Method and device for lossless quantitative characterization of flue-cured tobacco part oil content based on visible light spectrum

Abstract

The invention provides a flue-cured tobacco part oil content nondestructive quantitative characterization method and device based on visible light spectrum.The method comprises the steps that visible light spectrum data of flue-cured tobacco leaf samples at different parts are collected through a hyperspectral camera, and a series of pretreatment is carried out, the series of preprocessing at least comprises the steps of carrying out black and white board radiation correction on the visible light spectrum information, and extracting pixel points of which the characteristic wave bands are within 500.14-649.75 nm and the spectral reflectivity values are less than 0.3 or the characteristic wave bands are within 700.52-799.26 nm and the spectral reflectivity values are greater than or equal to 0.5 as regions of interest based on a threshold segmentation method; calculating a spectral reflectivity mean value of all pixel points in the region of interest as a single spectral curve representing the tobacco leaf sample of the tobacco bundle; and finally, on the basis of the preprocessed spectrum curve, constructing a flue-cured tobacco upper, middle and lower part classification model and constructing a part flue-cured tobacco oil score prediction model.

Description

Technical Field

[0001] The present invention relates to the technical field of digital detection of flue-cured tobacco quality, and in particular to a method and device for non-destructive quantitative characterization of oil content in different parts of flue-cured tobacco based on visible light spectroscopy. Background Art

[0002] Oil content is the appearance of the soft, liquid or semi-liquid substance inside tobacco leaves. It directly affects the physical and chemical properties and sensory quality of tobacco leaves and is a key indicator for tobacco leaf quality grading. Traditional manual grading methods are subject to strong subjectivity and poor consistency. In recent years, with the advancement of computer and instrumentation science, the use of machine learning methods to identify and predict the oil content grade of flue-cured tobacco leaves has become a hot topic in oil content research. For example, [202410025469.4] proposed a multimodal oil content characterization method for flue-cured tobacco leaves. This method digitizes the oil content characteristics of tobacco leaf samples through labeling. Then, image features, static weight, and color difference values ​​are obtained using industrial image acquisition, digital scale measurements, and colorimeter measurements. Correlation analysis is performed between oil content and other features to identify those significantly correlated with oil content. Furthermore, regression and path analysis are used to derive an oil content characterization equation.

[0003] Hyperspectral technology, as a fast, efficient and non-destructive qualitative and quantitative detection method, is currently also widely used in the tobacco industry. For example, Fu Guangming, Gao Ziting, Yang Jianxin, et al. proposed a study on the identification of flue-cured tobacco oil content based on near-infrared spectroscopy. In fact, the existing spectral technology for the detection of flue-cured tobacco oil content still has three major bottlenecks: (1) Existing methods overly rely on the analysis of chemical information in the near-infrared band (780-1700nm) and ignore the correlation between the apparent optical characteristics in the visible light region (400-780nm) and oil content, which restricts the synergistic effect of multi-source information; (2) Research is mostly limited to qualitative discrimination of oil grade and lacks quantitative prediction models; (3) Detection requires destructive pre-treatment such as grinding, which cannot meet the needs of non-destructive testing; (4) Modeling ignores the differences in oil distribution between tobacco leaf parts, resulting in insufficient model generalization ability. Therefore, it is urgent to establish a non-destructive testing system based on visible spectroscopy and part-by-part modeling to achieve accurate quantitative evaluation of flue-cured tobacco oil content and provide key technical support for intelligent grading.

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

[0005] Based on this, it is necessary to provide a non-destructive quantitative characterization method and device for the oil content in different parts of flue-cured tobacco based on visible light spectroscopy to address the above technical problems.

