Flue-cured tobacco flavor type distinguishing method and device based on visible-near infrared spectrum

Abstract

The invention provides a flue-cured tobacco flavor type discrimination method based on visible-near infrared spectroscopy, which comprises the following steps: realizing original state nondestructive detection of bundled tobacco leaves based on visible-near infrared spectroscopy without destructive pretreatment such as grinding and compaction; wherein when the visible-near infrared spectrum is processed, based on a threshold segmentation method, the characteristic wave band is extracted, the characteristic wave band is within 500.14-649.75 nm, and the spectral reflectivity value is smaller than or equal to 0.3, or the characteristic wave band is within 700.52-799.26 nm, and the spectral reflectivity value is larger than or equal to 0.5; taking a pixel point of which the characteristic wave band is within 1001.71-1298.17 nm and the spectral reflectivity value is less than or equal to 0.55, or the characteristic wave band is within 1502.16-1598.29 nm and the spectral reflectivity value is greater than or equal to 0.4, and the difference value between the upper and lower bounds of the spectral reflectivity in the wave band is greater than or equal to 0.1 as a region of interest, and calculating the mean value of the spectral reflectivity of all pixel points in the region of interest, as a single spectrum curve of a tobacco leaf sample of the tobacco bundle, a tobacco leaf stem and a shadow part are effectively removed, and a tobacco leaf surface area is effectively extracted and reserved.

Description

Technical Field

[0001] This invention relates to the field of digital detection technology for flue-cured tobacco quality, specifically to a method and apparatus for determining the aroma of flue-cured tobacco based on visible-near-infrared spectroscopy. Background Technology

[0002] The style and quality of tobacco leaves are the core elements determining their usability. Tobacco leaves with different aroma styles have different application values ​​in the cross-regional combination of raw materials and leaf blend formulations in the cigarette industry. The "National Study on Aroma Style Regionalization of Flue-cured Tobacco Leaves" is based on the needs of raw materials for Chinese cigarettes. Through the synergistic verification of the four-dimensional index system of ecology, sensory, chemistry and metabolism, it innovatively constructed a classification system of eight major aroma types of flue-cured tobacco leaves in China. Based on the principles of aroma style stability, superior ecological resources, standardized technical management and production scale, it finally determined the typical production areas of the eight major aroma types (National Flue-cured Tobacco Leaf Aroma Style Regionalization, China Tobacco Science, 2019, 25(4), 1-9.).

[0003] The current methods for judging the style of flue-cured tobacco mainly include traditional sensory evaluation and chromatographic fingerprinting. The former relies on the deep industry experience and rich expertise of the evaluation experts, and has problems such as strong subjectivity and poor reproducibility. The latter, although highly accurate, is difficult to apply on a large scale due to its cumbersome operation and high cost. In recent years, discrimination models based on near-infrared spectroscopy, electronic nose and other technologies have made progress, but there are still three major bottlenecks: (1) The detection requires grinding and compacting the tobacco leaves, which destroys the physical morphology of the sample and cannot meet the needs of industry. (2) Current models are mostly for the three basic aroma types of strong, medium and light, and are difficult to adapt to the current eight aroma type classification system. (3) The model training is mostly based on individual main varieties, and the generalization ability to multiple production areas and varieties across the country is insufficient.

[0004] In order to solve the above problems, people have been seeking an ideal technological solution. Summary of the Invention

[0005] Therefore, it is necessary to provide a method and apparatus for identifying the aroma of flue-cured tobacco based on visible-near-infrared spectroscopy to address the aforementioned technical problems.

[0006] To achieve the above objectives, the first aspect of the present invention provides a method for identifying the aroma type of flue-cured tobacco based on visible-near-infrared spectroscopy, comprising the following steps:

[0007] Obtain representative visible-near-infrared spectral information of tobacco leaf samples from different ecological zones;

[0008] The visible-near-infrared spectral information obtained in the previous step is subjected to black-and-white radiometric correction. Based on the threshold segmentation method, pixels with the following characteristics are extracted as regions of interest: 500.14–649.75 nm with a spectral reflectance value less than or equal to 0.3; 700.52–799.26 nm with a spectral reflectance value greater than or equal to 0.5; 1001.71–1298.17 nm with a spectral reflectance value less than or equal to 0.55; or 1502.16–1598.29 nm with a spectral reflectance value greater than or equal to 0.4; and the difference between the upper and lower bounds of the spectral reflectance within the band is greater than or equal to 0.1. The mean spectral reflectance of all pixels within the region of interest is calculated as a single spectral curve representing the tobacco leaf sample.

