacco leaves through experimental methods in the laboratory, otherwise it will bring huge costs and reduce production efficiency. From the above analysis, it can be seen that near-infrared prediction by adjusting the moisture content in the laboratory is not practical, and a method that can quickly determine the chemical composition of batches of tobacco leaves is urgently needed in actual work scenarios to reduce production costs and improve production efficiency. Summary of the invention

[0006] The purpose of the present invention is to provide a tobacco chemical composition prediction method based on spectral conversion, which is used to solve the problems of high cost and low efficiency in the actual scenario of determining the chemical composition of batch tobacco leaves; the purpose of the present invention is also to provide a tobacco chemical composition prediction device based on spectral conversion, which is used to quickly determine the chemical composition of batch tobacco leaves in actual working scenarios such as tobacco leaf purchase and delivery, aging progress, and tobacco production.

[0007] In order to achieve the above object, the present invention provides a method for predicting tobacco chemical components based on spectral conversion, comprising the following steps:

[0008] S1, obtaining a first near infrared spectrum of the tobacco to be tested;

[0009] S2. According to the actual moisture content of the tobacco to be tested, a correction parameter corresponding to the actual moisture content is obtained from the correspondence between different moisture contents and correction parameters obtained in advance;

[0010] S3, performing spectral conversion on the first near infrared spectrum according to the correction parameters obtained in step S2 to obtain a second near infrared spectrum of the tobacco to be tested at a standard moisture content;

[0011] S4, inputting the second near infrared spectrum into a pre-established near infrared prediction model, and outputting the content of chemical components in the tobacco to be tested; the near infrared prediction model is used to represent the corresponding relationship between the near infrared spectrum of tobacco with standard moisture content and the content of chemical components in tobacco with corresponding standard moisture content;

[0012] The correspondence between the different moisture contents and the correction parameters described in step S2 is obtained through the following process: obtaining the near-infrared spectra of tobacco samples with different moisture contents, selecting one of the tobacco samples with different moisture contents as the tobacco sample with standard moisture content, and respectively calculating the correction parameters for converting the near-infrared spectra of the tobacco samples with other moisture contents into the near-infrared spectra of the tobacco sample with standard moisture content, and then establishing the correspondence between the different moisture contents and the correction parameters according to the moisture content and the correction parameters corresponding to the moisture content.

[0013] In actual scenarios such as tobacco purchase, delivery, and processing, the amount of tobacco is large, and the moisture content in different tobaccos is different, which is difficult to measure quickly through experimental methods. If near-infrared prediction models corresponding to different moisture contents are established respectively, the modeling task is large and the cost is high. Therefore, the present invention pre-acquires the near-infrared spectra of tobacco samples with different moisture contents, selects one of the moisture content tobacco samples as the standard moisture content tobacco sample, and calculates the correction parameters when converting the near-infrared spectra of other moisture content tobacco samples to the standard moisture content, thereby obtaining the corresponding relationship between different moisture contents and the correction parameters. In the actual scenario, the real moisture content and the first near-infrared spectrum of the tobacco to be tested are obtained, and the corresponding correction parameters are extracted from the pre-acquired corresponding relationship according to the real moisture content. The first near-infrared spectrum under the real moisture content is converted to the second near-infrared spectrum under the standard moisture content by the correction parameters. The second near-infrared spectrum is input into the pre-established near-infrared prediction model, and the content of the chemical components in the tobacco to be tested is output. In this way, there is no need to establish a near-infrared prediction model corresponding to the actual moisture content in real time in the laboratory, and rapid measurements can be made in actual scenarios, thereby improving work efficiency. At the same time, there is also no need to establish multiple near-infrared prediction models corresponding to different moisture contents in advance, thus reducing the amount of modeling tasks and costs.

[0014] Furthermore, the method for calculating the correction parameters is: respectively splicing the near-infrared spectra of tobacco samples with different moisture contents with the near-infrared spectra of tobacco samples with standard moisture contents to obtain combined spectra corresponding to other different moisture contents; performing principal component analysis on each combined spectrum to obtain a first sub-load matrix corresponding to the current moisture content and a second sub-load matrix corresponding to the standard moisture content, and calculating the difference between the first sub-load matrix and the second sub-load matrix as the correction parameter corresponding to the current moisture content.

