Flue-cured tobacco real-time grading method and device based on near-infrared online detection data

Abstract

The invention relates to the technical field of tobacco leaf processing, in particular to a flue-cured tobacco real-time grading method and device based on near-infrared online detection data. Firstly, chemical indexes, obtained based on the near-infrared online detection data, of flue-cured tobacco leaves are associated with quality grades obtained in an appearance grading mode; the method comprises the following steps of: screening key chemical indexes through an information gain and a Fisher criterion, and then selecting an iterative algorithm (Informational gain and Fisher crisis restitution) to screen out the key chemical indexes through the information gain and the Fisher criterion; secondly, designing a tobacco leaf quality chemical index grading table aiming at a specific producing area on the basis of the screened key chemical indexes in combination with rich and historical tobacco leaf chemical index statistical data; finally, real-time grading of the flue-cured tobaccos is achieved on the basis of the combination of near-infrared online detection data and a designed chemical index grading table, and the grading efficiency and accuracy are improved. The problems that when tobacco leaves are graded in real time in the prior art, the grading quality is uneven and the grading efficiency is low in the flue-cured tobacco grading process due to the fact that chemical indexes referred to in the tobacco leaf grading process are different are solved.

Description

Technical Field

[0001] The invention relates to the technical field of tobacco leaf processing, and in particular to a tobacco leaf raw material quality characteristic analysis method based on data fusion. Background Art

[0002] In the process of tobacco leaf processing, in order to ensure the consistency and high quality of tobacco products, the internal quality grading of tobacco leaves has become an important issue in the tobacco industry. Common tobacco leaf quality grading methods include appearance quality grading, sensory quality grading (aroma, smoking evaluation) and chemical composition analysis. Among them, appearance quality grading and sensory quality grading require a large number of professionals to judge the appearance of tobacco leaves through visual, olfactory and smoking experience, which is highly subjective. For chemical composition analysis, the chemical composition in tobacco leaves, such as nicotine, sugar and volatile substances, is analyzed to scientifically evaluate the quality of tobacco leaves.

[0003] At present, chemical composition analysis based on near-infrared spectroscopy has become one of the most effective methods for tobacco quality analysis, and can quickly detect conventional chemical components such as raw sugar, total sugar, nicotine, total nitrogen, potassium and chlorine in tobacco leaves. For the conventional chemical composition of tobacco leaves, many researchers have pointed out dozens of chemical indicators that characterize the intrinsic quality of tobacco leaves based on conventional chemical composition, such as sugar-nitrogen ratio, sugar-alkali ratio, two sugar ratio, nitrogen-alkali ratio, nicotine-nitrogen ratio, ammonia nitrogen, potassium-chlorine ratio, aroma value, dolphin, free amino acids, petroleum ether extract and Amadori compounds; in addition, some experts have also conducted research on the correlation between chemical indicators and appearance and sensory evaluation. The results show that the intrinsic chemical indicators of tobacco leaves are highly correlated with appearance and sensory quality evaluation.

[0004] However, there are different opinions on which chemical indicators of tobacco leaves can be used to achieve efficient grading. So far, there is no unified and effective industry norm or standard for intrinsic quality evaluation based on the key chemical indicators of tobacco leaves, resulting in uneven grading quality of flue-cured tobacco.

[0005] Therefore, it is necessary to improve and design the grading methods and standards for tobacco leaves in order to solve the problems of uneven grading quality and low grading efficiency in the existing tobacco grading process due to different chemical indicators used as reference in the tobacco grading process during real-time grading. Summary of the invention

[0006] The purpose of the present invention is to propose a real-time grading method and device for flue-cured tobacco based on near-infrared online detection data, so as to solve the problems of uneven grading quality and low grading efficiency in the existing real-time grading of tobacco leaves due to different chemical indicators used as reference in the tobacco leaf grading process.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] The first aspect of the present application proposes a tobacco leaf raw material quality feature analysis method based on data fusion, comprising the following steps:

[0009] Step 1, collect near-infrared spectral data of different types of tobacco raw materials, measure the content of conventional chemical components of different types of tobacco raw materials, and build a near-infrared online quantitative model of the content of conventional chemical components of tobacco raw materials based on the near-infrared spectral data of tobacco raw materials and the measured values ​​of conventional chemical components of tobacco raw materials;

[0010] Step 2: Based on the near-infrared online quantitative model of the conventional chemical component content of tobacco leaf raw materials, the conventional chemical component content of tobacco leaf raw materials is predicted and calculated to obtain the predicted value of the conventional chemical component content and derived chemical indicators of tobacco leaf raw materials, and professional graders determine the appearance quality grade of the primary tobacco leaf raw materials based on the appearance of the tobacco leaf raw materials, such as color;

[0011] Step 3, establish the connection between the appearance quality grade of tobacco leaf raw materials and the quantitative chemical index data, use information gain and Fisher criterion to select iterative algorithm, screen the chemical index data with established connection, and obtain the key chemical indexes of tobacco leaf raw materials;

[0012] Step 4. Taking the key chemical indicators of tobacco leaf raw materials as a reference, statistically analyze the key chemical indicator data of tobacco leaf raw materials over the years, and combine the statistical quantities of the key chemical indicator data of tobacco leaf raw materials over the years to develop a tobacco leaf raw material quality chemical indicator grading table;

[0013] Step 5. Obtain the near-infrared spectral data of the tobacco leaf raw material sample, input the near-infrared spectral data of the tobacco leaf raw material sample into the near-infrared online quantitative model of the conventional chemical component content of the tobacco leaf raw material in Step 2, obtain the predicted values ​​of the derived chemical indicators of the tobacco leaf raw material sample, and compare the predicted values ​​of the key chemical indicators of the tobacco leaf raw material sample with the tobacco leaf raw material quality chemical indicator grading table in Step 4 to determine the quality grade of the tobacco leaf raw material sample.

