Method for determining spatial distribution of nicotine in reconstituted tobacco leaf based on hyperspectral imaging and r-r distribution model

Abstract

This invention discloses a method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the R-R distribution model. This method addresses the current industry challenge of non-destructively, rapidly, and visually quantifying the two-dimensional spatial distribution and uniformity of nicotine in reconstituted tobacco leaves. Specifically, near-infrared hyperspectral images of reconstituted tobacco leaf samples are acquired to obtain three-dimensional hyperspectral data. A chemometric model, combined with spectral preprocessing, is used to establish a quantitative inversion model for nicotine content, generating a two-dimensional distribution map of nicotine content. The R-R distribution model is then introduced into the nicotine distribution analysis, fitting the distribution map, extracting the cumulative distribution function, and calculating characteristic particle size parameters and coefficients of variation. Finally, a comprehensive uniformity index is constructed to quantitatively assess the spatial distribution uniformity of nicotine. This invention achieves a non-destructive, rapid, visual, and quantitative assessment of the spatial distribution of nicotine in reconstituted tobacco leaves, providing feasibility for homogenized production and intelligent processing scenarios in the tobacco industry.

Description

Technical Field

[0001] This invention relates to the field of optical response detection and tobacco quality analysis, and in particular to a non-contact, visual method for determining the spatial distribution of nicotine components in reconstituted tobacco leaves based on a combination of near-infrared hyperspectral imaging technology and the Rosin-Rahmler (RR) distribution model. This method falls under the category of intelligent quality detection and raw material homogenization characterization technology in the tobacco industry. Background Technology

[0002] Reconstituted tobacco, also known as tobacco sheet, is an indispensable key raw material in the cigarette industry. Its uniformity and stability directly affect the sensory quality and smoking experience of cigarettes. Among numerous quality indicators, nicotine, as the core functional component providing physiological satisfaction, is important in terms of its appropriate content, but even more crucial is its uniform and stable distribution within the two-dimensional space of the sheet. Uneven nicotine distribution will directly lead to significant differences in release behavior in different parts of a single cigarette, causing fluctuations in smoking taste and batch-to-batch quality instability. This severely restricts the consistency and controllability of product quality and is one of the core pain points in the tobacco industry's efforts to promote homogenized processing.

[0003] For a long time, the industry has heavily relied on wet chemical methods such as flow analysis for nicotine detection. For example, the industry has proposed using near-infrared spectroscopy to detect the physicochemical indicators of reconstituted tobacco extracts and tobacco pastes produced in the papermaking process. By collecting the near-infrared transmission spectra of liquid or paste samples and combining them with partial least squares methods to establish a quantitative model, rapid detection of nicotine and other physicochemical indicators has been achieved. However, this method has the following substantial limitations: First, the detection objects are intermediates in the production process (extracts, tobacco pastes), and the detection results represent the average chemical composition of the batch of materials, which cannot directly reflect the final product, the distribution of nicotine on the solid reconstituted tobacco leaf substrate; Second, this method is essentially an average spectral analysis of "points" or "homogenized samples," which can only provide a macroscopic average content and cannot provide information on the distribution of nicotine in two-dimensional space. This makes it impossible to accurately locate distribution defects caused by processes such as coating and drying during production, and even more impossible to target and optimize key processes.

[0004] In recent years, the development of near-infrared spectroscopy imaging technology has made it possible to acquire information on the spatial distribution of materials. However, existing technologies mostly remain at the level of intuitive interpretation of hyperspectral images or simple statistical descriptions, lacking a model and method capable of accurately mathematically representing and comprehensively quantifying the spatial distribution of chemical components. Therefore, facing the urgent need of the tobacco industry for homogenization control of raw materials, there is a pressing need to develop a solution that can rapidly, non-destructively, and quantitatively visualize the spatial distribution of nicotine content in reconstituted tobacco leaves, and accurately mathematically describe and comprehensively evaluate its uniformity. Summary of the Invention

[0005] In view of the above, the present invention aims to provide a method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and RR distribution model, so as to overcome the defects of existing methods for determining the distribution of nicotine in reconstituted tobacco leaves, such as strong destructiveness, only providing average content, and inability to quantitatively evaluate the two-dimensional spatial distribution and its uniformity.

