Evaluation method for tobacco shred blending uniformity

Abstract

The invention discloses a tobacco shred blending uniformity evaluation method. Comprising the following steps: A, respectively measuring the moisture content of each tobacco shred component by adopting a drying oven method, and establishing two-dimensional data sets of the moisture content of different types of tobacco shreds at different temperatures and different time; b, performing blending experiments of different tobacco shred blending gradients, and establishing tobacco shred moisture content prediction models of different types and different gradients; c, selecting cut tobacco on a cut tobacco production line as a sample, collecting spectral image data of cut tobacco and cut tobacco water loss, processing, inputting the data into the cut tobacco water content prediction model for prediction, performing contrast mapping on the data and the cut tobacco water content two-dimensional data set, evaluating the proportion of different types of cut tobacco, performing visualization or statistical analysis on the spectral image data, and determining the cut tobacco water content. The proportion and uniformity degree of cut tobacco components at different spatial positions are represented. Compared with the prior art, the evaluation method has higher quantitative and qualitative capabilities and higher detection precision and sensitivity, and realizes rapid and nondestructive detection.

Description

Technical Field

[0001] The present invention relates to the technical field of tobacco processing and detection, in particular to a method for evaluating tobacco blending uniformity. Background Art

[0002] Cigarette shred processing is a key process in cigarette production. It involves transforming tobacco leaves into qualified cut tobacco through a series of steps, based on the physical and chemical properties of the raw tobacco leaves and following a defined procedure. Within the shred production process, proportional blending involves the accurate and uniform blending of raw materials such as cut stems, expanded tobacco, and thin shredded tobacco into the dried cut tobacco leaves according to the designed recipe. This is a critical step in achieving consistent content of all tobacco components in the product formula, and this uniformity directly impacts the inherent quality and stability of the cigarette product. Therefore, mastering and controlling the precision of cigarette formulations is crucial to improving cigarette product quality. The uniformity of raw material formulations and the stability of cigarette products have become key issues that the tobacco industry urgently needs to address.

[0003] Tobacco blending is different from the mixing of general solid materials. Its blending ingredients include leaf cuts, stem cuts, expanded tobacco cuts, and thin slices. Because the chemical composition of tobacco cuts is similar, the existing technology has certain limitations and defects. For example, "Determination of tobacco blending uniformity" (YC / T 426-2012) uses the characteristic values constructed by total sugar, total alkaloids, and potassium in tobacco cuts as indicators for evaluating tobacco blending uniformity; this method has poor timeliness, limited scope of application, and cannot know the proportion of each component. Patent "A method for determining the uniformity of tobacco cuts, stem cuts, and reconstituted tobacco leaf formula cuts" with authorization announcement number CN106896032B is a method for characterizing the uniformity of tobacco cut blending based on the difference in cellulose and total sugar content. This method requires more chemical component testing, is time-consuming, and has high uncertainty. The patent "A method for detecting the blending ratio of various components in tobacco" with authorization announcement number CN108732127B uses a near-infrared model based on comprehensive chemical factor scores to detect the blending ratio of tobacco components. This method still requires further processing of the test samples and does not achieve non-destructive testing.

[0004] Therefore, there is an urgent need for a method for detecting tobacco blending uniformity that is fast, non-destructive, and has higher detection accuracy and sensitivity. Summary of the Invention

[0005] The object of the present invention is to provide a method for evaluating the blending uniformity of cut tobacco, which can perform rapid and non-destructive detection of cut tobacco and has higher detection accuracy and sensitivity.

[0006] The present invention provides a method for evaluating tobacco blending uniformity, comprising the following steps:

[0007] A. Based on the desired tobacco component types in cigarette products, the moisture content of each tobacco component was measured using the oven drying method. A two-dimensional dataset of moisture content of tobacco of different types at different temperatures and times was established.

[0008] B. Based on the desired tobacco component types in cigarette products, blending experiments were conducted at different tobacco blending gradients. Spectral image data of the blended samples was collected, and the average spectrum of the tobacco after noise reduction was extracted. The moisture content of different regions of the blended samples was measured using the oven method. The average moisture content was used as the true value. After processing the spectral image data, prediction models for the moisture content of tobacco of different types and gradients were established.