[0006] In order to achieve the above object, the present invention provides, in a first aspect, a method for non-destructive quantitative characterization of oil content in different parts of flue-cured tobacco based on visible light spectroscopy, comprising the following steps:

[0007] Obtain tobacco leaf samples from different parts of the tobacco leaf, obtain the oil content of the tobacco leaf samples through expert scoring, and use spectral technology to obtain the visible light spectrum information of the tobacco leaf samples;

[0008] The visible light spectrum information was corrected for black and white plate radiation. Based on the threshold segmentation method, pixels with characteristic bands within 500.14-649.75 nm and spectral reflectance values ​​less than 0.3, or with characteristic bands within 700.52-799.26 nm and spectral reflectance values ​​greater than or equal to 0.5 were extracted as regions of interest. The mean spectral reflectance of all pixels in the region of interest was calculated as a single spectral curve representing the tobacco leaf sample.

[0009] The spectral curve is preprocessed, and based on the preprocessed spectral curve, a part classification model is constructed using a support vector classification modeling method;

[0010] The spectral curves were classified according to the parts of the tobacco leaves. Based on the spectral curves of different parts and the corresponding oil content values, a partial least squares regression model was used to construct a prediction model for the oil content of the corresponding parts of the tobacco leaves.

[0011] For the tobacco leaves to be tested, spectral technology is used to obtain visible light spectral information, and the visible light spectral information is corrected for black and white plate radiation. Based on the threshold segmentation method, the main body of the tobacco leaf surface is extracted as the region of interest, and the mean spectral reflectance of all pixels in the region of interest is calculated as the spectral curve representing the tobacco leaves to be tested; the spectral curve is imported into the part classification model to identify the part of the tobacco leaves to be tested; according to the part identification result, the flue-cured tobacco oil content score prediction model of the corresponding part is called to process the spectral curve to obtain the oil content score prediction value of the tobacco leaves to be tested.

[0012] In a possible embodiment of the first aspect, the specific steps of obtaining visible light spectrum information using spectroscopy technology include:

[0013] Spread the tobacco leaves naturally on the stage, and use a hyperspectral camera to collect spectral data of the first measurement surface of the tobacco leaves at a wavelength of 400 to 1000 nm;

[0014] The tobacco leaves are turned upside down 180 degrees and laid flat again, and the spectral data of the second measurement surface of the tobacco leaves are collected again using a hyperspectral camera at a wavelength of 400 to 1000 nm;

[0015] The spectrum data of the first measuring surface and the spectrum data of the second measuring surface are averaged to serve as the visible light spectrum information of the tobacco leaves.

[0016] In a possible embodiment of the first aspect, obtaining the oil content value of a flue-cured tobacco leaf sample includes: balancing the moisture content of the flue-cured tobacco leaf sample to 16% to 18% in a constant humidity environment, pre-treating the flue-cured tobacco leaf sample according to the "Flue-cured Tobacco" (GB2635-1992) grading standard, and having multiple professional grading technicians score the oil content of the tobacco leaves according to a 10-point system, and taking the average of the scores as the oil content value of the flue-cured tobacco leaf sample.

[0017] In order to achieve the above-mentioned object, the second aspect of the present invention provides a non-destructive quantitative characterization device for oil content in different parts of flue-cured tobacco based on visible light spectroscopy, comprising:

[0018] The oil content value acquisition module is used to obtain the oil content value of tobacco leaf samples through expert scoring;

[0019] A spectrum acquisition module is used to obtain visible light spectrum information of a tobacco leaf sample or a tobacco leaf to be tested using spectral technology;

[0020] The interest extraction module is used to perform black and white plate radiation correction on the visible light spectrum information and extract pixels with characteristic bands within 500.14 to 649.75 nm and spectral reflectance values ​​less than 0.3, or with characteristic bands within 700.52 to 799.26 nm and spectral reflectance values ​​greater than or equal to 0.5 as regions of interest based on the threshold segmentation method. The mean spectral reflectance of all pixels within the region of interest is calculated as a single spectral curve representing the tobacco leaf sample.