[0006] The technical solution adopted in this invention is as follows:

[0007] This invention provides a method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the RR distribution model, including:

[0008] Obtain a three-dimensional hyperspectral data cube of reconstituted tobacco leaf samples;

[0009] Based on the three-dimensional hyperspectral data cube, combined with the pre-established quantitative inversion model of nicotine content, a two-dimensional distribution map of nicotine content in the sample is generated.

[0010] Each pixel in the two-dimensional distribution map is regarded as a particle, and its nicotine content value is regarded as the particle size. Based on the cumulative distribution function, the Rosin-Rahmler distribution model is used to fit the nicotine content distribution and calculate the distribution characteristic parameters. The distribution characteristic parameters represent the nicotine content value corresponding to the cumulative distribution reaching a preset percentage.

[0011] The coefficient of variation of nicotine content at all pixels in the two-dimensional distribution map is calculated, and a uniformity index is constructed in combination with the distribution characteristic parameters to quantitatively evaluate the spatial distribution uniformity of nicotine in the reconstituted tobacco sample.

[0012] In at least one of the possible implementation methods, the process of establishing a quantitative inversion model for nicotine content includes:

[0013] Obtain hyperspectral data of reconstituted tobacco calibration set samples with nicotine content reference values;

[0014] Hyperspectral data are preprocessed to remove noise and baseline drift;

[0015] Partial least squares regression or support vector machine is used to correlate the preprocessed hyperspectral data with the reference value of nicotine content to construct a quantitative inversion model.

[0016] In at least one of the possible implementations, the preprocessing of hyperspectral data includes at least: multivariate scattering correction and first derivative processing.

[0017] In at least one of the possible implementations, the three-dimensional hyperspectral data cube is acquired by non-contact scanning of the reconstituted tobacco sample in the wavelength range of 1000-2000 nm using a near-infrared hyperspectral imaging system.

[0018] In at least one of the possible implementations, the Rosin-Rahmler distribution model is used for fitting, as follows:

[0019]

[0020] Where x is the nicotine content of a pixel, F(x) is the percentage of the total nicotine content of pixels with a nicotine content less than or equal to x, and A and B are model parameters.

[0021] In at least one of the possible implementations, the coefficient of variation is calculated as follows:

[0022]

[0023] Where CV is the coefficient of variation of nicotine content at full pixel count; σ is the standard deviation of nicotine content at full pixel count; x mean This is the arithmetic mean of nicotine content across all pixels.

[0024] In at least one of the possible implementations, the distribution characteristic parameters are assumed to include: D10, D50, and D90; wherein, D10, D50, and D90 represent the nicotine content values ​​corresponding to the cumulative distribution reaching 10%, 50%, and 90%, respectively.

[0025] The uniformity index U is calculated as follows:

[0026]

[0027] in, The ratio of the distribution center to the distribution width represents the central tendency of the distribution; (1-CV) represents the penalty factor for the overall degree of variation.

[0028] In at least one possible implementation, the determination method further includes: determining the uniformity level of the spatial distribution of nicotine in reconstituted tobacco leaves based on the value of the uniformity index U, wherein a larger U value indicates a more uniform nicotine distribution.

[0029] In at least one possible implementation, the determination method further includes: after establishing a quantitative inversion model for nicotine content, evaluating the prediction accuracy and generalization ability of the model using the coefficient of determination R², root mean square error RMSE, and relative prediction deviation RPD.

[0030] Compared with existing technologies, this invention applies hyperspectral imaging technology to determine the spatial distribution of nicotine in reconstituted tobacco solid products. This overcomes the shortcomings of traditional methods, such as high destructiveness and inability to obtain spatial information, and achieves visualization of nicotine distribution, providing an intuitive basis for process diagnosis. In terms of technical conception, unlike existing near-infrared technology which only provides the average content of samples, this invention can acquire and characterize the distribution information of nicotine in a two-dimensional plane, providing a new dimension of data support for evaluating product homogeneity. Specifically, the Rosin-Rahmler distribution model, traditionally used for physical particle size analysis, is innovatively introduced into the characterization of the spatial distribution of chemical components. Through the design of parameters such as D10, D50, and D90, a precise mathematical description of the nicotine distribution morphology is achieved. Furthermore, by constructing a uniformity index U that includes the distribution center, distribution width, and overall variation, the complex two-dimensional distribution information is quantified into a simple and intuitive value, providing a clear quantitative basis for product quality grading and production process optimization. This is of great significance for promoting tobacco processing from experience-based control to digital, intelligent, and precise regulation. Attached Figure Description