[0009] C. Selecting tobacco that has been blended and mixed on the tobacco production line as a sample, collecting spectral image data of the tobacco, and conducting a water loss test on the tobacco using an oven method, and then collecting spectral image data of the water loss of the tobacco; preprocessing and feature extraction of the collected spectral image data, and inputting the spectral image data into the tobacco moisture content prediction model to predict the moisture content of the tobacco, and simultaneously performing a comparison and mapping with the two-dimensional tobacco moisture content data set; evaluating the proportions of different types of tobacco components based on the time, temperature, and rate values of the tobacco water loss, and visualizing or statistically analyzing the spectral image data to characterize the proportions and uniformity of tobacco components at different spatial locations.

[0010] Furthermore, the shredded tobacco components in step A are shredded tobacco leaves, shredded stems, expanded shredded tobacco or shredded tobacco sheets.

[0011] Furthermore, the specific process of step A is as follows:

[0012] A1. Place different tobacco components in a constant temperature and humidity chamber with the following parameters: 70% relative humidity, 25±2°C, and 48 hours of equilibration.

[0013] A2. The moisture content of different tobacco components was then measured using an oven method. The oven settings were: temperature control accuracy ±1°C, temperature uniformity ±1°C, temperature no higher than 40°C, and a temperature gradient of 5°C. A set of tobacco moisture contents was obtained every 10 minutes. A two-dimensional dataset of tobacco moisture contents of different types and temperatures, at different times, was established.

[0014] Furthermore, the specific process of step B is as follows:

[0015] B1. Conduct blending experiments with different tobacco blending gradients based on the desired tobacco component types in cigarette products. First, place the different tobacco components in a constant temperature and humidity chamber with the following parameters: 70% relative humidity, 25±2°C, and 48 hours of equilibration.

[0016] B2. Turn on the spectral imaging device and preheat it. Use a black and white calibration plate to reduce interference from light and dark current. Set the acquisition wavelength range to 960-1700nm, the spectral resolution to 10nm, and the number of spectral bands to 139. The black and white calibration plate formula is as follows:

[0017]

[0018] O is the original data, B is the all-black calibration image obtained by scanning the built-in blackboard, and W is the all-white calibration image obtained by scanning a standard white calibration plate with a reflectivity of 99%;

[0019] B3. Experiments were conducted according to the set blending ratio. Different types of tobacco components were laid flat on a conveyor belt and manually mixed. Spectral image data was collected during the mixing process, and the average spectrum of the tobacco after noise reduction was extracted. The coefficient of variation of the tobacco in different bands was statistically analyzed. Manual mixing was terminated until the coefficient of variation of the tobacco remained constant. Spectral image data of the blended sample was measured. The moisture content of the tobacco in different regions of the blended sample was measured using the oven method. The moisture content of the tobacco in different regions was obtained, and the average moisture content was used as the true value.

[0020] B4. Use the continuous projection algorithm to perform data dimension reduction on the denoised spectral image dataset;

[0021] B5. Partial least squares regression is used to model the spectral image data after dimension reduction, and a prediction model for the moisture content of tobacco of different types and blending gradients is established.

[0022] Furthermore, the weight gradient set in the blending test in step B1 is 2%.

[0023] Furthermore, the noise reduction process in step B3 is as follows: based on the collected spectral image data, the spectral reflectance in different bands is compared and analyzed, and according to the difference in spectral reflectance in a specific band, an appropriate threshold is set to perform threshold segmentation, and the spectral image information containing only the tobacco is extracted through the mask image, and the spectrum is averaged; the extracted average spectral data is preprocessed using wavelet transform or SG filtering to perform noise reduction on the data. The threshold segmentation algorithm formula is:

[0024]

[0025] x and y are the coordinates of the pixel in the image, f λ (x,y) is the original image, ɡ(x,y) is the mask image after segmentation, and T is the segmentation threshold.

[0026] Furthermore, the specific process of step B4 is as follows: by selecting the initial variable with the largest amount of information, constructing the projection matrix, calculating and selecting the variable with the largest projection length and updating the set until the selected variable reaches the set value, a set of optimal variables is gradually screened out; by selecting the characteristic wavelength and optimizing the model parameters, the redundancy of the model can be reduced, the model efficiency can be improved, and the key spectral features can be retained at the same time, thereby obtaining the characteristic band, characteristic band coefficient and bias of the model.