[0021] Spectrum preprocessing module, used to preprocess the spectrum curve;

[0022] A part classification model building module is used to build a part classification model based on the preprocessed spectral curve using a support vector classification modeling method;

[0023] The flue-cured tobacco oil content score prediction model construction module is used to classify the spectral curves according to the parts, and based on the spectral curves of different parts and the corresponding oil content values, use the partial least squares regression modeling method to construct the flue-cured tobacco oil content score prediction model for the corresponding parts;

[0024] The tobacco leaf oil content prediction module is used to call the spectrum acquisition module, the interest extraction module and the spectrum preprocessing module to process the tobacco leaves to be tested, obtain the spectral curve and call the part classification model to identify the parts of the tobacco leaves to be tested; according to the part identification results, the flue-cured tobacco oil content score prediction model of the corresponding part is called to process the spectral curve to obtain the oil content score prediction value of the tobacco leaves to be tested.

[0025] To achieve the above-mentioned objectives, the third aspect of the present invention provides a computer device comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; the memory is used to store computer programs; and the processor is used to implement the non-destructive quantitative characterization method for oil content in different parts of flue-cured tobacco as described in the first aspect when executing the program stored in the memory.

[0026] To achieve the above objectives, the fourth aspect of the present invention provides a computer-readable storage medium having instructions stored thereon, which, when executed by one or more processors, enables the processors to execute the non-destructive quantitative characterization method for oil content in different parts of flue-cured tobacco as described in the first aspect.

[0027] To achieve the above-mentioned object, the fifth aspect of the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the non-destructive quantitative characterization method for oil content in different parts of flue-cured tobacco as described in the first aspect.

[0028] The beneficial effects of the present invention are:

[0029] The present invention uses the visible light region (400-780nm) spectrum to explore its applicability for oil characterization, circumventing the high cost constraints of near-infrared imaging equipment while improving detection efficiency. Threshold segmentation is performed based on threshold conditions where the characteristic band is within 500.14-649.75nm and the spectral reflectance value is less than 0.3, or the characteristic band is within 700.52-799.26nm and the spectral reflectance value is greater than 0.5. This effectively removes the main stem and shadow part of the tobacco leaf and effectively extracts and retains the tobacco leaf surface area.

[0030] Independent prediction models for flue-cured tobacco oil content scores were constructed for the upper, middle, and lower leaves, solving the problem of high RMSE caused by traditional modeling that ignores the heterogeneity of tobacco leaf parts.

[0031] There is no need for destructive pre-treatment such as grinding and compaction, and non-destructive testing of the tobacco leaves in their original state can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a flow chart of the non-destructive quantitative characterization method for oil content in different parts of flue-cured tobacco.

[0033] Figure 2 Schematic diagram of the hyperspectral acquisition system;

[0034] Figure 3 Schematic diagram of hyperspectral acquisition of tobacco leaf images;

[0035] Figure 4 Filter map for the region of interest of tobacco leaves;

[0036] Figure 5is the average spectrum of the spectral parts;

[0037] Figure 6 is the confusion matrix of the training set and test set of the part model (fifth fold);

[0038] Figure 7 The scatter plot of the true value and predicted value of the upper leaf oil score model (fifth fold);

[0039] Figure 8 This is a scatter plot of the true and predicted values ​​of the oil content score model for the middle leaves (fifth fold);

[0040] Figure 9 This is a scatter plot of the true and predicted values ​​of the lower leaf oil score model (fifth fold).

[0041] In the figure: 1. Hyperspectral camera; 2. Halogen lamp; 3. Precision stage; 4. Control computer. DETAILED DESCRIPTION

[0042] To make the purpose, technical solutions and advantages of the present invention clearer, the implementation process of the present invention is described in detail below in conjunction with the experimental process. The present invention uses visible light hyperspectral imaging technology to predict the oil content of flue-cured tobacco leaves.