[0015] The near infrared spectra of tobacco samples with different moisture contents are different. After selecting tobacco samples with standard moisture contents, the near infrared spectra of tobacco samples with different current moisture contents are spliced ​​end to end with the near infrared spectra of tobacco samples with standard moisture contents to obtain a combined spectrum. Then, principal component analysis is performed on the combined spectrum to obtain a sub-load matrix corresponding to the current moisture content and a sub-load matrix corresponding to the standard moisture content. The difference between the two sub-load matrices is calculated as a correction parameter. When the actual moisture content of the tobacco to be tested is the same as one of the current moisture contents, the corresponding correction parameters can be extracted, and then the first near infrared spectrum is corrected to the second near infrared spectrum, which can reduce the workload of real-time measurement of tobacco chemical components and improve work efficiency.

[0016] Furthermore, the correction parameters are calculated by the following formula:

[0017]

[0018] In the formula, K 当前水分含量 represents the correction parameter corresponding to the current moisture content, X 当前水分含量 Near infrared spectrum of tobacco sample indicating current moisture content, represents the first sub-load matrix corresponding to the current moisture content, represents the second sub-load matrix corresponding to the standard moisture content, (·) + represents the Moore-Penrose generalized inverse of a matrix.

[0019] A set of specific formulas are provided to calculate the correction parameters, which is simple to calculate and convenient for the implementation of the present invention.

[0020] Furthermore, when the corresponding relationship between the different moisture contents and the calibration parameters is established, the tobacco sample with the lowest moisture content is selected as the tobacco sample with the standard moisture content.

[0021] The OH chemical bonds of water in tobacco samples affect the accuracy of near-infrared prediction. Therefore, tobacco samples with lower moisture content are selected as the standard. When extracting and establishing the near-infrared prediction model, tobacco samples with standard moisture content are selected. Compared with the near-infrared prediction model established with tobacco samples with higher moisture content, the accuracy is higher.

[0022] Furthermore, the real moisture content of the tobacco to be tested is obtained according to the first near infrared spectrum.

[0023] The true moisture content can also be obtained through the oven method, but for batches of tobacco to be tested, the detection time is long and the cost is high, which is not conducive to on-site implementation. Since water molecules in tobacco absorb near-infrared radiation of different wavelengths differently, the true moisture content can be obtained through the first near-infrared spectrum, which is low-cost and high-efficiency.

[0024] The present invention also provides a tobacco chemical component prediction device based on spectral conversion, comprising a processor and a memory, wherein the processor executes instructions in the memory to implement the above-mentioned tobacco chemical component prediction method based on spectral conversion.

[0025] The tobacco chemical composition prediction method based on spectral conversion is stored in a memory through programs and instructions. The method can be applied by reading the programs and instructions in the memory through a processor, so as to facilitate the determination of the chemical components in tobacco in actual scenarios and to evaluate the chemical quality of tobacco. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flowchart of a method for predicting tobacco chemical components based on spectral conversion in an embodiment of the method of the present invention;

[0027] Figure 2It is a schematic diagram of the structure of a tobacco chemical composition prediction device based on spectral conversion in an embodiment of the device of the present invention. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0029] Method Example:

[0030] The present invention provides a tobacco chemical component prediction method based on spectral conversion. The method measures the near-infrared spectrum of a sample with a "standard" moisture content, and then constructs a conversion relationship between the near-infrared spectra of samples with different moisture contents. The near-infrared spectra of samples with different moisture contents are converted into near-infrared spectra under a "standard" moisture content. A near-infrared prediction model established according to the sample with the "standard" moisture content is then used for prediction, thereby reducing the influence of moisture content on near-infrared prediction, improving prediction accuracy, and enhancing the applicability of the prediction method in actual working scenarios.