[0027] Furthermore, the specific process of step B5 is as follows: by extracting latent variables that are representative of the original variables and explanatory of the dependent variables in the independent variable space and the dependent variable space, and then iteratively extracting the latent variables, gradually establishing the regression relationship between the principal components of the independent variables and the principal components of the dependent variables, and back-projecting the results back to the original variable space, finally establishing tobacco moisture content prediction models of different types and gradients.

[0028] Furthermore, the specific process of step C is as follows:

[0029] C1. Select the blended tobacco from the tobacco production line as a sample, spread it evenly on a tray, and use spectral imaging equipment to collect spectral image data of the tobacco;

[0030] C2. Perform a water loss test on the sample in C1 using the oven method, and then use a spectral imaging device to collect spectral image data of the water loss of the tobacco;

[0031] C3 preprocesses and extracts features from the spectral image data collected in steps C1 and C2 according to the noise reduction process in step B3 and the dimensionality reduction process in step B4;

[0032] C4. Input the data from step C3 into the tobacco moisture content prediction model, and then predict the tobacco moisture content pixel by pixel, and compare and map it with the two-dimensional tobacco moisture content dataset;

[0033] C5. Integrate the results of step C4 pixel by pixel, based on the time, temperature, and rate of dehydration of the tobacco, with an error of 0.5%, to evaluate the proportions of different tobacco components. Visualize or statistically analyze the spectral image data to characterize the proportions and uniformity of tobacco at different spatial locations.

[0034] Furthermore, the visualization in step C5 is to generate a pseudo-color image, and the statistical analysis includes the variance, coefficient of variation and mean absolute percentage error of the moisture content of the cut tobacco.

[0035] In summary, the present invention has the following advantages:

[0036] The method for evaluating tobacco blending uniformity provided by this invention leverages the varying dehydration rates of different tobacco components and employs a high-precision prediction model for tobacco moisture content based on spectral imaging. This model, based on the differences in dehydration rates and moisture distribution of tobacco at different temperatures, time periods, and blending ratios, enables rapid detection of tobacco blending ratios and spatial distribution. Compared to other methods, such as near-infrared (NIR) technology, digital imaging, and terahertz (THz), this method offers stronger quantitative and qualitative capabilities, higher detection accuracy and sensitivity, and improved efficiency, enabling rapid, non-destructive testing. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0038] Figure 1 A flow chart of the evaluation method provided by the present invention;

[0039] Figure 2 The two-dimensional dataset of moisture content of cut tobacco at different temperatures and times established in step A2 of the embodiment of the present invention;

[0040] Figure 3 This is the threshold segmentation of the spectral image in step B3 of the embodiment of the present invention: the left image is a pseudo-color image of the tobacco, the middle image is the spectral curve of the tobacco and the background, and the right image is the threshold segmentation processing image;

[0041] Figure 4 This is a graph of spectral data after noise reduction using different preprocessing methods in step B3 of the embodiment of the present invention;

[0042] Figure 5 This is a pixel-by-pixel distribution diagram of the moisture content of the tobacco in step C5 in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present application belongs.

[0044] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular also includes the plural. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0045] The following will clearly and completely describe the technical solutions of the present invention in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] Example

[0047] A method for evaluating the uniformity of tobacco blending, such as Figure 1 As shown, the specific process is as follows:

[0048] A. Establish a two-dimensional data set of moisture content of tobacco of different types at different temperatures and times

[0049] A1. For each desired tobacco component (leaf, stem, expanded tobacco, or flake) in a cigarette product, place the different tobacco components onto trays and place them in a constant temperature and humidity chamber. The parameters are set to: 70% relative humidity, 25±2°C, and 48 hours of equilibration.