[0014] Preferably, the Step 1, collecting near-infrared spectral data of different categories of tobacco raw materials, measuring the contents of conventional chemical components of different categories of tobacco raw materials, and constructing a near-infrared online quantitative model of the contents of conventional chemical components of tobacco raw materials based on the measured values ​​of conventional chemical components of tobacco raw materials and near-infrared spectral data, comprises the following steps:

[0015] Step 1.1, collect multiple different varieties of first-cured tobacco raw materials from different origins, and select first-cured tobacco raw materials from different growth parts among the different varieties of first-cured tobacco raw materials;

[0016] Step 1.2, using near infrared spectroscopy technology, performing near infrared spectroscopy scanning on the raw materials of the first-cured tobacco leaves at different growth parts in Step 1.1, and obtaining near infrared spectroscopy data of the raw materials of the first-cured tobacco leaves;

[0017] Step 1.3, chemically analyze the raw materials of the first-cured tobacco leaves from different growth parts in Step 1.1, and determine the actual values ​​of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine;

[0018] Step 1.4: Based on the near-infrared spectral data of the freshly cured tobacco leaf raw materials in Step 1.2 and the measured values ​​of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in Step 1.3, a near-infrared online quantitative model for the content of conventional chemical components in tobacco leaf raw materials is constructed through modeling methods such as partial least squares (PLS).

[0019] Preferably, the Step 2, based on the near infrared online quantitative model of the conventional chemical component content of the tobacco leaf raw material, predicts and calculates the conventional chemical component content of the tobacco leaf raw material, obtains the predicted value of the conventional chemical component content of the tobacco leaf raw material and the derived chemical index, and determines the appearance quality grade of the tobacco leaf raw material based on the appearance of the tobacco leaf raw material, such as color, by professional graders, including the following steps:

[0020] Step 2.1, using the near infrared online quantitative model of the conventional chemical components of tobacco raw materials, predict the conventional chemical components of new tobacco raw material samples, and obtain the predicted contents of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in the new tobacco raw material samples;

[0021] Step 2.2, based on the predicted contents of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in the new tobacco leaf raw material samples, the derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, total nitrogen / nicotine in the new tobacco leaf raw material samples are calculated to obtain the derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, total nitrogen / nicotine in the new tobacco leaf raw material samples;

[0022] Step 2.3. Professional graders determine the appearance quality grade of the new tobacco leaf raw material samples based on their appearance and color.

[0023] Preferably, the Step 3, establishing a connection between the appearance quality grade of tobacco leaf raw materials and the quantitative chemical index data, using information gain and Fisher criterion to select an iterative algorithm, screening the chemical index data with established connections, and obtaining the key chemical indexes of tobacco leaf raw materials, comprises the following steps:

[0024] Step 3.1, establish the connection between the tobacco leaf raw material appearance quality grade and the quantitative chemical index data, and obtain the fused tobacco leaf raw material appearance quality grade and quantitative chemical index data set;

[0025] Step 3.2, calculate the information gain value of each chemical index in the fused tobacco leaf raw material appearance quality grade and quantitative chemical index data set, and obtain the information gain value of each chemical index of the tobacco leaf raw material. The calculation formula is as follows:

[0026] Gain(D, A)=Ent(D)-Ent(D|A)

[0027]

[0028] Where K is the number of different levels; |D| represents the total number of samples; |C k |Indicates grade C k Number of samples

[0029] Step 3.3, sort the information gain values ​​of the chemical indicators of tobacco raw materials from large to small, and select the first M chemical indicators with high information gain values ​​as the primary chemical indicators;

[0030] Step 3.4, calculate the Fisher scores of the first M chemical indicators with high information gain values, and obtain the Fisher scores of the first M chemical indicators with high information gain values. The calculation formula is as follows:

[0031]

[0032] Where K is the number of different levels; |C k |Indicates grade C k The number of samples; m k Indicates level C k The average value of each index in the sample; m is the average value of each index in all samples; x n Indicates level C k The values ​​of each indicator

[0033] Step 3.5, sort the Fisher scores of the first M chemical indicators with high information gain values ​​from large to small, and select N chemical indicators with high Fisher scores as a candidate subset;

[0034] Step 3.6, let N = N + 1, repeat steps Step 3.3 to Step 3.6 until N is equal to M;

[0035] Step 3.7, let M = M + 1, repeat Step 3.3 to Step 3.7 until M is equal to the number of chemical indicators, and obtain a series of chemical indicator subsets;

[0036] Step 3.8, count the frequencies of different indicators in the series of chemical indicator subsets, and arrange the frequencies of each chemical indicator from high to low. The higher the frequency, the more critical the indicator is. Select several chemical indicators with higher frequencies;