[0031] Before implementation, the present invention needs to first obtain the near-infrared spectra of tobacco samples with different moisture contents, select one of the moisture contents as the standard moisture content, calculate the correction parameters when converting the near-infrared spectra of other tobacco samples with different moisture contents to the near-infrared spectrum of the tobacco sample with the standard moisture content, thereby establishing a corresponding relationship between different moisture contents and the correction parameters.

[0032] According to the actual situation, a group of tobacco samples with high moisture content are obtained, and tobacco samples with different moisture contents are prepared by temperature and humidity control, such as baking, sun drying, etc. In this embodiment, multiple tobacco samples with different moisture contents such as 7.0%, 9.0%, 10.0%, 11.0% and 12.0% are prepared, the moisture content variation range is controlled to be ±0.4%, and the tobacco sample with the lowest moisture content of 7.0% is selected as the tobacco sample with standard moisture content.

[0033] The near infrared spectra of tobacco samples with different moisture contents are collected by a near infrared spectrometer to obtain near infrared spectra corresponding to tobacco samples with different moisture contents. In this embodiment, the near infrared spectra are represented in a matrix form, and the near infrared spectra corresponding to the moisture contents of 7.0%, 9.0%, 10.0%, 11.0% and 12.0% are represented as X 7％ , X 9％ , X 10％ , X 11％ and X 12％ Then, through spectral space conversion, the correction parameters for converting the near infrared spectra of tobacco samples with different moisture contents into the near infrared spectra of tobacco samples with standard moisture content are determined.

[0034] Near infrared spectroscopy 9％ Switch to near infrared spectroscopy 7％ Take as an example to illustrate the process of obtaining the correction parameters.

[0035] Near infrared spectroscopy 9％ and near infrared spectroscopy 7％ The combined spectrum X is obtained by splicing the head and tail together, which is expressed as:

[0036] X=[X 7％ X 9％ ]

[0037] By performing principal component analysis on the combined spectrum X, we can obtain:

[0038]

[0039] Where T represents the score matrix of the combined spectrum, represents the loading matrix of the combined spectrum, represents the sub-loading matrix corresponding to the tobacco sample with a moisture content of 7.0% in the loading matrix, The number of columns and X 7％ The number of columns is the same, represents the sub-loading matrix corresponding to the tobacco sample with a moisture content of 9.0% in the loading matrix, The number of columns and X 9％ The number of columns is the same as that of the combined spectrum, and E represents the noise part of the combined spectrum. When the load matrix is ​​divided into sub-load matrices, it is divided according to the number of tobacco samples with corresponding moisture content and the way of head-to-tail splicing. For example, the near infrared spectrum X 9％ The first near infrared spectroscopy 7％ The tails of are spliced ​​together, and there are 30 groups of 9.0% moisture content samples and 30 groups of 7.0% moisture content tobacco samples, then the first 30 columns of the loading matrix are divided into Divide the last 30 columns of the loading matrix into

[0040] X can be calculated by the following formula 9％ Correction parameter K converted to 7% moisture content 9％ :

[0041]

[0042] In the above table, “+” indicates the Moore-Penrose generalized inverse.

[0043] The correction parameter K is explained by the following derivation process: 9％ Can be used to represent near infrared spectrum X 9％ Near infrared spectroscopy 7％The reason for the difference between them. According to Lambert's absorption law, the combined spectrum X can be expressed as:

[0044]

[0045] Where C represents the concentration matrix of sample components, It represents the pure spectrum matrix corresponding to the sample moisture content of 7.0%. It represents the pure spectrum matrix corresponding to the sample moisture content of 9.0%.

[0046] Combining formula (1) and formula (3), we get:

[0047]

[0048]

[0049] In the formula, R=(C T C) -1 C T T, R are full-rank matrices and can be converted into unit matrices.

[0050] The difference between the near infrared spectra of tobacco samples with two moisture contents is reflected in the difference between the pure spectra. Assuming that the near infrared spectra X 9％ The matrix after spectral transformation is X 9％to7％ , then X 9％ With X 9％to7％ The difference between them is expressed by the following formula:

[0051]

[0052] In the formula, Indicates near infrared spectrum X 9％ Estimated values ​​of the concentration matrix corresponding to the chemical components.