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described below with reference to the accompanying drawings, wherein:

[0032] Figure 1 A schematic flowchart of the method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and RR distribution model provided in an embodiment of the present invention;

[0033] Figure 2 A physical image of reconstituted tobacco leaves provided in an embodiment of the present invention;

[0034] Figure 3 The original reflectance spectra of different reconstituted tobacco leaf samples in the wavelength range of 1000-2000 nm provided in the embodiments of the present invention;

[0035] Figure 4 The above are the spectra of different reconstituted tobacco leaf samples after multivariate scattering correction preprocessing provided in the embodiments of the present invention.

[0036] Figure 5 The following are spectral images of different reconstituted tobacco leaf samples after first-order derivative preprocessing, provided in an embodiment of the present invention.

[0037] Figure 6 The diagram shows the optimal model fitting effect of partial least squares regression on nicotine content provided in the embodiments of the present invention.

[0038] Figure 7 The diagram shows the optimal model fitting effect of the support vector machine regression model on nicotine content provided in the embodiments of the present invention.

[0039] Figure 8 Two-dimensional inversion distribution pseudo-color images of nicotine content in three different grades of reconstituted tobacco leaf samples, corresponding hyperspectral grayscale images and visible light images provided for embodiments of the present invention;

[0040] Figure 9 A schematic diagram of RR model fitting and characteristic particle size analysis of nicotine content distribution in low-grade reconstituted tobacco leaf samples provided in an embodiment of the present invention.

[0041] Figure 10 A schematic diagram of RR model fitting and characteristic particle size analysis of nicotine content distribution in mid-grade reconstituted tobacco leaf samples provided in an embodiment of the present invention;

[0042] Figure 11 This is a schematic diagram of the RR model fitting and characteristic particle size analysis of the nicotine content distribution of high-grade reconstituted tobacco leaf samples provided in an embodiment of the present invention. Detailed Implementation

[0043] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0044] This invention proposes an embodiment of a method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the RR distribution model. Specifically, as follows: Figure 1 As shown, it includes:

[0045] Step S1: Place the reconstituted tobacco sample within the field of view of the near-infrared hyperspectral imaging system for non-contact scanning, acquire the reflectance spectral data of each pixel, and form a three-dimensional hyperspectral data cube.

[0046] This embodiment selects 282 reconstituted tobacco leaf samples provided by tobacco sheet suppliers within the industry, such as... Figure 2 The following is an illustration; the thin-slice samples are then manually screened by experts to remove samples with obvious appearance defects such as damage, mold, or severe wrinkles, in order to ensure data quality.

[0047] Next, a near-infrared hyperspectral imaging system was used for data acquisition. The system's wavelength range was preferably set to 1000-2000 nm, with a spectral resolution of 6.7 nm and a scanning resolution of 50 μm / pixel to ensure precise coverage of the non-uniform areas on the sample surface. Before acquisition, the system was preheated for 30 minutes to ensure light source stability. The reconstituted tobacco sample was placed flat on the moving stage, the lens-sample distance was adjusted to 300 mm, the light source intensity was set to 250 W, the scanning speed to 21 mm / s, and the image spatial resolution to 320 × 256 pixels. The reflectance spectral data of each pixel was acquired, forming a three-dimensional hyperspectral data cube.

[0048] To eliminate the effects of uneven light source and dark current noise, a standard polytetrafluoroethylene (PTFE) white plate (99.99% reflectance) was scanned under the same conditions to obtain a fully white calibration image W, and a fully black calibration image B was obtained by covering the lens cap. The original hyperspectral image R0 was then subjected to black-and-white correction to obtain the relative reflectance image R of the sample. The black-and-white correction formula is:

[0049]

[0050] In the formula: R represents the corrected signal strength, R0 represents the original signal strength, B represents the calibration signal strength in complete black, and W represents the calibration signal strength in complete white.

[0051] After acquisition, the region of interest for each sample is extracted using an image segmentation algorithm, and the spectrum of all pixels in that region is averaged to obtain the representative spectrum of the sample.

[0052] Step S2: Preprocess the spectral data and use chemometrics to correlate the preprocessed spectral data with the nicotine content reference value to establish a high-precision nicotine content quantitative inversion model. This model is used to invert the spectrum of each pixel in the three-dimensional hyperspectral data cube of the unknown sample and generate a two-dimensional distribution map of the nicotine content of the sample.

