Tobacco leaf location detection and identification method

Abstract

The invention relates to the technical field of chemometrics and molecular spectrum analysis, in particular to a tobacco leaf location detection and identification method, which comprises the following steps: collecting a batch of raw tobacco samples to be processed, averagely dividing the raw tobacco samples into three locations, namely a leaf tip location, a leaf middle location and a leaf base location, respectively scanning a sample spectrum of each location, pre-treating the sample spectrums, and determining the location of the tobacco leaf according to the pre-treated sample spectrums; the three location spectrums are divided into three classification models in combination with a genetic algorithm and a fuzzy K-means algorithm, sample spectrums at all the air separation openings are collected and projected into the classification models for prediction, the threshing deviation location of each air separation opening is judged, follow-up threshing air separation can be finely controlled, and the tobacco redrying quality is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of chemometrics and molecular spectral analysis, and more specifically, to a tobacco leaf location detection and identification method. Background Art

[0002] The physical characteristics of tobacco leaves, such as moisture content, thickness and width, have a significant impact on the parameters of leaf threshing and air separation. This means that by carefully air-sorting tobacco leaves at different locations (leaf tip, leaf middle, leaf base), the threshing machine speed and air separation speed can be adjusted according to their physical characteristics, thereby optimizing the redrying process.

[0003] The three different locations of each tobacco leaf have significant differences in the distribution of chemical components, and these differences have an important impact on the quality and processing effect of the tobacco leaves. By accurately testing the leaf tip, middle and base of the tobacco leaves, the redrying process can be optimized according to the physical properties and chemical composition of the tobacco leaves, the yield and quality of the tobacco leaves can be improved, and the tobacco resources can be effectively utilized. This process not only involves the consideration of physical properties, but also includes the analysis of chemical composition and the evaluation of sensory quality. It is a key link for the redrying plant to improve the quality of tobacco leaves and optimize the production process.

[0004] In the existing redrying production process, the leaf beating process generally uses the image method to detect the leaf structure (large pieces, medium pieces, small pieces, and fragments) of the tobacco leaves, thereby controlling the leaf structure. The location (leaf tip, leaf middle, leaf base) of the tobacco leaves beaten at each air outlet is not identified, and data support cannot be provided for classified redrying.

[0005] The prior art discloses an automatic detection device and control method for the leaf structure of online tobacco leaves during threshing and redrying. The automatic detection device can intercept the leaves at the outlets of each air separator of the threshing unit in real time during the production process, perform online tobacco leaf imaging through two sets of imaging devices, namely visible light and X-ray, perform image analysis and processing, detect the leaf structure data and establish a data model, adjust the production parameters in real time according to the model, and guide the operation of production. Although the scheme detects the leaf structure at the air separator outlet, it does not distinguish and identify the tobacco leaves, cannot perform detailed air separation, and cannot guarantee the quality of tobacco leaf redrying. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for detecting and identifying tobacco leaf positions, which divides tobacco leaves into three positions: leaf tip, leaf middle and leaf base, and establishes spectral models for each position. The spectrum can be sampled and scanned at the wind separation port to determine the leaf threshing bias of each wind separation port, and the subsequent leaf threshing wind separation can be finely controlled to improve the quality of tobacco leaf redrying.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0008] A method for detecting and identifying tobacco leaf locations is provided, comprising the following steps:

[0009] S1: Collect n raw tobacco samples from the batch to be processed, divide the raw tobacco samples from the batch to be processed into three locations: leaf tip, leaf middle, and leaf base, and collect p samples from each location, and take Y slices from each sample;

[0010] S2: Superimpose the same samples on Y sheets, scan the spectrum of the samples in each area respectively, and scan the spectrum of the samples at the tip of the blade. 11 、x 12 ,……,x 1p}, scan the sample spectrum X2 in the leaf area = {x 21 、x 22 ,……,x 2p}, scan the sample spectrum of the leaf base X3 = {x 31 、x 32 ,……,x 3p}, save the spectra separately by location;

[0011] S3: preprocessing the spectra of the three location samples in step S2;

[0012] S4: Combine genetic algorithm and fuzzy K-means algorithm to classify the preprocessed spectra, and divide the spectra of the leaf tip, leaf middle and leaf base into three classification models, namely L 叶尖 , L 叶中 , L 叶基 ;

[0013] S5: Collecting sample spectra of each air outlet in the leaf threshing air separation process;

[0014] S6: Project the sample spectra of each air outlet in step S5 to the classification model of step S4 for prediction, count the locations corresponding to each sample of each air outlet, and determine the leaf beating bias location of each air outlet.