[0037] Step 3.9, using the Pearson correlation coefficient, eliminate the linear correlation between different chemical indicators in the series of chemical indicator subsets, and obtain several chemical indicators with high frequency after eliminating the linear correlation, which are used as the key chemical indicators for measuring tobacco leaf quality. The elimination formula is as follows:

[0038]

[0039] Where N is the total number of samples, is the content of the ith sample in index a, and is the average value of the content of index a in all samples

[0040] Preferably, the Step 4, taking the key chemical indicators of tobacco leaf raw materials as a reference, statistically analyzing the key chemical indicator data of tobacco leaf raw materials over the years, and combining the statistical quantities of the key chemical indicator data of tobacco leaf raw materials over the years to formulate a tobacco leaf raw material quality chemical indicator grading table, comprises the following steps:

[0041] Step 4.1. Using the key chemical indicators of tobacco leaf raw materials as a reference, screen out the key chemical indicator data of tobacco leaf raw materials over the years, and statistically analyze the key chemical indicator data of tobacco leaf raw materials over the years to obtain the statistical quantities such as the average value, standard deviation and coefficient of variation of the key chemical indicator data of tobacco leaf raw materials over the years;

[0042] Step 4.2, based on the statistics of key chemical index data of tobacco leaf raw materials over the years, determine the precise thresholds of key chemical indexes for each quality grade of tobacco leaf raw materials;

[0043] Step 4.3: Based on the precise thresholds of key chemical indicators for each tobacco leaf raw material quality grade in Step 4.2, develop a tobacco leaf raw material quality chemical indicator grading table.

[0044] Preferably, the Step 5, obtaining near infrared spectral data of the tobacco leaf raw material sample, inputting the near infrared spectral data of the tobacco leaf raw material sample into the near infrared online quantitative model of the conventional chemical component content of the tobacco leaf raw material in Step 2, obtaining the predicted values ​​of the key chemical indicators of the tobacco leaf raw material sample, comparing the predicted values ​​of the key chemical indicators of the tobacco leaf raw material sample with the tobacco leaf raw material quality chemical indicator grading table in Step 4, and determining the quality grade of the tobacco leaf raw material, comprises the following steps:

[0045] Step 5.1, using near infrared spectroscopy technology to scan the tobacco leaf raw material sample to obtain near infrared spectral data in the tobacco leaf raw material sample;

[0046] Step 5.2, using the near infrared online quantitative model of the conventional chemical composition content of tobacco leaf raw materials in Step 2, predict the conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in tobacco leaf raw material samples and the derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, total nitrogen / nicotine in tobacco leaf raw material samples, and obtain the predicted values ​​of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in tobacco leaf raw material samples and the predicted values ​​of derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, total nitrogen / nicotine in tobacco leaf raw material samples;

[0047] Step 5.3. Compare the conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in the tobacco leaf raw material sample in Step 5.2 and the predicted values ​​of derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, total nitrogen / nicotine with the precise threshold values ​​of key chemical indicators of each tobacco leaf raw material quality grade in the tobacco leaf raw material quality chemical indicator grading table formulated in Step 4 to determine the quality grade of the tobacco leaf raw material sample.

[0048] The second aspect of the present application proposes a real-time flue-cured tobacco grading device based on near infrared online detection data, comprising:

[0049] A data acquisition module is used to collect near-infrared spectral data of different types of tobacco raw materials and obtain the measured values ​​of conventional chemical components of different types of tobacco raw materials;

[0050] The quantitative analysis module is connected with the data acquisition module. The quantitative analysis module is used to predict and calculate the content of conventional chemical components of tobacco raw materials based on the near-infrared spectral data of different categories of tobacco raw materials, and determine the appearance quality grade of the primary tobacco raw materials based on the appearance of the tobacco raw materials, such as color, by professional graders;

[0051] The key data screening module, the quantitative analysis module is connected with the key data screening module, and the key data screening module is used to screen the chemical index data related to the appearance quality grade of the tobacco leaf raw material to obtain the key chemical index of the tobacco leaf raw material;

[0052] The grading reference module, the key data screening module is connected with the grading module, the grading module is used to take the key chemical indicators of tobacco leaf raw materials as a reference, to collect the key chemical indicator data of tobacco leaf raw materials over the years, and to formulate a grading table of tobacco leaf raw material quality chemical indicators in combination with the statistical quantities of the key chemical indicator data of tobacco leaf raw materials over the years;

[0053] The automatic grading decision module is connected with the grading reference module, and the automatic grading decision module is used to obtain the predicted values ​​of key chemical indicators of tobacco leaf raw material samples, compare the predicted values ​​of key chemical indicators of tobacco leaf raw material samples with the tobacco leaf raw material quality chemical indicator grading table, and determine the quality grade of the tobacco leaf raw material samples.

[0054] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0055] 1. A real-time grading method for flue-cured tobacco based on near-infrared online detection data in the present invention relates the chemical indexes of the first-cured tobacco leaves obtained based on the near-infrared online detection data to the quality grades obtained by the appearance grading method, and screens the key chemical indexes through the information gain and Fisher criterion reselection iteration algorithm; secondly, based on the screened key chemical indexes and in combination with the rich and historical statistical data of tobacco chemical indexes, a tobacco quality chemical index grading table for a specific origin is designed; finally, based on the near-infrared online detection data combined with the designed chemical index grading table, real-time grading of flue-cured tobacco is realized, and grading efficiency and accuracy are improved, so as to solve the problem that in the existing real-time grading of tobacco leaves, the grading quality is uneven and the grading efficiency is low in the tobacco grading process due to the different chemical indexes used as reference in the tobacco grading process.