[0009] Preprocessing and characteristic band selection of the spectral curves;

[0010] Based on the feature bands obtained in the previous step, SVC, RF, and KNN models were constructed respectively, and the established models were qualitatively analyzed to select the model with the best performance as the flue-cured tobacco aroma classification model.

[0011] For the tobacco leaves to be detected, visible-near-infrared spectral information is obtained using spectral technology. The visible-near-infrared spectral information is then subjected to black-and-white radiometric correction. The main body of the tobacco leaf surface is extracted as the region of interest based on the threshold segmentation method. The average spectral reflectance of all pixels in the region of interest is calculated as the spectral curve representing the tobacco leaf to be detected. After preprocessing and feature band screening of the spectral curve, it is imported into the flue-cured tobacco aroma classification model selected in the previous step for further processing to obtain the aroma classification result.

[0012] In one possible embodiment of the first aspect, the specific steps for obtaining visible-near-infrared spectral information using spectral techniques include:

[0013] The tobacco leaves are laid flat on the stage, and the first hyperspectral camera (400-1000 nm wavelength) and the second hyperspectral camera (900-1700 nm wavelength) are used to collect spectral data of the first measurement surface of the tobacco leaves.

[0014] The tobacco leaves will be flipped 180° up and down and laid flat again. The first hyperspectral camera (wavelength 400-1000nm) and the second hyperspectral camera (wavelength 900-1700nm) will be used to collect spectral data of the second measurement surface of the tobacco leaves.

[0015] The spectral data from the first measurement surface and the spectral data from the second measurement surface are averaged to obtain the visible-near-infrared spectral information of the tobacco leaf.

[0016] To achieve the above objectives, a second aspect of the present invention provides a device for determining the aroma of flue-cured tobacco based on visible-near-infrared spectroscopy, comprising:

[0017] The spectral acquisition module is used to acquire the visible-near-infrared spectral information of representative tobacco leaf samples from different ecological zones;

[0018] The Interest Extraction (IOI) module performs black-and-white radiometric correction on the visible-near-infrared spectral information and extracts pixels with the following characteristics based on threshold segmentation: 500.14–649.75 nm with a spectral reflectance value less than or equal to 0.3; 700.52–799.26 nm with a spectral reflectance value greater than or equal to 0.5; 1001.71–1298.17 nm with a spectral reflectance value less than or equal to 0.55; or 1502.16–1598.29 nm with a spectral reflectance value greater than or equal to 0.4, and the difference between the upper and lower bounds of the spectral reflectance within the band is greater than or equal to 0.1. These pixels are then designated as regions of interest (ROIs). The mean spectral reflectance of all pixels within the ROI is calculated as a single spectral curve representing the tobacco leaf sample.

[0019] The spectral preprocessing module is used to preprocess the spectral curves.

[0020] A characteristic band screening method is used to screen characteristic bands of preprocessed spectral curves.

[0021] The module for constructing a flue-cured tobacco aroma classification model is used to construct SVC, RF, and KNN models respectively, and to perform qualitative analysis on the established models to select the model with the best performance as the flue-cured tobacco aroma classification model.

[0022] The tobacco aroma prediction module is used to call the spectrum acquisition module, the extraction of interest module, and the spectrum preprocessing module to process the tobacco leaves to be detected, obtain the spectral curve, and call the flue-cured tobacco aroma classification model to identify the aroma of the tobacco leaves to be detected.

[0023] To achieve the above objectives, a third aspect of the present invention provides a computer device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory is used to store computer programs; and the processor is used to execute the program stored in the memory to implement the flue-cured tobacco aroma discrimination method as described above.

[0024] To achieve the above objectives, a fourth aspect of the present invention provides a computer-readable storage medium having instructions stored thereon that, when executed by one or more processors, cause the processors to perform the flue-cured tobacco aroma discrimination method as described above.