[0053] Substituting formula (5) into formula (6), we get X 9％ The matrix after spectral transformation is X 9％to7％ It is expressed as:

[0054]

[0055] Therefore, the correction parameter K 9％ Indicates near infrared spectrum X 9％ Near infrared spectroscopy 7％ The difference between.

[0056] When the correction parameter K is obtained for converting the near infrared spectra of tobacco samples with different moisture contents to the standard moisture content 9％ , K 10％ , K 11％ and K12％ , the corresponding relationship between different moisture contents and the correction parameters of the corresponding moisture contents can be established.

[0057] Before the present invention is implemented, a near infrared prediction model corresponding to the standard moisture content needs to be established. First, tobacco samples with standard moisture content are screened by the KS sequential algorithm (Kennard-Stone) to obtain a group of better modeling samples to establish a near infrared prediction model with better effect. The near infrared spectra of the modeling samples are collected by a near infrared spectrometer.

[0058] Then, according to relevant national standards or industry standards, the content of chemical components in the modeling sample is obtained by using existing chemical analysis methods such as continuous flow method, ion chromatography, ICP-MS method, amino acid analyzer method, liquid chromatography tandem mass spectrometry, gravimetric method, high performance liquid chromatography and gas chromatography (all of which are well known to technicians in this field, and the specific implementation process of these methods is not described here).

[0059] In this embodiment, the contents of chemical components in the modeling samples as shown in Table 1 below are mainly obtained, including conventional chemical components, anions and cations, tobacco polyphenols, polyacids and higher fatty acids, amino acids, Amadori compounds and dichloromethane extracts, solanesol, neophytadiene, etc., a total of 69 chemical components.

[0060] Table 1 List of chemical components in modeling samples

[0061]

[0062]

[0063] Next, the near-infrared spectra and chemical content of the modeled samples were used as training sets, and the most commonly used partial least squares (PLS) method was used to establish a near-infrared prediction model for characterizing the correspondence between near-infrared spectra and chemical content. The number of factors in the near-infrared prediction model was determined by variance explanation, cross-validation, principal component investigation and long-term experience. The near-infrared prediction model was optimized by the MSC-SG derivation joint method and variable selection. The root mean square error, determination coefficient and relative analytical error were used to evaluate the performance and stability of the near-infrared prediction model.

[0064] like Figure 1 As shown, the tobacco chemical component prediction method based on spectral conversion of the present invention comprises the following steps:

[0065] S1. Obtaining the actual moisture content and the first near infrared spectrum of the tobacco to be tested.

[0066] Since water molecules in tobacco have obvious absorption of 1.94 μm near-infrared radiation and almost no absorption of the adjacent 1.84 μm near-infrared radiation, in this embodiment, the actual moisture content of the tobacco to be tested can be determined based on the absorption of near-infrared radiation in the first near-infrared spectrum.

[0067] S2. Extracting the correction parameters corresponding to the actual moisture content from the correspondence between different moisture contents and correction parameters obtained in advance.

[0068] S3. Perform spectral conversion on the first near infrared spectrum according to the correction parameters obtained in step S2, and add the correction parameters to the first near infrared spectrum to obtain a second near infrared spectrum of the tobacco to be tested at a standard moisture content.

[0069] S4. Input the second near infrared spectrum into a pre-established near infrared prediction model, and output the content of chemical components in the tobacco to be tested.

[0070] S5. Use the root mean square error to evaluate the near infrared prediction value of the chemical components of the tobacco to be tested.

[0071] In this embodiment, the tobacco to be tested can be the first-cured tobacco leaves in the tobacco leaf purchasing and delivery link, or it can be the aged tobacco leaves or processed tobacco in the tobacco processing process, and the actual working environment of the tobacco is not limited.

[0072] Taking alcoholized tobacco strips as an example, according to the actual situation, it is necessary to produce tobacco samples with a series of moisture contents of 7.0%, 9.0%, 10.0%, 11.0% and 12.0%, and select the tobacco leaf sample with a moisture content of 7.0% as the standard moisture content tobacco leaf sample, and calculate the corresponding correction parameters to establish the corresponding relationship between different moisture contents and correction parameters. A near-infrared prediction model corresponding to the tobacco leaf sample with a moisture content of 7.0% is also established.