[0015] The tobacco leaf position detection and identification method of the present invention collects raw tobacco samples of a batch to be processed and divides them into three positions, namely, leaf tip, leaf middle and leaf base, respectively, scans the sample spectrum of each position, and after preprocessing the sample spectrum, divides the three position spectra into three classification models in combination with a genetic algorithm and a fuzzy K-means algorithm, collects sample spectra at each wind separation port, projects them into the classification model for prediction, and judges the leaf threshing bias position of each wind separation port, which can finely control the subsequent leaf threshing wind separation, improve the quality of tobacco leaf redrying, and effectively separate tobacco leaves of different qualities, thereby improving the original modularization and homogenization level of tobacco leaves and realizing refined management.

[0016] Preferably, in step S2 and step S5, a handheld near-infrared instrument is used to scan the spectrum of the sample.

[0017] Preferably, the spectra at the tip, middle and base of the leaf collected in step S2 have drift phenomenon, but because the handheld near-infrared instrument adopts a scanning method of directly attaching to the surface of the tobacco leaf and the external environment remains relatively stable, the complexity of the collected spectrum is low. Therefore, the first-order derivative preprocessing method is used in step S3 to improve the accuracy of the spectrum.

[0018] Preferably, in step S4, the process of classifying the preprocessed spectrum includes:

[0019] S401: Define the fitness function, set the crossover probability, population size, number of iterations b, mutation probability, number of classifications k, weighted index a and convergence threshold ε; X = {x 11 , x 12 , ..., x 1p ,x 21 , x 22 , ..., x 2p ,x 31 , x 32 , ..., x 3p}

[0020] S402: Randomly select m samples from n samples as the membership matrix B of the initial classification (0) , and calculate the center point Y of m samples 初始 ; Membership matrix B of the initial classification (0) The membership degree A of the samples in ij Requirements:

[0021]

[0022] In the formula, k represents the number of categories, i is one of the categories, A ij Indicates the membership of the jth sample of the i-th category to the i-th category;

[0023] S403: Calculate the distance between each remaining sample and the initial classification center of the m samples, and classify the samples with the closest distance into one category, so that the total number of samples is divided into k categories;

[0024] S404: Calculate the fitness of k types of samples;

[0025] S405: Calculate the mean of each type of sample as the new center point Y i , after selection, crossover, and mutation, the distance between each remaining sample and the center point is calculated, and it is reclassified into k categories according to the minimum error function Q;

[0026] S406: According to the number of iterations b and the classification center Y i , membership matrix B (0) , and obtain the membership matrix B after b iterations (b)and classification center Y i b ;

[0027] S407: Based on the obtained classification center Y i b and membership degree A ij , find the membership matrix B (b+1) ;

[0028] S408: If max{|B (b) -B (b+1) |}≤ε, then stop the iteration and output the membership matrix B (b+1) and the corresponding classification center Y i b , go to step S409; if not, return to step S402;

[0029] S409: In the obtained membership matrix B (b+1) , let the maximum element in each column be 1 and the rest be 0, and the best sample is the final fuzzy k-means 3-class classification model, which are L 叶尖 , L 叶中 , L 叶基 .

[0030] Preferably, in step S405, the error function calculation formula is:

[0031]

[0032] In the formula, k represents the number of classifications; n represents the total number of samples; A ij represents the membership of the jth sample of the i-th category to the i-th category; a represents the weighted index; x ij represents the spectrum of the jth sample of the i-th category; Y i represents the classification center of the i-th category; ‖x ij -Y i ‖ represents sample x ij With Classification Center Y i The Euclidean distance between them; Q represents the sample x ij With Classification Center Y i The weighted sum of squared distances;

[0033] The best classification is to minimize the error function Q, so we need to get the best classification and the best membership degree A. ij and classification center Y i ;

[0034] When a>1 and x ij ≠Y i When the membership degree A is calculated iteratively ij and classification center Y i :

[0035]

[0036] Preferably, in step S4, the spectra X1, X2, X3 of the three location samples are preprocessed and then projected into space, so that the distribution differences of the three locations can be observed.