[0056] 2. In the present invention, by associating the chemical indicators detected online with the traditional grading results, it can not only achieve an efficient integration of algorithm theory and practical experience, but also enhance the scientificity and practicality of grading. The algorithm used can also effectively screen out the key chemical indicators that can best distinguish the quality grades of tobacco leaves.

[0057] 3. The customized grading table in the present invention not only fully considers the characteristics of tobacco leaves from specific origins, but also combines historical big data analysis to make the grading standards more in line with regional realities, ensuring the pertinence and accuracy of the grading results, which is conducive to maintaining the characteristics of tobacco leaves in the producing area and enhancing market competitiveness.

[0058] 4. The present invention uses online detection with the help of near-infrared technology, combined with a designed chemical index grading table, which can not only achieve real-time flue-cured tobacco grading, shorten the cycle from detection to grading and improve production efficiency, but also ensure the accuracy and consistency of grading.

[0059] 5. Through the deep integration of technology and data, the present invention can not only simplify the grading process and reduce manual dependence, but also improve the automation and intelligence level of grading, providing a new idea for tobacco quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1The present invention is a flowchart of a real-time flue-cured tobacco grading method based on near-infrared online detection data according to an embodiment of the present invention.

[0061] Figure 2 It is the information gain and Fisher score of multiple chemical indicators in the specific embodiment of the present invention.

[0062] Figure 3 It is the number of times that multiple chemical indicators are selected in the specific embodiment of the present invention.

[0063] Figure 4 Schematic diagram of the structure of a real-time flue-cured tobacco grading device based on near-infrared online detection data according to an embodiment of the present invention.

[0064] Table 1 is a chemical index classification table based on reducing sugar total nitrogen ratio and nicotine DETAILED DESCRIPTION

[0065] like Figure 1-4 In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0066] Example

[0067] Regarding the conventional chemical composition of tobacco leaves, many researchers have pointed out dozens of chemical indicators that characterize the intrinsic quality of tobacco leaves based on the conventional chemical composition, such as sugar-nitrogen ratio, sugar-alkali ratio, two-sugar ratio, nitrogen-alkali ratio, nicotine-nitrogen ratio to ammonia nitrogen, potassium-chlorine ratio, aroma value, dolphin, free amino acids, petroleum ether extracts and Amadori compounds; in addition, some experts have also conducted research on the correlation between chemical indicators and appearance and sensory evaluation. The results show that the intrinsic chemical indicators of tobacco leaves are highly correlated with the appearance and sensory quality evaluation. However, there are many different opinions on which chemical indicators of tobacco leaves can be used to achieve efficient grading. So far, there has been no unified and effective industry specification or standard for intrinsic quality evaluation based on key chemical indicators of tobacco leaves, resulting in uneven grading quality of flue-cured tobacco.

[0068] Therefore, a real-time grading method for flue-cured tobacco based on near-infrared online detection data is proposed to solve the problems of uneven grading quality and low grading efficiency in the existing real-time grading of tobacco leaves due to different reference chemical indicators in the tobacco leaf grading process.

[0069] Specifically, the first aspect of the present application proposes a real-time grading method for flue-cured tobacco based on near-infrared online detection data, see Figure 1 , including the following steps:

[0070] The first step is to collect near-infrared spectral data of different types of tobacco raw materials, measure the content of conventional chemical components of different types of tobacco raw materials, and construct a near-infrared online quantitative model of the content of conventional chemical components of tobacco raw materials based on the near-infrared spectral data of tobacco raw materials and the measured values ​​of conventional chemical components of tobacco raw materials. The model specifically includes the following steps:

[0071] 1) Collecting multiple different varieties of first-cured tobacco raw materials from different production areas in Yuxi, Yunnan, and selecting first-cured tobacco raw materials from different growth parts among the different varieties of first-cured tobacco raw materials;

[0072] 2) Using near infrared spectroscopy technology, perform near infrared spectroscopy scanning on the raw materials of freshly cured tobacco leaves from different growth parts to obtain near infrared spectroscopy data of the raw materials of freshly cured tobacco leaves;

[0073] 3) Conduct chemical analysis on raw materials of first-cured tobacco leaves from different growth parts to determine the actual values ​​of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine;

[0074] 4) Based on the near-infrared spectral data of freshly cured tobacco leaf raw materials and the measured values ​​of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine, a near-infrared online quantitative model for the content of conventional chemical components in tobacco leaf raw materials was constructed through modeling methods such as the partial least squares (PLS).