[0073] In an actual working scenario, when the actual moisture content is obtained, the correction parameters corresponding to the actual moisture content are selected, and the spectrum conversion is performed using the corresponding formula in the following formula.

[0074] X 9％to7％ =X 9％ -K 9％

[0075] X 10％to7％ =X 10％ -K 10％

[0076] X 11％to7％ =X 11％ -K 11％

[0077] X 12％to7％ =X12％ -K 12％

[0078] For example, when the actual moisture content is 11%, select the corresponding correction parameter, and then use the correction parameter K 11％ The first near infrared spectrum X 11％ Correction is performed to obtain the corresponding second near-infrared spectrum X 11％to7％ . X 11％to7％ Input the near infrared prediction model, output the chemical composition content of the alcoholized tobacco strips to be tested with a true moisture content of 11.0% at a standard moisture content of 7.0%, and then evaluate it through RMSEP. 9％ , K 10％ , K 11％ and K 12％ It is a calibration parameter relative to the standard moisture content of 7.0%.

[0079] Taking the first-cured tobacco leaves as an example, according to the actual situation, it is necessary to produce tobacco samples with a series of moisture contents of 6.0%, 8.0%, 9.0%, 10.0%, 11.0% and 12.0%, and select the tobacco leaf sample with a moisture content of 6.0% as the standard moisture content tobacco leaf sample, and calculate the corresponding correction parameters to establish the corresponding relationship between different moisture contents and correction parameters. A near-infrared prediction model corresponding to the tobacco leaf sample with a moisture content of 6% is also established.

[0080] In an actual working scenario, when the actual moisture content is obtained, the correction parameters corresponding to the actual moisture content are selected, and the spectrum conversion is performed using the corresponding formula in the following formula.

[0081] X 8％to6％ =X 8％ -K 8％

[0082] X 9％to6％ =X 9％ -K 9％

[0083] X 10％to6％ =X 10％ -K 10％

[0084] X 11％to6％ =X 11％ -K 11％

[0085] X 12％to6％ =X 12％ -K 12％

[0086] For example, when the actual moisture content is 10.0%, the corresponding correction parameters are selected, and then the first near infrared spectrum X is corrected by the correction parameters. 10％ Correction is performed to obtain the corresponding second near-infrared spectrum X 10％to7％ . X 10％to7％ Input the near infrared prediction model, output the chemical composition content of the tested freshly cured tobacco leaves with a true moisture content of 10% at a standard moisture content of 6.0%, and then evaluate it through RMSEP. 8％ , K 9％ , K 10％ , K 11％ and K 12％ It is a calibration parameter relative to the standard moisture content of 6.0%.

[0087] The present invention can predict the chemical composition of first-cured tobacco leaves, redried tobacco strips, aged tobacco strips, and cigarette tobacco in different work links such as tobacco leaf purchase and delivery, aging progress, process formula, and cigarette product quality. It can also reduce the impact of moisture on near-infrared prediction, and compared with directly establishing near-infrared prediction models for different moisture ranges, it has lower costs and higher efficiency.

[0088] In order to further illustrate the effect of the present invention, the above method is applied to a specific example. Taking the prediction before and after 11% moisture correction as an example, the comparison of the prediction results before and after correction is shown in Table 2, where the root mean square error is the mean of the square root of the error between the predicted value and the true value, the smaller the better; R 2 is the coefficient of determination, which is how much better the predicted value is compared with the case where only the mean is used. It is between 0 and 1, and the larger the better. RPD is the relative analytical error. The larger it is, the more accurate the prediction is.

[0089] Table 2 Comparison of prediction results before and after moisture correction

[0090]

[0091]

[0092]

[0093] It can be seen from Table 2 that, overall, the prediction results of each component after correction are improved to a certain extent compared with those before correction. Although the indicators of some components are not as good as before correction, the overall prediction results are closer to the actual ones, indicating that the correction method of the present invention can improve the prediction accuracy of each component.