[0025] To achieve the above objectives, a fifth aspect of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for identifying the aroma of flue-cured tobacco.

[0026] The beneficial effects of this invention are as follows:

[0027] This invention utilizes visible-near-infrared spectroscopy to achieve non-destructive detection of bundled tobacco leaves without the need for destructive pretreatment such as grinding or compaction. Furthermore, during region of interest extraction, threshold segmentation is performed based on the following conditions: a spectral reflectance value of less than or equal to 0.3 within the range of 500.14–649.75 nm; a spectral reflectance value of greater than or equal to 0.5 within the range of 700.52–799.26 nm; a spectral reflectance value of less than or equal to 0.55 within the range of 1001.71–1298.17 nm; or a spectral reflectance value of greater than or equal to 0.4 within the range of 1502.16–1598.29 nm; and a difference between the upper and lower bounds of the spectral reflectance within the band greater than or equal to 0.1. This effectively removes the main stem and shaded areas of the tobacco leaves, effectively extracting and preserving the leaf surface area. By employing a double-sided sampling strategy with the tobacco leaves rotated 180°, the error of single-sided measurement is effectively overcome. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating the method for determining the aroma of flue-cured tobacco according to the present invention;

[0029] Figure 2 This is a schematic diagram of hyperspectral acquisition of tobacco leaf images;

[0030] Figure 3 Filter the map for regions of interest in tobacco leaves;

[0031] Figure 4 The average spectrum of the eight major fragrance types;

[0032] Figure 5 The confusion matrix constructed for the optimal model (fifth fold). Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the implementation process of this invention will be described in detail below in conjunction with experimental procedures. This invention is based on visible light hyperspectral imaging technology to achieve the classification and prediction of the aroma type of flue-cured tobacco. Specifically, as... Figure 1As shown, the method for identifying the aroma type of flue-cured tobacco based on visible-near-infrared spectroscopy acquires the visible-near-infrared spectral information of tobacco leaf samples through a double-sided acquisition strategy with the tobacco leaves rotated 180°, overcoming the error of single-sided measurement. In data processing, a threshold segmentation method combining characteristic bands and spectral reflectance values ​​is used to accurately extract the main region of tobacco leaf (ROI). One or more combined preprocessing methods are used to process the spectral curves, and then standardized SS is used to unify the feature scale, which significantly improves the quality of spectral data. In model construction, Spearman correlation analysis (|ρ|>0.3) and cluster analysis (CA) are used to screen characteristic bands. Combined with SVM, RF, and KNN model construction methods and 5-fold cross-validation grid search parameter tuning, a high-precision aroma type identification system is constructed. Multi-dimensional evaluation indicators such as F1 score are introduced to verify the robustness of the model, providing technical support for digital characterization of tobacco leaf formulation, leaf threshing and re-drying.

[0034] Example 1

[0035] Experimental Design:

[0036] S0: Collection of flue-cured tobacco samples

[0037] Sample collection: 205 representative sampling points were selected from 22 tobacco-growing provinces and regions across the country, covering 27 tobacco varieties and eight major aroma types, and a total of 615 samples were obtained, as shown in Table 1.

[0038]

[0039] Table 1. Distribution of flue-cured tobacco samples by eight aroma types and ecological zones.

[0040] S1: Spectral acquisition of flue-cured tobacco samples

[0041] Instrument configuration: Spectral data were acquired using a first hyperspectral camera and a second hyperspectral camera; the first hyperspectral camera was a FigSpec-17 hyperspectral camera with a wavelength of 400–1000 nm and a resolution of 2.02 nm, and the second hyperspectral camera was a FigSpec-2X hyperspectral camera with a wavelength of 900–1700 nm and a resolution of 3.33 nm.

[0042] The acquisition system consists of a closed acquisition chamber, halogen lamps (50W power, 3000K yellow light, 12V input voltage) distributed on both sides of the chamber, a precision stage, and a control computer. By fixing the central axis of the halogen lamp light source at a 25° angle to the plane of the stage, the center of the light source on both sides is ensured to point towards the dark area of ​​the light source on the other side, so that the light source illuminates the tobacco leaf sample to be tested evenly and avoids local overexposure. The FigSpec-17 hyperspectral camera lens is 90cm away from the stage, and the FigSpec-2X hyperspectral camera is 77cm away from the stage, ensuring that the camera's field of view covers the sample.