[0050] A2. The moisture content of different types of cut tobacco components was determined using the oven method (YC / T 31-1996). The oven parameters were: temperature control accuracy ±1°C, temperature uniformity ±1°C, temperature not exceeding 40°C, and a temperature gradient of 5°C. A set of cut tobacco moisture content was obtained every 10 minutes. A two-dimensional dataset of cut tobacco moisture content at different temperatures, times, and types was established (e.g. Figure 2 As shown, Figure 2 Only the two-dimensional dataset of the moisture content of the tobacco leaves is shown). The formula for measuring the moisture content of the tobacco leaves using the oven method is:

[0051]

[0052] Among them, m1 is the mass of the sample tobacco and glass dish before drying, m2 is the mass of the sample tobacco and glass dish after drying, and m3 is the mass of the glass dish.

[0053] B. Establish a prediction model for moisture content of tobacco of different types and gradients through blending experiments and spectral image data acquisition and processing technology

[0054] B1. Based on the desired tobacco components in cigarette products, different tobacco blending gradients (weight gradient set to 2%) were conducted. Before the experiment, the tobacco was spread onto trays (500g of each sample was taken) and placed in a constant temperature and humidity chamber. The ambient parameters were: 70% relative humidity, 25±2°C, and 48 hours of equilibration.

[0055] B2. Turn on the spectral imaging device and preheat it. The spectral imaging device in this invention uses a hyperspectral acquisition system with a collection wavelength range of 960-1700nm, a spectral resolution of 10nm, and 139 spectral bands. Before acquisition, turn on the spectral imaging device and preheat it to ensure that the detection system is in good condition. A black and white plate calibration is used to reduce interference from light and dark current. The formula is as follows:

[0056]

[0057] Wherein: O is the original data, B is the all-black calibration image obtained by scanning the built-in blackboard, and W is the all-white calibration image obtained by scanning a standard white calibration plate with a reflectivity of 99%.

[0058] B3. The experiment was conducted according to the set blending ratio. The tobacco was spread flat on a conveyor belt with a width of 55 cm and a sample thickness of 2-3 cm. The tobacco was then manually mixed. Spectral image data was repeatedly measured during the mixing process. The average spectrum of the tobacco after noise reduction was extracted. The coefficient of variation of the tobacco in different bands was statistically analyzed. Manual mixing was terminated until the coefficient of variation of the tobacco remained unchanged. Spectral image data was measured for five randomly selected blended samples within the conveyor belt space for modeling research. The moisture content of the tobacco in different regions of the blended samples was measured using the oven method. The moisture content of the different regions of the blended samples was obtained, and the average moisture content was used as the true value.

[0059] The specific process of noise reduction is as follows: Based on the collected hyperspectral imaging data, the spectral reflectance in different bands is compared and analyzed. According to the difference in spectral reflectance in a specific band, an appropriate threshold is set and threshold segmentation is performed. The spectral image information containing only the tobacco is extracted through the mask image (such as Figure 3 As shown, the band position is 20, the threshold is 0.2), and the spectrum is averaged; the threshold segmentation algorithm formula is:

[0060]

[0061] Among them, x and y are the coordinates of the pixel in the image, f λ (x,y) is the original image, ɡ(x,y) is the mask image after segmentation, and T is the segmentation threshold;

[0062] For the extracted average spectral data, wavelet transform, SG filtering and other methods are used for preprocessing and noise reduction, which can improve the accuracy of feature extraction and regression. Figure 4 As shown, the preprocessing methods in the first row from left to right are the first-order derivative (1 st D), second-order derivative (2 nd D) Standard normal variable transformation. The preprocessing methods in the second row from left to right are SG filtering, multivariate scatter correction, and wavelet transform. The preprocessing methods in the third row from left to right are moving average smoothing, mean centering, and maximum and minimum normalization.

[0063] B4. A continuous projection algorithm is used to reduce the dimensionality of the spectral image dataset. Initial variables with the highest information content are selected, a projection matrix is constructed, and the variable with the longest projection length is calculated and selected. This set of variables is updated until the selected variable reaches the set value, gradually selecting an optimal set of variables. By selecting characteristic wavelengths and optimizing model parameters, model redundancy can be reduced, improving model performance while retaining key spectral features. This results in the model's characteristic bands, characteristic band coefficients, and biases.