[0037] Preferably, the classification centers of the three locations of leaf tip, leaf middle and leaf base are calculated respectively, each sample corresponds to an error function, and the error function is projected onto a two-dimensional plane. The samples in each location form an elliptical distribution, i.e., a confidence ellipse interval. The leaf tip, leaf middle and leaf base are classified by constructing and applying a confidence ellipse model.

[0038] Preferably, in step S5, the method for collecting the sample spectrum of each air separation port in the leaf beating process is: after the leaf beating and re-roasting production is stable, cross-sectional sampling is performed at the leaf beating air separation port, F samples are taken from each air separation port, each sample is evenly spread out, and the spectrum of each sample is scanned with a handheld near-infrared instrument.

[0039] Preferably, when cross-sectional sampling is performed at the leaf-beating wind outlet, 2.7 kg to 3.3 kg is sampled each time.

[0040] Preferably, the F samples are scanned Z times respectively, and the average spectrum of the Z spectra is taken as the sample spectrum, and F spectra are obtained from each wind outlet respectively.

[0041] Compared with the prior art, the tobacco leaf location detection and identification method of the present invention has the following beneficial effects:

[0042] The tobacco leaves are divided into three areas: leaf tip, leaf middle and leaf base, and spectral models are established for each area. The spectrum can be sampled and scanned at the wind outlet to determine the leaf threshing bias of each wind outlet, and the subsequent leaf threshing wind can be finely controlled to improve the quality of tobacco leaf re-roasting. It can also effectively separate tobacco leaves of different qualities, thereby improving the original modularization and homogenization level of the tobacco leaves and realizing refined management. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Flow chart of the tobacco leaf location detection and identification method in an embodiment of the present invention;

[0044] Figure 2 This is a schematic diagram of tobacco leaf location division in an embodiment of the present invention;

[0045] Figure 3 This is a schematic diagram of the spectrum of tobacco leaf samples in an embodiment of the present invention;

[0046] Figure 4 Schematic diagram of the spectrum of the tobacco leaf sample after pretreatment in the embodiment of the present invention;

[0047] Figure 5 is a location distribution diagram of tobacco leaf samples in an embodiment of the present invention;

[0048] Figure 6 This is the location ellipse distribution diagram of tobacco leaf samples in the embodiment of the present invention;

[0049] Figure 7 This is a quantitative diagram of the location distribution of tobacco leaf samples at each wind outlet in the embodiment of the present invention;

[0050] Figure 2 In the diagram, A represents the leaf base; B represents the middle of the leaf; and C represents the leaf tip. DETAILED DESCRIPTION

[0051] The present invention will be further described below in conjunction with specific implementation modes.

[0052] Embodiment 1

[0053] A tobacco leaf location detection and identification method, such as Figure 1 As shown, the following steps are included:

[0054] S1: Collect n raw tobacco samples from the batch to be processed, and divide the raw tobacco samples from the batch to be processed into three locations: leaf tip, leaf middle, and leaf base. Figure 2 As shown, there are p samples in each location, and Y slices are taken for each sample;

[0055] S2: Overlay the samples with the same Y slices and scan the sample spectrum in each area separately, such as Figure 3 As shown, the sample spectrum X1 of the scanning blade tip area is 11 、x 12 ,……,x 1p}, scan the sample spectrum X2 in the leaf area = {x 21 、x 22 ,……,x 2p}, scan the sample spectrum of the leaf base X3 = {x 31 、x 32 ,……,x 3p}, save the spectra separately by location;

[0056] S3: preprocessing the spectra of the three location samples in step S2;

[0057] S4: Combine genetic algorithm and fuzzy K-means algorithm to classify the preprocessed spectra, and divide the spectra of the leaf tip, leaf middle and leaf base into three classification models, namely L 叶尖 , L 叶中 , L 叶基 ;

[0058] S5: Collecting sample spectra of each air outlet in the leaf threshing air separation process;

[0059] S6: Project the sample spectra of each air outlet in step S5 to the classification model of step S4 for prediction, count the locations corresponding to each sample of each air outlet, and determine the leaf beating bias location of each air outlet.