[0053] First, the nicotine content of all 282 samples was determined using a flow analyzer (standard method) as a reference value for modeling, for example, but not limited to, dividing the sample set into a calibration set and a prediction set. Figure 3 The original spectra of some samples are shown. The original reflectance spectra of tobacco sheets generally show a gradual trend in the 1000–2000 nm band, but there are obvious baseline drifts and differences in reflectance between samples, especially in the high-band region where there is a certain degree of overlap and interference. To eliminate these effects, the spectra are preprocessed, for example, multivariate scattering correction is used to eliminate baseline drift caused by scattering from the sample surface and particle size differences, and first derivative is used to enhance the characteristic absorption peaks of nicotine, thereby suppressing noise and irrelevant interference. Figure 4The spectrum after multivariate scattering correction effectively eliminates baseline shift; Figure 5 The spectrum processed using the first derivative further enhanced the resolution of nicotine's characteristic absorption peaks: in key regions such as 1100-1200 nm (CH overtones), 1400-1500 nm (NH overtones), and 1800-1900 nm (CH combination), the peak-valley characteristics of the absorption peaks were significantly amplified, effectively distinguishing the spectral differences between nicotine and matrix components. It is evident that the combination of the two preprocessing methods highlighted the vibrational information of nicotine functional groups, providing technical support for accurately identifying characteristic absorption peaks and improving the model's anti-interference capability.

[0054] In some preferred embodiments of the present invention, combined with Figure 6 and Figure 7 The diagram illustrates how partial least squares regression (PLS) and support vector machine (SVM) regression were used to establish nicotine quantification models. Combining the preprocessing methods mentioned earlier with and without preprocessing, the model performance was evaluated using the coefficient of determination (R²), root mean square error (RMSE), and relative prediction deviation (RPD). Models with R² > 0.90 and RPD > 3.0 were selected as the preferred models. The results are shown in Table 1.

[0055] Table 1. Effect Indicators of Nicotine Models

[0056] As shown in Table 1, both models exhibit excellent prediction accuracy and robustness under different spectral preprocessing conditions. Specifically, without preprocessing, the correlation coefficients (R²) for nicotine prediction by the PLS and SVM models reached 0.9083 and 0.9156, respectively, indicating good basic discrimination ability. After introducing multivariate scattering correction (MSC), the performance of both models was further improved. The validation set R² of the PLS model increased to 0.9137, RMSEP decreased to 0.1953%, and RPD reached 3.4037; while the SVM model was even better, with a validation set R² of 0.9202, RMSEP of 0.1840%, and RPD of 3.5398, showing stronger generalization ability and prediction stability. Although first-derivative preprocessing was slightly inferior to MSC in some indicators, it was still significantly better than the original spectrum, further verifying the key role of preprocessing in enhancing the characteristic bands of nicotine and suppressing irrelevant noise. Overall, all the preferred models have an R² greater than 0.90 and an RPD greater than 3.0, which fully meet the requirements for high-precision nicotine content inversion and lay a reliable model foundation for the accurate generation of nicotine spatial distribution in the future.

[0057] From three samples with different coating rates, one representative slice was selected from each sample. The optimal model described above was used to invert the nicotine content at each pixel, generating a two-dimensional pseudo-color image of the nicotine content distribution. Figure 8 Three image representations of tobacco sheet samples with different coating rates are shown: (1) pseudo-color image based on prediction results; (2) corresponding hyperspectral image grayscale distribution; (3) actual visible light image. From the pseudo-color image, it can be observed that the model successfully distinguishes tobacco sheets from different nicotine ranges, with a clear increasing trend in color gradient. The predicted value range covers low, medium, and high grades of reconstituted tobacco well, indicating that this method has high discriminative power and prediction consistency at the image level. Comparison with the visible light image also shows a high degree of consistency between the model's prediction results and the actual color depth trend, further verifying the reliability of its visualization inversion capability.