[0043] The tobacco leaf samples were laid flat on the stage and kept in a naturally spread state. Spectral data for each tobacco leaf sample were acquired using a FigSpec-17 hyperspectral camera (wavelength 400–1000 nm, resolution 2.02 nm) and a FigSpec-2X hyperspectral camera (wavelength 900–1700 nm, resolution 3.33 nm). The acquisition results are shown below. Figure 2 As shown.

[0044] In particular, spectral data were collected twice for each tobacco leaf sample. Specifically, the tobacco leaves were laid flat on the stage, and spectral data of the first measurement surface of the tobacco leaves were collected using a FigSpec-17 hyperspectral camera with a wavelength of 400-1000 nm and a FigSpec-2X hyperspectral camera with a wavelength of 900-1700 nm.

[0045] The tobacco leaves will be flipped 180° up and down and laid flat again. Then, the spectral data of the second measurement surface of the tobacco leaves will be collected again using a FigSpec-17 hyperspectral camera with a wavelength of 400-1000 nm and a FigSpec-2X hyperspectral camera with a wavelength of 900-1700 nm.

[0046] The spectral data from the first measurement surface and the spectral data from the second measurement surface are averaged to obtain the visible-near-infrared spectral information.

[0047] A total of 1230 spectral scans were performed, with effective channels ranging from 400.43 to 779.14 nm and from 904.94 nm to 1700.12 nm, totaling 428 bands.

[0048] Step S2: Hyperspectral data processing;

[0049] To eliminate noise interference, black and white board radiometric correction was performed after data acquisition: Dark reference data (reflectivity close to 0%) was acquired in a completely dark environment with all light sources turned off. After calibration with the light sources turned on and a high-reflectivity white board (reflectivity ≥ 99%) covered, the acquired hyperspectral images were corrected based on the following formula:

[0050]

[0051] In the formula: I is the corrected hyperspectral data; I0 is the original hyperspectral data collected; B is the calibration data in a completely dark environment; W is the standard white board data.

[0052] To accurately select the region of interest (ROI) on the tobacco leaf surface, it is necessary to avoid the main stem of the tobacco leaf and remove the shaded areas caused by the stacking of tobacco leaves, and to cover the effective leaf surface of the tobacco leaf as completely as possible.

[0053] In this embodiment, a threshold segmentation method is used to extract the main body of the tobacco leaf surface as the region of interest. Specifically, the threshold conditions for the threshold segmentation method are as follows: the characteristic band is located within 500.14–649.75 nm and the spectral reflectance value is less than or equal to 0.3; or the characteristic band is located within 700.52–799.26 nm and the spectral reflectance value is greater than or equal to 0.5; or the characteristic band is located within 1001.71–1298.17 nm and the spectral reflectance value is less than or equal to 0.55; or the characteristic band is located within 1502.16–1598.29 nm and the spectral reflectance value is greater than or equal to 0.4, and the difference between the upper and lower bounds of the spectral reflectance within the band is greater than or equal to 0.1.

[0054] Therefore, this embodiment uses a threshold segmentation method to extract pixels with the following characteristics: a spectral reflectance value less than or equal to 0.3 in the 500.14–649.75 nm range; a spectral reflectance value greater than or equal to 0.5 in the 700.52–799.26 nm range; a spectral reflectance value less than or equal to 0.5 in the 1001.71–1298.17 nm range; or a spectral reflectance value greater than or equal to 0.55 in the 1502.16–1598.29 nm range; and a spectral reflectance value greater than or equal to 0.4 in the 1502.16–1598.29 nm range; and a difference between the upper and lower bounds of the spectral reflectance within the spectral reflectance range greater than or equal to 0.1 as regions of interest. Figure 3 This refers to the region of interest that has been extracted.

[0055] After extracting the region of interest, the mean spectral reflectance of pixels within the region of interest that satisfy all the above conditions is calculated and used as a single spectral curve representing the tobacco leaf sample, as shown below. Figure 4 The figure shows the average spectral curves of the eight major fragrance types.