[0075] Step 2: Please refer to Figure 2 and Figure 3 Based on the near-infrared online quantitative model of the conventional chemical component content of tobacco leaf raw materials, the conventional chemical component content of tobacco leaf raw materials is predicted and calculated to obtain the predicted value of the conventional chemical component content and derived chemical indicators of tobacco leaf raw materials, and the appearance quality grade of the primary tobacco leaf raw materials is determined by professional graders based on the appearance of tobacco leaf raw materials, such as color, etc., which specifically includes the following steps:

[0076] 1) Using the near-infrared online quantitative model of the conventional chemical composition content of tobacco leaf raw materials, the conventional chemical composition content of 2559 new tobacco leaf raw material samples was predicted, and the predicted contents of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in the new tobacco leaf raw material samples were obtained;

[0077] 2) Based on the predicted contents of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in the new tobacco leaf raw material samples, the derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, and total nitrogen / nicotine in the new tobacco leaf raw material samples are calculated to obtain the derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, and total nitrogen / nicotine in the new tobacco leaf raw material samples;

[0078] 3) Professional graders determine the appearance quality grade of the new tobacco leaf raw material samples based on their appearance and color.

[0079] The third step is to establish the connection between the appearance quality grade of tobacco leaf raw materials and the quantitative chemical index data, and use information gain and Fisher criterion to select iterative algorithm to screen the chemical index data with established connection to obtain the key chemical index of tobacco leaf raw materials, which specifically includes the following steps:

[0080] 1) Establish the connection between the appearance quality grade of tobacco leaf raw materials and the quantitative chemical index data, and obtain the fused appearance quality grade of tobacco leaf raw materials and the quantitative chemical index data set;

[0081] 2) Calculate the information gain value of each chemical index in the fused tobacco leaf raw material appearance quality grade and quantitative chemical index data set to obtain the information gain value of each chemical index of the tobacco leaf raw material. The calculation formula is as follows:

[0082] Gain(D, A)=Ent(D)-Ent(D|A)

[0083]

[0084] Where K is the number of different levels; |D| represents the total number of samples; |C k |Indicates grade C k Number of samples

[0085] 3) Sort the information gain values ​​of the chemical indicators of tobacco raw materials from large to small, and select the first M chemical indicators with high information gain values ​​as the primary chemical indicators;

[0086] 4) Calculate the Fisher scores of the first M chemical indicators with high information gain values, and obtain the Fisher scores of the first M chemical indicators with high information gain values. The calculation formula is as follows:

[0087]

[0088] Where K is the number of different levels; |C k |Indicates grade C k The number of samples; m k Indicates level C k The average value of each index in the sample; m is the average value of each index in all samples; x n Indicates level C k The values ​​of each indicator

[0089] 5) Sort the Fisher scores of the first M chemical indicators with high information gain values ​​from large to small, and select N chemical indicators with high Fisher scores as a candidate subset;

[0090] 6) Let N = N + 1, and repeat Step 3.3 to Step 3.6 until N is equal to M;

[0091] 7) Let M = M + 1, repeat Step 3.3 to Step 3.7 until M is equal to the number of chemical indicators, and obtain a series of chemical indicator subsets;

[0092] 8) Count the frequencies of different indicators in the series of chemical indicator subsets, and arrange the frequencies of each chemical indicator from high to low. The higher the frequency, the more critical the indicator is. Select several chemical indicators with higher frequencies;

[0093] 9) Pearson correlation coefficient is used to eliminate the linear correlation between different chemical indicators in the series of chemical indicator subsets, and several chemical indicators with high frequency of occurrence after eliminating the linear correlation are obtained as key chemical indicators for measuring tobacco leaf quality. The elimination formula is as follows:

[0094]

[0095] Where N is the total number of samples, is the content of the ith sample in index a, and is the average value of the content of index a in all samples

[0096] The fourth step is to use the key chemical indicators of tobacco leaf raw materials as a reference, collect the key chemical indicator data of tobacco leaf raw materials over the years, and combine the statistical quantities of the key chemical indicator data of tobacco leaf raw materials over the years to develop a tobacco leaf raw material quality chemical indicator grading table, which specifically includes the following steps:

[0097] 1) Using the key chemical indicators of tobacco leaf raw materials as a reference, screen out the key chemical indicator data of tobacco leaf raw materials over the years, and statistically analyze the key chemical indicator data of tobacco leaf raw materials over the years to obtain the statistical quantities such as the average value, standard deviation and coefficient of variation of the key chemical indicator data of tobacco leaf raw materials over the years;

[0098] 2) Based on the statistics of key chemical index data of tobacco leaf raw materials over the years, determine the precise thresholds of key chemical indexes for each quality grade of tobacco leaf raw materials;

[0099] 3) Based on the precise thresholds of key chemical indicators for each tobacco leaf raw material quality grade, a tobacco leaf raw material quality chemical indicator grading table is developed, as shown in Table 1.