[0060] The above-mentioned tobacco leaf position detection and identification method collects raw tobacco samples from the batch to be processed and divides them into three positions: leaf tip, leaf middle and leaf base. The sample spectrum of each position is scanned respectively. After preprocessing the sample spectrum, the three position spectra are divided into three classification models by combining genetic algorithm and fuzzy K-means algorithm. The sample spectrum at each wind outlet is collected and projected into the classification model for prediction. The leaf threshing bias position of each wind outlet is judged, which can finely control the subsequent leaf threshing wind fraction and improve the quality of tobacco leaf redrying.

[0061] In step S2 and step S5, a handheld near infrared instrument is used to scan the spectrum of the sample.

[0062] The spectra collected in step S2 at the tip, middle and base of the leaf have drift phenomenon. However, since the handheld near-infrared instrument adopts a scanning method that directly attaches to the surface of the tobacco leaf and the external environment remains relatively stable, the complexity of the collected spectrum is low. Therefore, the first-order derivative preprocessing method is used in step S3 to improve the accuracy of the spectrum, such as Figure 4 shown.

[0063] In step S5, the method for collecting the sample spectrum of each air separation port in the leaf beating process is as follows: after the leaf beating and re-roasting production is stable, cross-sectional sampling is performed at the leaf beating air separation port, F samples are taken from each air separation port, each sample is evenly spread, and the spectrum of each sample is scanned with a handheld near-infrared instrument.

[0064] Scan F samples Z times respectively, take the average spectrum of the Z spectra as the sample spectrum, and obtain F spectra from each wind outlet respectively.

[0065] Embodiment 2

[0066] This embodiment is similar to the first embodiment, except that in step S4, the process of classifying the preprocessed spectrum includes:

[0067] S401: define the fitness function, set the crossover probability, population size, number of iterations b, mutation probability, number of classifications k, weighted index a and convergence threshold ε;

[0068] S402: Randomly select m samples from n samples as the membership matrix B of the initial classification (0) , and calculate the center point Y of m samples 初始 ; Membership matrix B of the initial classification (0) The membership degree A of the samples in ij Requirements:

[0069]

[0070] In the formula, k represents the number of categories, i is one of the categories, A ij Indicates the membership of the jth sample of the i-th category to the i-th category;

[0071] S403: Calculate the distance between each remaining sample and the initial classification center of the m samples, and classify the samples with the closest distance into one category, so that the total number of samples is divided into k categories;

[0072] S404: Calculate the fitness of k types of samples;

[0073] S405: Calculate the mean of each type of sample as the new center point Y i , after selection, crossover, and mutation, the distance between each remaining sample and the center point is calculated, and it is reclassified into k categories according to the minimum error function Q;

[0074] Error function calculation formula:

[0075]

[0076] In the formula, k represents the number of classifications; n represents the total number of samples; A ij represents the membership of the jth sample of the i-th category to the i-th category; a represents the weighted index; x ij represents the spectrum of the jth sample of the i-th category; Y i represents the classification center of the i-th category; ‖x ij -Y i ‖ represents sample x ij With Classification Center Y i The Euclidean distance between them; Q represents the sample x ij With Classification Center Y i The weighted sum of squared distances;

[0077] Where a>1, to strengthen the sample x ij The contrast of each degree, the larger the value of a, the greater the fuzziness of the resulting classification matrix; generally the initial value of m is 1 or 2;

[0078] The best classification is to minimize the error function Q, so we need to get the best classification and the best membership degree A. ij and classification center Y i ;

[0079] When a>1 and x ij ≠Y i When the membership degree A is calculated iteratively ij and classification center Y i :

[0080]

[0081] S406: According to the number of iterations b and the classification center Y i , membership matrix B (0) , and obtain the membership matrix B after b iterations (b) and classification center Y i b ;

[0082] S407: Based on the obtained classification center Y i b and membership degree A ij , find the membership matrix B (b+1) ;

[0083] S408: If max{|B (b) -B (b+1) |}≤ε, then stop the iteration and output the membership matrix B (b+1) and the corresponding classification center Y i b , go to step S409; if not, return to step S402;

[0084] S409: In the obtained membership matrix B (b+1) , let the maximum element in each column be 1 and the rest be 0, and the best sample is the final fuzzy k-means 3-class classification model, which are L 叶尖 , L 叶中 , L 叶基 .