[0094] Device Example:

[0095] A tobacco chemical component prediction device based on spectral conversion of the present invention is as follows Figure 2As shown, it includes a memory, a processor and an internal bus, and the processor and the memory complete data and communication interaction with each other through the internal bus. The processor executes the instructions in the memory to implement the tobacco chemical component prediction method based on spectral conversion in the method embodiment. The implementation of the method has been clearly introduced in the method embodiment and will not be repeated here.

[0096] The processor may be a microprocessor MCU, a programmable logic device FPGA or other processing device. The memory may be various memories that store information using electrical energy, such as RAM, ROM, etc.; it may also be various memories that store information using magnetic energy, such as hard disks, floppy disks, magnetic tapes, magnetic core memories, magnetic bubble memories, USB flash drives, etc.; it may also be various memories that store information using optical methods, such as CDs, DVDs, etc.; of course, it may also be other types of memories, such as quantum memories, graphene memories, etc.

Claims

1. A method for predicting tobacco chemical components based on spectral conversion, characterized in that: The steps include: S1, obtaining a first near infrared spectrum of the tobacco to be tested; S2. According to the actual moisture content of the tobacco to be tested, the near infrared spectra of the tobacco samples with different moisture contents are respectively spliced ​​end to end with the near infrared spectra of the tobacco samples with standard moisture content to obtain combined spectra corresponding to other different moisture contents; principal component analysis is performed on each combined spectrum to obtain a first sub-load matrix corresponding to the current moisture content and a second sub-load matrix corresponding to the standard moisture content, and the difference between the first sub-load matrix and the second sub-load matrix is ​​calculated as a correction parameter corresponding to the current moisture content; the correction parameter corresponding to the actual moisture content is obtained from the correspondence between the different moisture contents and the correction parameters obtained in advance; S3, performing spectral conversion on the first near infrared spectrum according to the correction parameters obtained in step S2 to obtain a second near infrared spectrum of the tobacco to be tested at a standard moisture content; S4, inputting the second near infrared spectrum into a pre-established near infrared prediction model, and outputting the content of chemical components in the tobacco to be tested; the near infrared prediction model is used to represent the corresponding relationship between the near infrared spectrum of tobacco with standard moisture content and the content of chemical components in tobacco with corresponding standard moisture content; The correspondence between the different moisture contents and the correction parameters described in step S2 is obtained through the following process: obtaining the near-infrared spectra of tobacco samples with different moisture contents, selecting one of the tobacco samples with different moisture contents as the tobacco sample with standard moisture content, and respectively calculating the correction parameters for converting the near-infrared spectra of the tobacco samples with other moisture contents into the near-infrared spectra of the tobacco sample with standard moisture content, and then establishing the correspondence between the different moisture contents and the correction parameters according to the moisture content and the correction parameters corresponding to the moisture content.

2. The method for predicting tobacco chemical components based on spectral conversion according to claim 1, characterized in that: The correction parameters are calculated by the following formula: In the formula, K 当前水分含量 represents the correction parameter corresponding to the current moisture content, X 当前水分含量 Near infrared spectrum of tobacco sample indicating current moisture content, represents the first sub-load matrix corresponding to the current moisture content, represents the second sub-load matrix corresponding to the standard moisture content, (·) + represents the Moore-Penrose generalized inverse of a matrix.

3. The tobacco chemical component prediction method based on spectral conversion according to claim 1 or 2, characterized in that: When the corresponding relationship between the different moisture contents and the calibration parameters is established, the tobacco sample with the lowest moisture content is selected as the tobacco sample with the standard moisture content.

4. The method for predicting tobacco chemical components based on spectral conversion according to claim 1, characterized in that: The real moisture content of the tobacco to be tested is obtained according to the first near infrared spectrum.

5. A tobacco chemical component prediction device based on spectral conversion, characterized in that: The method comprises a processor and a memory, wherein the processor executes instructions in the memory to implement the tobacco chemical component prediction method based on spectral conversion according to any one of claims 1 to 4.

Images