[0100] Table 1: Chemical index classification table based on reducing sugar-to-total nitrogen ratio and nicotine

[0101]

[0102] Step 5: Obtain near infrared spectral data of the tobacco leaf raw material sample, input the near infrared spectral data of the tobacco leaf raw material sample into the near infrared online quantitative model of the conventional chemical component content of the tobacco leaf raw material, obtain the predicted values ​​of the key chemical indicators of the tobacco leaf raw material sample, compare the predicted values ​​of the key chemical indicators of the tobacco leaf raw material sample with the tobacco leaf raw material quality chemical indicator grading table in Step 4, and determine the quality grade of the tobacco leaf raw material sample, which specifically includes the following steps:

[0103] 1) Scan the tobacco leaf raw material sample using near infrared spectroscopy technology to obtain near infrared spectral data in the tobacco leaf raw material sample;

[0104] 2) Using the near-infrared online quantitative model of the conventional chemical composition content of tobacco leaf raw materials, conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in tobacco leaf raw material samples and derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, and total nitrogen / nicotine in tobacco leaf raw material samples are predicted to obtain the predicted values ​​of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in tobacco leaf raw material samples and the predicted values ​​of derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, and total nitrogen / nicotine in tobacco leaf raw material samples;

[0105] 3) Compare the predicted values ​​of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in the tobacco raw material sample and derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, total nitrogen / nicotine with the precise threshold values ​​of key chemical indicators of each tobacco raw material quality grade in the tobacco raw material quality chemical indicator grading table formulated in Step 4 to determine the quality grade of the tobacco raw material sample.

[0106] The above is a real-time flue-cured tobacco grading method based on near-infrared online detection data proposed in the first aspect of the present application, and the following is a real-time flue-cured tobacco grading device based on near-infrared online detection data proposed in the second aspect of the present application.

[0107] For details, please refer to Figure 4 , including data collection module, quantitative analysis module, key data screening module, grading reference module and automatic grading decision module;

[0108] Among them, the data acquisition module is used to collect near-infrared spectral data of different types of tobacco raw materials and obtain the measured values ​​of conventional chemical components of different types of tobacco raw materials;

[0109] The quantitative analysis module is connected with the data acquisition module. The quantitative analysis module is used to predict and calculate the content of conventional chemical components of tobacco raw materials based on the near-infrared spectral data of different types of tobacco raw materials, and determine the appearance quality grade of the primary tobacco raw materials based on the appearance of the tobacco raw materials, such as color, by professional graders;

[0110] The key data screening module, the quantitative analysis module is connected with the key data screening module, and the key data screening module is used to screen the connection between the appearance quality grade of the tobacco leaf raw material and the quantitative chemical index data to obtain the key chemical index of the tobacco leaf raw material;

[0111] The grading reference module, the key data screening module is connected with the grading module, the grading module is used to take the key chemical indicators of tobacco leaf raw materials as a reference, to collect the key chemical indicator data of tobacco leaf raw materials over the years, and to formulate a grading table of tobacco leaf raw material quality chemical indicators in combination with the statistical quantities of the key chemical indicator data of tobacco leaf raw materials over the years;

[0112] The automatic grading decision module is connected with the grading reference module, and the automatic grading decision module is used to obtain the predicted values ​​of key chemical indicators of tobacco leaf raw material samples, compare the predicted values ​​of key chemical indicators of tobacco leaf raw material samples with the tobacco leaf raw material quality chemical indicator grading table, and determine the quality grade of the tobacco leaf raw material samples.

[0113] The present application discloses a real-time grading method for flue-cured tobacco based on near-infrared online detection data. The method relates the chemical indicators of the first-cured tobacco leaves obtained based on the near-infrared online detection data to the quality grades obtained by the appearance grading method, and screens the key chemical indicators through the information gain and Fisher criterion reselection iteration algorithm. Secondly, based on the screened key chemical indicators and in combination with the rich and historical statistical data of tobacco chemical indicators, a tobacco quality chemical indicator grading table for a specific origin is designed. Finally, based on the near-infrared online detection data and the designed chemical indicator grading table, real-time grading of flue-cured tobacco is realized, and the grading efficiency and accuracy are improved, so as to solve the problems of uneven grading quality and low grading efficiency in the flue-cured tobacco grading process due to different chemical indicators used as reference in the tobacco grading process when grading tobacco leaves in real time.

[0114] Although the present invention is described herein with reference to a number of illustrative embodiments of the present invention, it will be appreciated that those skilled in the art may devise many other modifications and implementations that fall within the scope and spirit of the principles disclosed herein. More specifically, within the scope of the present disclosure, drawings, and claims, a variety of variations and modifications may be made to the components and / or layout of the subject combination layout. In addition to the variations and modifications made to the components and / or layout, other uses will also be apparent to those skilled in the art.

Claims

1. A real-time grading method for flue-cured tobacco based on near-infrared online detection data, characterized in that: The following steps are involved: Step 1, collect near-infrared spectral data of different types of tobacco raw materials, measure the content of conventional chemical components of different types of tobacco raw materials, and build a near-infrared online quantitative model of the content of conventional chemical components of tobacco raw materials based on the near-infrared spectral data of tobacco raw materials and the measured values ​​of conventional chemical components of tobacco raw materials; Step 2: Based on the near-infrared online quantitative model of the conventional chemical component content of tobacco leaf raw materials, the conventional chemical component content of tobacco leaf raw materials is predicted and calculated to obtain the predicted value of the conventional chemical component content and derived chemical indicators of tobacco leaf raw materials, and professional graders determine the appearance quality grade of the primary tobacco leaf raw materials based on the appearance of the tobacco leaf raw materials, such as color; Step 3, establish the connection between the appearance quality grade of tobacco leaf raw materials and the quantitative chemical index data, use information gain and Fisher criterion to select iterative algorithm, screen the chemical index data with established connection, and obtain the key chemical indexes of tobacco leaf raw materials; Step 4. Taking the key chemical indicators of tobacco leaf raw materials as a reference, statistically analyze the key chemical indicator data of tobacco leaf raw materials over the years, and combine the statistical quantities of the key chemical indicator data of tobacco leaf raw materials over the years to develop a tobacco leaf raw material quality chemical indicator grading table; Step 5. Obtain near-infrared spectral data of tobacco leaf raw material samples, input the near-infrared spectral data of tobacco leaf raw material samples into the near-infrared online quantitative model of conventional chemical component content of tobacco leaf raw materials in Step 2, obtain predicted values ​​of key chemical indicators of tobacco leaf raw material samples, compare the predicted values ​​of key chemical indicators of tobacco leaf raw material samples with the tobacco leaf raw material quality chemical indicator grading table in Step 4, and determine the quality grade of tobacco leaf raw material samples.