[0085] In step S4, the spectra of the three location samples X1, X2, and X3 are preprocessed and projected into space, such as Figure 5 As shown, the distribution differences among the three locations can be observed.

[0086] The classification centers of the three locations of leaf tip, leaf center, and leaf base are calculated respectively. Each sample corresponds to an error function. The error function is projected onto a two-dimensional plane. The samples in each location form an elliptical distribution, namely, the confidence ellipse interval, as shown in Figure 6 As shown, the leaf tip, leaf middle, and leaf base classification are performed by constructing and applying a confidence ellipse model.

[0087] Embodiment 3

[0088] This embodiment is similar to the second embodiment, except that in step S5, the leaf threshing and redrying equipment is provided with a plurality of air separation ports. Taking the first air separation port, the fourth air separation port, the seventh air separation port and the eleventh air separation port as examples, after the leaf threshing and redrying production is stabilized, cross-sectional sampling is performed at the first air separation port, the fourth air separation port, the seventh air separation port and the eleventh air separation port, 2.7 kg to 3.3 kg are taken each time, 50 samples are taken from each air separation port, each sample is evenly spread, and a handheld near-infrared instrument is used to scan the spectrum 60 times close to the tobacco leaves. The average spectrum of the 60 spectra is calculated for each sample as the sample spectrum, then 50 spectra are obtained from each of the four air separation ports, and the sample spectra of the first air separation port are x respectively. 片0101 、x 片0102 ,……,x 片0150 , the sample spectra of the fourth wind outlet are x 片0401 、x 片0402 ,……,x 片0450 The sample spectra of the seventh wind outlet are x 片0701 、x 片0702 ,……,x 片0750 , the sample spectra of the eleventh wind outlet are x 片1101 、x 片1102 ,……,x 片1150 .

[0089] The sample spectra of the first, fourth, seventh and eleventh wind ports are projected into three classification models for prediction, and the location of each sample of each wind port in the three positions of leaf tip, leaf middle and leaf base is counted, such as Figure 7 As shown, the different locations of leaf beating at each wind outlet lead to differences in the distribution of chemical components, which affects the differences in the final quality.

[0090] In the specific contents of the above-mentioned specific implementation methods, the various technical features can be combined in any non-contradictory manner. In order to make the description concise, not all possible combinations of the above-mentioned technical features are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0091] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A tobacco leaf location detection and identification method, characterized in that: The following steps are involved: S1: Collect n raw tobacco samples from the batch to be processed, divide the raw tobacco samples from the batch to be processed into three locations: leaf tip, leaf middle, and leaf base, and collect p samples from each location, and take Y slices from each sample; S2: Superimpose the same samples on Y sheets, scan the sample spectrum of each position respectively, and scan the sample spectrum of the blade tip position X1={x 11 、x 12 ,……,x 1p }, scan the sample spectrum X2 in the leaf area = {x 21 、x 22 ,……,x 2p }, scan the sample spectrum of the leaf base X3 = {x 31 、x 32 ,……,x 3p }, save the spectra separately by location; S3: preprocessing the spectra of the three location samples in step S2; S4: Combine genetic algorithm and fuzzy K-means algorithm to classify the preprocessed spectra, and divide the spectra of the leaf tip, leaf middle and leaf base into three classification models, namely L 叶尖 , L 叶中 , L 叶基 ; S5: Collecting sample spectra of each air outlet in the leaf threshing air separation process; S6: Project the sample spectra of each air outlet in step S5 to the classification model of step S4 for prediction, count the locations corresponding to each sample of each air outlet, and determine the leaf beating bias location of each air outlet.

2. The tobacco leaf location detection and identification method according to claim 1, characterized in that: In step S2 and step S5, a handheld near infrared instrument is used to scan the spectrum of the sample.

3. The tobacco leaf location detection and identification method according to claim 2, characterized in that: The spectra at the tip, middle and base of the leaf collected in step S2 have drift phenomenon. However, since the handheld near-infrared instrument adopts a scanning method that directly attaches to the surface of the tobacco leaf and the external environment remains relatively stable, the complexity of the collected spectra is low. Therefore, the first-order derivative preprocessing method is used in step S3 to improve the accuracy of the spectrum.