[0064] B5. Partial least squares regression is used to model the dimensionality-reduced spectral image data. This model extracts latent variables that are representative of the original variables and explanatory of the dependent variables from the independent and dependent variable spaces. It then iteratively extracts these latent variables, gradually establishing a regression relationship between the principal components of the independent and dependent variables. The results are then back-projected back into the original variable space, ultimately establishing prediction models for moisture content of different tobacco types and gradients.

[0065] C. Select tobacco from the tobacco production line for measurement and analyze the measurement data

[0066] C1. Select the blended tobacco from the tobacco production line as a sample, spread it evenly on a tray, and use spectral imaging equipment to collect spectral image data of the tobacco.

[0067] C2. Perform a water loss test on the tobacco cut in step C1 using an oven method, and then use a spectral imaging device to collect spectral image data of the water loss of the tobacco cut.

[0068] C3. Preprocess and extract features from the spectral images collected in steps C1 and C2 according to the noise reduction method in step B3 and the dimensionality reduction method in step B4.

[0069] C4. Input the data from step C3 into the tobacco moisture content prediction model from step B5, and then predict the tobacco moisture content pixel by pixel, and compare and map it with the two-dimensional tobacco moisture content dataset from step A2.

[0070] C5. Integrate the results of step C4 pixel by pixel based on the time, temperature, and rate of dehydration of the cut tobacco, with an error of 0.5%, to evaluate the proportions of different types of cut tobacco (leaf cut, stem cut, expanded cut, and thin cut tobacco). Visualize the spectral image data (pseudo-color processing) or perform statistical analysis (variance, coefficient of variation, and mean absolute percentage error of the moisture content of the cut tobacco) to characterize the proportion and uniformity of the cut tobacco at different spatial locations. Figure 5 As shown, Figure 5 This is a pixel-by-pixel distribution diagram of the moisture content of tobacco after pseudo-color processing. The moisture content increases in the direction of the arrow.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating tobacco blending uniformity, characterized in that: The following steps are involved: A. Based on the desired tobacco component types in cigarette products, the moisture content of each tobacco component was measured using the oven drying method. A two-dimensional dataset of moisture content of tobacco of different types at different temperatures and times was established. B. Conduct blending experiments with different tobacco blending gradients based on the desired tobacco component types in cigarette products. Collect spectral image data of the blended samples and extract the average spectrum of the tobacco after noise reduction. Simultaneously, measure the moisture content of different regions of the blended samples using the oven method. The average moisture content is used as the true value. After processing the spectral image data, a prediction model for moisture content of tobacco of different types and gradients is established; C. Selecting tobacco that has been blended and mixed on the tobacco production line as a sample, collecting spectral image data of the tobacco, and conducting a water loss test on the tobacco using an oven method, and then collecting spectral image data of the water loss of the tobacco; preprocessing and feature extraction of the collected spectral image data, and inputting the spectral image data into the tobacco moisture content prediction model to predict the moisture content of the tobacco, and simultaneously performing a comparison and mapping with the two-dimensional tobacco moisture content data set; evaluating the proportions of different types of tobacco components based on the time, temperature, and rate values of the tobacco water loss, and visualizing or statistically analyzing the spectral image data to characterize the proportions and uniformity of tobacco components at different spatial locations.

2. The method for evaluating tobacco blending uniformity according to claim 1, characterized in that: The tobacco components in step A include leaf shreds, stem shreds, expanded shreds or thin shreds.

3. The method for evaluating tobacco blending uniformity according to claim 1, characterized in that: The specific process of step A is as follows: A1. Place different tobacco components in a constant temperature and humidity chamber with the following parameters: 70% relative humidity, 25±2°C, and 48 hours of equilibration. A2. The moisture content of different tobacco components was then measured using an oven method. The oven settings were: temperature control accuracy ±1°C, temperature uniformity ±1°C, temperature no higher than 40°C, and a temperature gradient of 5°C. A set of tobacco moisture contents was obtained every 10 minutes. A two-dimensional dataset of tobacco moisture contents of different types and temperatures, at different times, was established.