2. The method for real-time grading of flue-cured tobacco based on near-infrared online detection data according to claim 1, characterized in that: The step 1, collecting near infrared spectral data of different types of tobacco raw materials, measuring the content of conventional chemical components of different types of tobacco raw materials, and constructing a near infrared online quantitative model of the content of conventional chemical components of tobacco raw materials based on the measured values ​​of conventional chemical components of tobacco raw materials and near infrared spectral data, comprises the following steps: Step 1.1, collect multiple different varieties of first-cured tobacco raw materials from different origins, and select first-cured tobacco raw materials from different growth parts among the different varieties of first-cured tobacco raw materials; Step 1.2, using near infrared spectroscopy technology, performing near infrared spectroscopy scanning on the raw materials of the first-cured tobacco leaves at different growth parts in Step 1.1, and obtaining near infrared spectroscopy data of the raw materials of the first-cured tobacco leaves; Step 1.3, chemically analyze the raw materials of the first-cured tobacco leaves from different growth parts in Step 1.1, and determine the actual values ​​of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine; Step 1.4: Based on the near-infrared spectral data of the freshly cured tobacco leaf raw materials in Step 1.2 and the measured values ​​of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in Step 1.3, a near-infrared online quantitative model for the content of conventional chemical components in tobacco leaf raw materials is constructed through modeling methods such as partial least squares (PLS).

3. The method for real-time grading of flue-cured tobacco based on near-infrared online detection data according to claim 1, characterized in that: In step 2, based on the near infrared online quantitative model of the conventional chemical component content of tobacco raw materials, the conventional chemical component content of tobacco raw materials is The content of conventional chemical components of tobacco raw materials is predicted and calculated to obtain the predicted values ​​of conventional chemical components and derived chemical indicators of tobacco raw materials, and professional graders determine the appearance quality grade of tobacco raw materials based on the appearance of tobacco raw materials, such as color, etc., including the following steps: Step 2.1, using the near infrared online quantitative model of the conventional chemical components of tobacco raw materials, predict the conventional chemical components of new tobacco raw material samples, and obtain the predicted contents of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in the new tobacco raw material samples; Step 2.2, based on the predicted contents of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in the new tobacco leaf raw material samples, the derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, total nitrogen / nicotine in the new tobacco leaf raw material samples are calculated to obtain the derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, total nitrogen / nicotine in the new tobacco leaf raw material samples; Step 2.3: Professional graders determine the appearance quality grade of the new tobacco leaf raw material samples based on their appearance and color.

4. The method for real-time grading of flue-cured tobacco based on near-infrared online detection data according to claim 1, characterized in that: The step 3, establishing the connection between the appearance quality grade of tobacco leaf raw materials and the quantitative chemical index data, using information gain and Fisher criterion to select iterative algorithm, screening the chemical index data with established connection, and obtaining the key chemical index of tobacco leaf raw materials; comprises the following steps: Step 3.1, establish the connection between the tobacco leaf raw material appearance quality grade and the quantitative chemical index data, and obtain the fused tobacco leaf raw material appearance quality grade and quantitative chemical index data set; Step 3.2, calculate the information gain value of each chemical index in the fused tobacco leaf raw material appearance quality grade and quantitative chemical index data set, and obtain the information gain value of each chemical index of the tobacco leaf raw material. The calculation formula is as follows: Gain(D,A)=Ent(D)-Ent(D|A) Where K is the number of different levels; |D| represents the total number of samples; |C k |Indicates grade C k Number of samples Step 3.3, sort the information gain values ​​of the chemical indicators of tobacco raw materials from large to small, and select the first M chemical indicators with high information gain values ​​as the primary chemical indicators; Step 3.4, calculate the Fisher scores of the first M chemical indicators with high information gain values, and obtain the Fisher scores of the first M chemical indicators with high information gain values. The calculation formula is as follows: Where K is the number of different levels; |C k |Indicates grade C k The number of samples; m k Indicates level C k The average value of each index in the sample; m is the average value of each index in all samples; x n Indicates level C k The values ​​of each indicator Step 3.5, sort the Fisher scores of the first M chemical indicators with high information gain values ​​from large to small, and select N chemical indicators with high Fisher scores as a candidate subset; Step 3.6, let N = N + 1, repeat steps Step 3.3 to Step 3.6 until N is equal to M; Step 3.7, let M = M + 1, repeat Step 3.3 to Step 3.7 until M is equal to the number of chemical indicators, and obtain a series of chemical indicator subsets; Step 3.8, count the frequencies of different indicators in the series of chemical indicator subsets, and arrange the frequencies of each chemical indicator from high to low. The higher the frequency, the more critical the indicator is. Select several chemical indicators with higher frequencies; Step 3.9, using the Pearson correlation coefficient, eliminate the linear correlation between different chemical indicators in the series of chemical indicator subsets, and obtain several chemical indicators with high frequency after eliminating the linear correlation, which are used as the key chemical indicators for measuring tobacco leaf quality. The elimination formula is as follows: Where N is the total number of samples, is the content of the ith sample in index a, and It is the average value of the content of index α in all samples.