4. The tobacco leaf location detection and identification method according to claim 1, characterized in that: In step S4, the process of classifying the preprocessed spectrum includes: S401: Define the fitness function, set the crossover probability, population size, number of iterations b, mutation probability, number of classifications k, weighted index a and convergence threshold ε; X = {x 11 , x 12 , ..., x 1p ,x 21 , x 22 , ..., x 2p ,x 31 , x 32 , ..., x 3p } S402: Randomly select m samples from n samples as the membership matrix B of the initial classification (0) , and calculate the center point Y of m samples 初始 ; Membership matrix B of the initial classification (0) The membership degree A of the samples in ij Requirements: 0≤A ij ≤1; (i=1,2,……,k); In the formula, k represents the number of categories; i is one of the categories, A ij Indicates the membership of the jth sample of the i-th category to the i-th category; S403: Calculate the distance between each remaining sample and the initial classification center of the m samples, and classify the samples with the closest distance into one category, so that the total number of samples is divided into k categories; S404: Calculate the fitness of k types of samples; S405: Calculate the mean of each type of sample as the new center point Y i , after selection, crossover, and mutation, the distance between each remaining sample and the center point is calculated, and it is reclassified into k categories according to the minimum error function Q; S406: According to the number of iterations b and the classification center Y i , membership matrix B (0) , and obtain the membership matrix B after b iterations (b) and classification center Y i b ; S407: Based on the obtained classification center Y i b and membership degree A ij , find the membership matrix B (b+1) ; S408: If max{|B (b) -B (b+1) |}≤ε, then stop the iteration and output the membership matrix B (b+1) and the corresponding classification center Y i b , go to step S409; if not, return to step S402; S409: In the obtained membership matrix B (b+1) , let the maximum element in each column be 1 and the rest be 0, and the best sample is the final fuzzy k-means 3-class classification model, which are L 叶尖 , L 叶中 , L 叶基 .

5. The tobacco leaf location detection and identification method according to claim 4, characterized in that: In step S405, the error function calculation formula is: In the formula, k represents the number of classifications; n represents the total number of samples; A ij represents the membership of the jth sample of the i-th category to the i-th category; a represents the weighted index; x ij represents the spectrum of the jth sample of the i-th category; Y i represents the classification center of the i-th category; ‖x ij -Y i ‖ represents sample x ij With Classification Center Y i The Euclidean distance between them; Q represents the sample x ij With Classification Center Y i The weighted sum of squared distances; The best classification is to minimize the error function Q, so we need to get the best classification and the best membership degree A. ij and classification center Y i ; When a>1 and x ij ≠Y i When the membership degree A is calculated iteratively ij and classification center Y i :

6. The tobacco leaf location detection and identification method according to claim 5, characterized in that: In step S4, the spectra of the three location samples X1, X2, and X3 are preprocessed and projected into space, and the distribution differences of the three locations can be observed.

7. The tobacco leaf location detection and identification method according to claim 6, characterized in that: The classification centers of the three locations of leaf tip, leaf middle and leaf base are calculated respectively. Each sample corresponds to an error function, and the error function is projected onto a two-dimensional plane. The samples in each location form an elliptical distribution, namely the confidence ellipse interval. The leaf tip, leaf middle and leaf base are classified by constructing and applying the confidence ellipse model.

8. The tobacco leaf location detection and identification method according to any one of claims 1 to 7, characterized in that: In step S5, the method for collecting the sample spectrum of each air separation port in the leaf beating process is as follows: after the leaf beating and re-roasting production is stable, cross-sectional sampling is performed at the leaf beating air separation port, F samples are taken from each air separation port, each sample is evenly spread, and the spectrum of each sample is scanned with a handheld near-infrared instrument.

9. The tobacco leaf location detection and identification method according to claim 8, characterized in that: When taking cross-sectional samples at the leaf-beating wind outlet, take 2.7kg to 3.3kg each time.

10. The tobacco leaf location detection and identification method according to claim 8, characterized in that: Scan F samples Z times respectively, take the average spectrum of the Z spectra as the sample spectrum, and obtain F spectra from each wind outlet respectively.

Images