[0058] Step S3: Calculate the characteristic parameters based on the fitting of nicotine distribution;

[0059] In practice, the two-dimensional nicotine content distribution map generated in the previous step can be regarded as a collection of countless particles, with each pixel representing a particle and its nicotine content value representing the particle size. The nicotine content of all pixels is statistically analyzed, and a one-dimensional cumulative distribution function F(x) is constructed, where F(x) represents the percentage of pixels with nicotine content less than or equal to a specific value x out of the total content of all pixels. F(x) is fitted using the Rosin-Rahmler distribution model. The optimal values ​​of A and B in F(x) can be determined through fitting. Based on the fitted cumulative distribution curve, characteristic parameters are defined and calculated: D10 is the x value corresponding to F(x) = 0.1, D50 is the x value corresponding to F(x) = 0.5, and D90 is the x value corresponding to F(x) = 0.9; simultaneously, the coefficient of variation (CV) of the nicotine content of all pixels in the two-dimensional distribution map is calculated.

[0060] To elaborate further, three reconstituted tobacco leaf samples, representing low, medium, and high grades respectively, were selected. Figure 8 The three samples were processed to generate a two-dimensional nicotine distribution map. Nicotine content data for all pixels was extracted, outliers were removed, and the percentage F(x) of pixels with nicotine content less than or equal to a certain value x was calculated to construct a one-dimensional cumulative distribution function curve. The RR distribution model was then referenced. The curve was fitted, and the optimal parameters A and B were determined through iterative calculation. Based on the fitted curve, the feature parameters D10 (F(x)=0.1), D50 (F(x)=0.5), D90 (F(x)=0.9), and the coefficient of variation (CV) for all pixels were calculated. The results are shown in Table 2, and the corresponding fitted curves are shown in the figure. Figure 9 , Figure 10 , Figure 11 As shown.

[0061] Table 2. Results of Nicotine Distribution Analysis

[0062] From Table 2 and Figures 9-11 It can be seen that:

[0063] (1) The empirical cumulative distribution curves of the three samples highly coincide with the fitting curves of the RR model, indicating that the RR model can well describe the spatial distribution law of nicotine.

[0064] (2) The D50 value (distribution center) increases sequentially with the sample grade (low, medium and high) (1.1011% → 1.2011% → 1.2158%), which is completely consistent with the actual total content gradient of the sample.

[0065] (3) The D90-D10 value (distribution width) decreased from 0.0362% in the low grade and 0.0362% in the medium grade to 0.0273% in the high grade, while the coefficient of variation (CV) also decreased from 0.0105 and 0.0096 to 0.0071. This indicates that the nicotine distribution in the high grade sample is more concentrated, the difference between pixels is smaller, and the uniformity is better.

[0066] Step S4: Construct a simplified alkalinity uniformity index to comprehensively reflect the distribution pattern, in order to fully assess the spatial distribution uniformity of nicotine.

[0067] Based on the previous examples, the calculation formula is as follows: U = [D50 / (D90 - D10)] × (1 - CV); In this formula, D50 / (D90-D10) represents the ratio of the distribution center to the distribution width. The larger this value, the more concentrated the nicotine content distribution, meaning that the content values ​​of most pixels are close to the median value, and the dispersion is low; (1-CV) is a reward factor for the overall variability. The smaller the overall variability (the smaller the CV), the larger this factor. Therefore, the larger the product U value, the more uniform and concentrated the spatial distribution of nicotine in the reconstituted tobacco sample, and the better the quality consistency. Conversely, the smaller the U value, the more dispersed the distribution and the worse the uniformity.

[0068] Specifically, based on the above steps, in order to comprehensively and quantitatively evaluate the uniformity of nicotine distribution, the simplified alkalinity uniformity index U proposed in this invention is calculated. Taking the high-grade sample as an example, its D50=1.2158, D90-D10=0.0273, CV=0.0071, then U=[1.2158 / 0.0273]×(1-0.0071)≈44.3068. The calculation results of U values ​​for low, medium, and high-grade samples are listed in Table 2.

[0069] The results showed that the U-value increased significantly with the improvement of sample grade (30.0506→32.7917→44.3068), indicating that the U-value can sensitively reflect the difference in the uniformity of nicotine distribution: high-grade samples not only have higher content levels, but their pixel-level distribution is also more concentrated and uniform. The larger the U-value, the better the spatial distribution uniformity of nicotine in reconstituted tobacco leaves.