[0056] Step S3: Spectral curve preprocessing;

[0057] There are many preprocessing methods for spectral curves, such as smoothing, noise reduction, spectral line restoration, standardization and spectral normalization, standard normal variable (SNV), multiplicative scattering correction (MSC), Fourier transform (FT), wavelet transform (WT), and orthogonal signal correction (OSC). Specifically, in this embodiment, single and combined preprocessing methods such as SG smoothing, D1 derivative, multivariate scattering correction (MSC), and vector normalization (SNV) are used to process the spectral data, and then standardization (SS) is applied to unify the feature scale. Savitzky-Golay (SG) smoothing and first derivative (D1) eliminate baseline drift, while multivariate scattering correction (MSC) and standard normal variable transformation (SNV) eliminate the influence of granularity differences. Taking SVC modeling as an example, the impact of different preprocessing methods on model performance is shown in Table 2.

[0058]

[0059] As can be seen, in this embodiment, the optimal preprocessing combination is SG-D1-SS. Using this optimal preprocessing combination to preprocess the obtained spectral curve can ensure optimal data processing, obtain the best model result, and improve prediction accuracy.

[0060] Step S4: Feature band selection;

[0061] Spearman correlation coefficient was used to screen feature bands that were significantly correlated with the target variable, and the final feature subset was selected with |ρ|>0.3 as the boundary to reduce the feature dimensionality. K-means clustering was performed on the bands using CA, and the bands closest to the cluster center were selected from each cluster as representative features to eliminate redundant information.

[0062] Based on the selected bands, SVC, RF, and KNN models were established respectively. Model performance evaluation is shown in Table 3. Evaluation metrics included accuracy, precision, recall, and F1 score. Specifically, the evaluation metrics were confusion matrix, accuracy, precision, recall, and F1 score; the F1 score is a combined indicator of precision and recall, with values ​​ranging from 0 to 1, where 1 is the best and 0 is the worst. A higher F1 score indicates a more robust model. The calculation formula is shown below:

[0063]

[0064] Table 3. Model Performance Evaluation Table After Feature Band Selection

[0065]

[0066]

[0067] As can be seen from Table 3, the CA feature band selection method is the optimal selection method. Under this method, the number of feature bands is the fewest, and the accuracy, precision, recall, and F1 score are all the highest.

[0068] Step S5: Construction of the classification model for the eight major aroma types

[0069] We employ three algorithms: Support Vector Regression (SVC), Random Forest, and K-Nearest Neighbors (KNN). We determine the optimal hyperparameters for each model using a grid search method with 5-fold cross-validation. Finally, we evaluate the model performance on an independent test set and output a complete evaluation report that includes cross-validation accuracy, test set accuracy, and feature importance ranking.

[0070] Specifically, the evaluation metrics used are confusion matrix, accuracy, precision, recall, and F1 score.

[0071] Referring to Table 3, the SVC-based flue-cured tobacco aroma classification model outperforms Random Forest and K-Nearest Neighbors (KNN). Therefore, the SVC-based model is selected as the flue-cured tobacco aroma classification model for subsequent processing. Specifically, as shown in Table 3... Figure 5 The image shows the confusion matrix (fifth fold) constructed by the SVC flue-cured tobacco aroma classification model.

[0072] Specifically, based on SG-D1-SS as preprocessing and combined with the CA feature selection method, the evaluation index results of the SVC model are shown in Table 4.

[0073]

[0074]

[0075] As can be seen from the table above, the visible-near-infrared spectroscopy method for identifying the aroma of flue-cured tobacco proposed in this embodiment has good identification accuracy.

[0076] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0077] Example 2

[0078] Based on the same inventive concept, this application also provides a flue-cured tobacco aroma discrimination device for implementing the flue-cured tobacco aroma discrimination method described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more flue-cured tobacco aroma discrimination device embodiments provided below can be found in the limitations of the flue-cured tobacco aroma discrimination method above, and will not be repeated here.