5. The method for real-time grading of flue-cured tobacco based on near-infrared online detection data according to claim 1, characterized in that: The Step 4, taking the key chemical indicators of tobacco leaf raw materials as a reference, statistically analyzing the key chemical indicator data of tobacco leaf raw materials over the years, and combining the statistical quantities of the key chemical indicator data of tobacco leaf raw materials over the years, formulates a tobacco leaf raw material quality chemical indicator grading table, including the following steps: Step 4.

1. Using the key chemical indicators of tobacco leaf raw materials as a reference, screen out the key chemical indicator data of tobacco leaf raw materials over the years, and statistically analyze the key chemical indicator data of tobacco leaf raw materials over the years to obtain the statistical quantities such as the average value, standard deviation and coefficient of variation of the key chemical indicator data of tobacco leaf raw materials over the years; Step 4.2, based on the statistics of key chemical index data of tobacco leaf raw materials over the years, determine the precise thresholds of key chemical indexes for each quality grade of tobacco leaf raw materials; Step 4.3: Based on the precise thresholds of key chemical indicators for each tobacco leaf raw material quality grade in Step 4.2, develop a tobacco leaf raw material quality chemical indicator grading table.

6. The method for real-time grading of flue-cured tobacco based on near-infrared online detection data according to claim 1, characterized in that: The step 5, obtaining near infrared spectral data of the tobacco leaf raw material sample, inputting the near infrared spectral data of the tobacco leaf raw material sample into the near infrared online quantitative model of the conventional chemical component content of the tobacco leaf raw material in step 2, obtaining the predicted values ​​of the key chemical indicators of the tobacco leaf raw material sample, comparing the predicted values ​​of the key chemical indicators of the tobacco leaf raw material sample with the tobacco leaf raw material quality chemical indicator grading table in step 4, and determining the quality grade of the tobacco leaf raw material, comprises the following steps: Step 5.1, using near infrared spectroscopy technology to scan the tobacco leaf raw material sample to obtain near infrared spectral data in the tobacco leaf raw material sample; Step 5.2, using the near infrared online quantitative model of the conventional chemical composition content of tobacco leaf raw materials in Step 2, predict the conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in tobacco leaf raw material samples and the derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, total nitrogen / nicotine in tobacco leaf raw material samples, and obtain the predicted values ​​of conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in tobacco leaf raw material samples and the predicted values ​​of derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, total nitrogen / nicotine in tobacco leaf raw material samples; Step 5.

3. Compare the predicted values ​​of key chemical indicators in the conventional chemical components such as total sugar, reducing sugar, total nitrogen, nicotine, moisture, potassium and chlorine in the tobacco raw material samples in Step 5.2 and the derived chemical indicators such as total sugar / total nitrogen, reducing sugar / total nitrogen, total sugar / nicotine, reducing sugar / nicotine, total nitrogen / nicotine in the tobacco raw material samples with the precise threshold values ​​of key chemical indicators for each tobacco raw material quality grade in the tobacco raw material quality chemical indicator grading table formulated in Step 4 to determine the quality grade of the tobacco raw material samples.

7. A real-time flue-cured tobacco grading device based on near-infrared online detection data, characterized in that: include: The data acquisition module is used to collect near-infrared spectral data of different types of tobacco raw materials and obtain the measured values ​​of conventional chemical components of different types of tobacco raw materials; The quantitative analysis module is connected with the data acquisition module. The quantitative analysis module is used to predict and calculate the content of conventional chemical components of tobacco raw materials based on the near-infrared spectral data of different types of tobacco raw materials, and determine the appearance quality grade of the primary tobacco raw materials based on the appearance of the tobacco raw materials, such as color, by professional graders; The key data screening module, the quantitative analysis module is connected with the key data screening module, and the key data screening module is used to screen the chemical index data related to the appearance quality grade of the tobacco leaf raw material to obtain the key chemical index of the tobacco leaf raw material; The grading reference module, the key data screening module is connected with the grading module, the grading module is used to take the key chemical indicators of tobacco leaf raw materials as a reference, to collect the key chemical indicator data of tobacco leaf raw materials over the years, and to formulate a grading table of tobacco leaf raw material quality chemical indicators in combination with the statistical quantities of the key chemical indicator data of tobacco leaf raw materials over the years; The automatic grading decision module is connected with the grading reference module, and the automatic grading decision module is used to obtain the predicted values ​​of key chemical indicators of tobacco leaf raw material samples, compare the predicted values ​​of key chemical indicators of tobacco leaf raw material samples with the tobacco leaf raw material quality chemical indicator grading table, and determine the quality grade of the tobacco leaf raw material samples.

Images