[0070] In summary, this invention integrates near-infrared hyperspectral imaging technology, chemometric modeling, and the RR distribution model to construct a complete method for determining the spatial distribution of nicotine in reconstituted tobacco leaves, encompassing data acquisition, content inversion, distribution fitting, and uniformity quantification. This method not only achieves non-destructive and visualized detection of nicotine distribution but also, for the first time, provides a precise mathematical description and comprehensive quantitative evaluation method for this key quality indicator through the RR model and the U-index. This has significant theoretical and applied value for improving the homogenization level of reconstituted tobacco products and promoting the intelligent upgrading of tobacco processing technology.

[0071] In this invention, when directional terms are mentioned, they are relative concepts based on the embodiments. Furthermore, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects have an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0072] The above description of the structure, features, and effects of the present invention is based on the embodiments shown in the figures. However, the above are only preferred embodiments of the present invention. It should be noted that the technical features involved in the above embodiments and their preferred methods can be reasonably combined and matched by those skilled in the art to form a variety of equivalent solutions without departing from or changing the design concept and technical effects of the present invention. Therefore, the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments, that do not exceed the spirit covered by the specification and figures, should be within the protection scope of the present invention.

Claims

1. A method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the RR distribution model, characterized in that, include: Obtain a three-dimensional hyperspectral data cube of reconstituted tobacco leaf samples; Based on the three-dimensional hyperspectral data cube, combined with the pre-established quantitative inversion model of nicotine content, a two-dimensional distribution map of nicotine content in the sample is generated. Each pixel in the two-dimensional distribution map is regarded as a particle, and its nicotine content value is regarded as the particle size. Based on the cumulative distribution function, the Rosin-Rahmler distribution model is used to fit the nicotine content distribution and calculate the distribution characteristic parameters. The distribution characteristic parameters represent the nicotine content value corresponding to the cumulative distribution reaching a preset percentage. The coefficient of variation of nicotine content at all pixels in the two-dimensional distribution map is calculated, and a uniformity index is constructed in combination with the distribution characteristic parameters to quantitatively evaluate the spatial distribution uniformity of nicotine in the reconstituted tobacco sample.

2. The method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the RR distribution model according to claim 1, characterized in that, The process of establishing a quantitative inversion model for nicotine content includes: Obtain hyperspectral data of reconstituted tobacco calibration set samples with nicotine content reference values; Hyperspectral data are preprocessed to remove noise and baseline drift; Partial least squares regression or support vector machine is used to correlate the preprocessed hyperspectral data with the reference value of nicotine content to construct a quantitative inversion model.

3. The method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the RR distribution model according to claim 2, characterized in that, Preprocessing methods for hyperspectral data include at least: multivariate scattering correction and first derivative processing.

4. The method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the RR distribution model according to claim 1, characterized in that, The three-dimensional hyperspectral data cube was acquired by non-contact scanning of reconstituted tobacco samples in the wavelength range of 1000-2000 nm using a near-infrared hyperspectral imaging system.

5. The method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the RR distribution model according to claim 1, characterized in that, The Rosin-Rahmler distribution model was used for fitting, as follows: Where x is the nicotine content of a pixel, F(x) is the percentage of the total nicotine content of pixels with a nicotine content less than or equal to x, and A and B are model parameters.

6. The method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the RR distribution model according to claim 1, characterized in that, The coefficient of variation is calculated as follows: Where CV is the coefficient of variation of nicotine content at full pixel count; σ is the standard deviation of nicotine content at full pixel count; x mean This is the arithmetic mean of nicotine content across all pixels.

7. The method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the RR distribution model according to claim 1, characterized in that, ... The distribution characteristic parameters include: D10, D50 and D90; wherein, D10, D50 and D90 represent the nicotine content values ​​corresponding to the cumulative distribution reaching 10%, 50% and 90%, respectively. The uniformity index U is calculated as follows: in, The ratio of the distribution center to the distribution width represents the central tendency of the distribution; (1-CV) represents the penalty factor for the overall degree of variation.

8. The method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the RR distribution model according to claim 7, characterized in that, The determination method further includes: determining the uniformity level of nicotine spatial distribution in reconstituted tobacco leaves based on the value of the uniformity index U, wherein a larger U value indicates a more uniform nicotine distribution.

9. The method for determining the spatial distribution of nicotine in reconstituted tobacco leaves based on hyperspectral imaging and the RR distribution model according to any one of claims 1 to 8, characterized in that, The determination method further includes: after establishing a quantitative inversion model for nicotine content, evaluating the prediction accuracy and generalization ability of the model using the coefficient of determination R², root mean square error RMSE, and relative prediction deviation RPD.

Images