[0079] Specifically, the flue-cured tobacco aroma discrimination device based on visible-near-infrared spectroscopy includes:

[0080] The spectral acquisition module is used to acquire the visible-near-infrared spectral information of representative tobacco leaf samples from different ecological zones;

[0081] The Interest Extraction (IOI) module performs black-and-white radiometric correction on the visible-near-infrared spectral information and extracts pixels with the following characteristics based on threshold segmentation: 500.14–649.75 nm with a spectral reflectance value less than or equal to 0.3; 700.52–799.26 nm with a spectral reflectance value greater than or equal to 0.5; 1001.71–1298.17 nm with a spectral reflectance value less than or equal to 0.55; or 1502.16–1598.29 nm with a spectral reflectance value greater than or equal to 0.4, and the difference between the upper and lower bounds of the spectral reflectance within the band is greater than or equal to 0.1. These pixels are then designated as regions of interest (ROIs). The mean spectral reflectance of all pixels within the ROI is calculated as a single spectral curve representing the tobacco leaf sample.

[0082] The spectral preprocessing module is used to preprocess the spectral curves.

[0083] A characteristic band screening method is used to screen characteristic bands of preprocessed spectral curves.

[0084] The module for constructing a flue-cured tobacco aroma classification model is used to construct SVC, RF, and KNN models respectively, and to perform qualitative analysis on the established models to select the model with the best performance as the flue-cured tobacco aroma classification model.

[0085] The tobacco aroma prediction module is used to call the spectrum acquisition module, the extraction of interest module, and the spectrum preprocessing module to process the tobacco leaves to be detected, obtain the spectral curve, and call the flue-cured tobacco aroma classification model to identify the aroma of the tobacco leaves to be detected.

[0086] Example 3

[0087] This embodiment provides a computer device, which can be a terminal. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are connected to the system bus via the input / output interface.

[0088] The processor in this computer device provides computing and control capabilities.

[0089] The computer device's memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium.

[0090] The computer device's input / output interface is used for exchanging information between the processor and external devices.

[0091] The communication interface of this computer device is used to communicate with external terminals via wired or wireless means. Wireless communication can be achieved through WIFI, mobile cellular networks, NFC (Near Field Communication), or other technologies.

[0092] When the computer program is executed by the processor, it implements the method for identifying the aroma of flue-cured tobacco as described in Example 1.

[0093] The display unit of this computer device is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of this computer device can be a touch layer covering the display screen, or buttons, a trackball, or a touchpad set on the casing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0094] Example 4

[0095] Based on the above embodiments, this embodiment provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method for identifying the aroma of flue-cured tobacco as described in Embodiment 1.

[0096] Example 5

[0097] Based on the above embodiments, this embodiment provides a computer program product, including a computer program that, when executed by a processor, implements the method for identifying the aroma of flue-cured tobacco as described in Embodiment 1.

[0098] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0099] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments 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.

[0100] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation of the present invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications and substitutions should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. A method for identifying the aroma type of flue-cured tobacco based on visible-near-infrared spectroscopy, characterized in that, Includes the following steps: Obtain representative visible-near-infrared spectral information of tobacco leaf samples from different ecological zones; The visible-near-infrared spectral information obtained in the previous step is subjected to black-and-white radiometric correction. Based on the threshold segmentation method, pixels with the following characteristics are extracted as regions of interest: 500.14 ~ 649.75 nm with a spectral reflectance value less than or equal to 0.3; 700.52 ~ 799.26 nm with a spectral reflectance value greater than or equal to 0.5; 1001.71 ~ 1298.17 nm with a spectral reflectance value less than or equal to 0.55; or 1502.16 ~ 1598.29 nm with a spectral reflectance value greater than or equal to 0.4; and the difference between the upper and lower bounds of the spectral reflectance within the band is greater than or equal to 0.

1. The mean spectral reflectance of all pixels within the region of interest is calculated as a single spectral curve representing the tobacco leaf sample. Preprocessing and characteristic band selection of the spectral curves; Based on the feature bands obtained in the previous step, SVC, RF, and KNN models were constructed respectively, and the established models were qualitatively analyzed to select the model with the best performance as the flue-cured tobacco aroma classification model. For the tobacco leaves to be detected, visible-near-infrared spectral information is obtained using spectral technology. The visible-near-infrared spectral information is then subjected to black-and-white radiometric correction. The main body of the tobacco leaf surface is extracted as the region of interest based on the threshold segmentation method. The average spectral reflectance of all pixels in the region of interest is calculated as the spectral curve representing the tobacco leaf to be detected. After preprocessing and feature band screening of the spectral curve, it is imported into the flue-cured tobacco aroma classification model selected in the previous step for further processing to obtain the aroma classification result.