4. The method for evaluating tobacco blending uniformity according to claim 1, characterized in that: The specific process of step B is as follows: B1. Conduct blending experiments with different tobacco blending gradients based on the desired tobacco component types in cigarette products. First, place the different tobacco components in a constant temperature and humidity chamber with the following parameters: 70% relative humidity, 25±2°C, and 48 hours of equilibration. B2. Turn on the spectral imaging device and preheat it. Use a black and white calibration plate to reduce interference from light and dark current. Set the acquisition wavelength range to 960-1700 nm, the spectral resolution to 10 nm, and the number of spectral bands to 139. B3. Conduct the experiment according to the set blending gradient. Different types of tobacco components are laid flat on a conveyor belt and manually mixed. Spectral image data is collected during the mixing process, and the average spectrum of the tobacco after noise reduction is extracted. The coefficient of variation of the tobacco in different bands is statistically analyzed. Manual mixing is terminated until the coefficient of variation of the tobacco remains unchanged. Spectral image data of the blended sample is collected. Simultaneously, the moisture content of the tobacco in different regions of the blended sample is measured using the oven method. The moisture content of the tobacco in different locations is obtained, and the average moisture content is used as the true value. B4. Use the continuous projection algorithm to perform data dimension reduction on the denoised spectral image dataset; B5. Partial least squares regression is used to model the spectral image data after dimension reduction, and a prediction model for the moisture content of tobacco of different types and blending gradients is established.

5. The method for evaluating tobacco blending uniformity according to claim 4, characterized in that: The weight gradient set in the blending test in step B1 is 2%.

6. The method for evaluating tobacco blending uniformity according to claim 4, characterized in that: The noise reduction process in step B3 is as follows: based on the collected spectral image data, the spectral reflectance in different bands is compared and analyzed, and according to the difference in spectral reflectance in a specific band, an appropriate threshold is set to perform threshold segmentation, and the spectral image information containing only the tobacco is extracted through the mask image, and spectral averaging is performed. The extracted average spectral data is preprocessed using wavelet transform or SG filtering to perform noise reduction operations on the data.

7. The method for evaluating tobacco blending uniformity according to claim 6, characterized in that: The specific process of step B4 is as follows: by selecting the initial variable with the largest amount of information, constructing the projection matrix, calculating and selecting the variable with the largest projection length and updating the set until the selected variable reaches the set value, a set of optimal variables is gradually screened out; By selecting characteristic wavelengths and optimizing model parameters, the redundancy of the model can be reduced, the model efficiency can be improved, and the key spectral features can be retained, thereby obtaining the characteristic bands, characteristic band coefficients and bias of the model.

8. The method for evaluating tobacco blending uniformity according to claim 4, characterized in that: The specific process of step B5 is as follows: by extracting latent variables that are representative of the original variables and explanatory of the dependent variables in the independent variable space and the dependent variable space, and then iteratively extracting the latent variables, gradually establishing the regression relationship between the principal components of the independent variables and the principal components of the dependent variables, and back-projecting the results back to the original variable space, finally establishing tobacco moisture content prediction models of different types and gradients.

9. The method for evaluating tobacco blending uniformity according to claim 7, characterized in that: The specific process of step C is as follows: C1. Select the blended tobacco from the tobacco production line as a sample, spread it evenly on a tray, and use spectral imaging equipment to collect spectral image data of the tobacco; C2. The water loss test of the tobacco in step C1 is performed using an oven method, and then spectral imaging equipment is used to collect spectral image data of the water loss of the tobacco; C3 preprocesses and extracts features from the spectral image data collected in steps C1 and C2 according to the noise reduction process in step B3 and the dimensionality reduction process in step B4; C4. Input the data from step C3 into the tobacco moisture content prediction model, and then predict the tobacco moisture content pixel by pixel, and compare and map it with the two-dimensional tobacco moisture content dataset; C5. Integrate the results of step C4 pixel by pixel, based on the time, temperature, and rate of dehydration of the tobacco, with an error of 0.5%, to evaluate the proportions of different tobacco components. Visualize or statistically analyze the spectral image data to characterize the proportions and uniformity of tobacco at different spatial locations.

10. The method for evaluating tobacco blending uniformity according to claim 9, characterized in that: The visualization in step C5 is to generate a pseudo-color image, and the statistical analysis includes the variance, coefficient of variation and mean absolute percentage error of the moisture content of the cut tobacco.

Images