2. The method for determining the aroma type of flue-cured tobacco based on visible-near-infrared spectroscopy according to claim 1, characterized in that, The specific steps for obtaining visible-near-infrared spectral information using spectroscopic techniques include: The tobacco leaves will be laid flat on the stage, and the first hyperspectral camera (400-1000 nm wavelength) and the second hyperspectral camera (900-1700 nm wavelength) will be used to collect the spectral data of the first measurement surface of the tobacco leaves. The tobacco leaves will be flipped 180° up and down and laid flat again. The first hyperspectral camera (wavelength 400-1000 nm) and the second hyperspectral camera (wavelength 900-1700 nm) will be used to collect the spectral data of the second measurement surface of the tobacco leaves. The spectral data from the first measurement surface and the spectral data from the second measurement surface are averaged to obtain the visible-near-infrared spectral information of the tobacco leaf.

3. A method for identifying the aroma type of flue-cured tobacco based on visible-near-infrared spectroscopy according to claim 1 or 2, characterized in that, The spectral curves were processed using one or more of the following preprocessing methods: Savitzky-Golay smoothing, D1 derivative, multivariate scattering correction (MSC), and vector normalization (SNV). Then, the SS standardization was used to unify the characteristic scale.

4. The method for determining the aroma type of flue-cured tobacco based on visible-near-infrared spectroscopy according to claim 1, characterized in that, The specific steps for feature band selection include: Spearman correlation coefficient was used to screen feature bands that were significantly correlated with the target variable, and the final feature subset was selected with |ρ|>0.3 as the boundary. K-means clustering of the bands was performed using CA, and the band closest to the cluster center was selected as the representative feature from each cluster.

5. The method for determining the aroma type of flue-cured tobacco based on visible-near-infrared spectroscopy according to claim 1, characterized in that, The aroma classification model for flue-cured tobacco was trained using 5-fold cross-validation. During model training, the validation set accuracy is used as the optimization metric. The optimal combination of hyperparameters of the model is determined by grid search. The confusion matrix, accuracy, precision, recall, and F1 score are used as evaluation metrics to qualitatively evaluate the model performance.

6. A device for identifying the aroma of flue-cured tobacco based on visible-near-infrared spectroscopy, characterized in that, include: The spectral acquisition module is used to acquire the visible-near-infrared spectral information of representative tobacco leaf samples from different ecological zones; The Interest Extraction (IOI) module performs black-and-white radiometric correction on the visible-near-infrared spectral information and extracts pixels with the following characteristics based on threshold segmentation: 500.14 ~ 649.75 nm with a spectral reflectance value less than or equal to 0.3; 700.52 ~ 799.26 nm with a spectral reflectance value greater than or equal to 0.5; 1001.71 ~ 1298.17 nm with a spectral reflectance value less than or equal to 0.55; or 1502.16 ~ 1598.29 nm with a spectral reflectance value greater than or equal to 0.4, and the difference between the upper and lower bounds of the spectral reflectance within the band is greater than or equal to 0.

1. These pixels are then used as regions of interest (ROIs). The mean spectral reflectance of all pixels within the ROI is calculated as a single spectral curve representing the tobacco leaf sample. The spectral preprocessing module is used to preprocess the spectral curves. A characteristic band screening method is used to screen characteristic bands of preprocessed spectral curves. The module for constructing a flue-cured tobacco aroma classification model is used to construct SVC, RF, and KNN models respectively, and to perform qualitative analysis on the established models to select the model with the best performance as the flue-cured tobacco aroma classification model. The tobacco aroma prediction module is used to call the spectrum acquisition module, the extraction of interest module, and the spectrum preprocessing module to process the tobacco leaves to be detected, obtain the spectral curve, and call the flue-cured tobacco aroma classification model to identify the aroma of the tobacco leaves to be detected.

7. A computer device, characterized in that: It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor, when executing a program stored in memory, implements the method for determining the aroma of flue-cured tobacco as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for identifying the aroma of flue-cured tobacco as described in any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for identifying the aroma of flue-cured tobacco as described in any one of claims 1 to 5.

Images