[0043] Specifically, such as Figure 1 As shown, this application uses a hyperspectral camera to collect visible light spectral data and performs a series of preprocessing and modeling, such as black and white plate radiation correction, threshold screening to extract the effective tobacco leaf surface part, and SG convolution smoothing and SS normalization combined spectral preprocessing, using support vector classification and partial least squares regression modeling, etc., which can achieve accurate prediction of the classification of the upper, middle and lower parts of flue-cured tobacco and the oil content score of flue-cured tobacco by part. During the data processing process, 5-fold cross-validation is used to divide the data set to enhance the generalization of the model, and the grid search method is used to optimize the hyperparameters to improve the model performance, ensuring the robustness and accuracy of the model. It provides a reliable technical means for the quantitative characterization of flue-cured tobacco oil content, improves the scientificity and accuracy of flue-cured tobacco quality assessment, and provides innovative technical support for the tobacco industry in flue-cured tobacco quality control.

[0044] Example 1

[0045] Experimental design:

[0046] Tobacco leaf sampling: 205 representative sampling sites across 22 tobacco-growing provinces nationwide were selected, and samples of stem tobacco leaves were collected from the upper, middle, and lower parts of the leaves, resulting in a total of 615 stem tobacco leaf samples (205 from each part). Forty-four stem tobacco leaf samples from tobacco-growing areas in Henan Province were also selected as an external validation set.

[0047] Instrument Configuration: Spectral data were collected using a FigSpec-17 hyperspectral camera 1 (wavelength 400-1000nm, resolution 2.02nm). The spectral acquisition system consists of a closed acquisition cabinet, halogen lamps 2 located on both sides of the cabinet, a precision stage 3, and a control computer 4. The cabinet is sealed except for the opening where the stage faces the camera. A fixed-angle light source layout was used to eliminate ambient light interference and ensure uniform illumination on the tobacco leaf surface. Figure 2 .

[0048] S1: oil score assignment;

[0049] The tobacco leaf samples were pretreated according to the grading standard of "Flue-cured Tobacco" (GB2635-1992). After the moisture content was balanced to 16%-18% in a constant humidity environment, at least three professional grading technicians independently scored the tobacco leaf oil content of each tobacco leaf sample according to a 10-point system, and the arithmetic mean was taken as the final oil score.

[0050] The statistical characteristics of the oil content scores of different parts of flue-cured tobacco are shown in Table 1. The oil content scores of different parts showed significant differences (P < 0.01), with the upper leaves having the highest score (6.60 ± 0.37) and the lower leaves having the lowest score (5.16 ± 0.86).

[0051] Table 1 Statistical analysis of the oil content of flue-cured tobacco leaves with different sample sizes ①

[0052]

[0053] Note: ① In the table, numbers in the same column followed by different letters indicate that the differences reached extremely significant levels (P < 0.01).

[0054] S2: Flue-cured tobacco hyperspectral collection;

[0055] Before collection, the tobacco leaf samples were spread flat on the stage and kept in a naturally stretched state. In particular, the spectrum data of each tobacco leaf was collected twice. The collected spectrum data can be found in Figure 3 .

[0056] Specifically, the collection steps are as follows: spreading the tobacco leaf sample naturally on the stage, and using a hyperspectral camera to collect spectral data of the first measurement surface of the tobacco leaf sample at a wavelength of 400 to 1000 nm;

[0057] The tobacco leaf sample is flipped upside down 180 degrees and laid flat again, and the spectral data of the second measurement surface of the tobacco leaf sample is collected again using a hyperspectral camera at a wavelength of 400 to 1000 nm;

[0058] The spectrum data of the first measurement surface and the spectrum data of the second measurement surface are averaged to obtain the visible light spectrum information.

[0059] In this embodiment, the spectrum is scanned 1230 times in total, and the effective channels are 400.43 to 779.14 nm, with a total of 183 bands.

[0060] Step S3: hyperspectral data processing;

[0061] To eliminate noise interference, a black and white plate radiometric calibration was performed after data acquisition: dark reference data (reflectance close to 0%) was collected in a completely dark environment with all light sources turned off. After calibration with the light source turned on and a high-reflectance white plate (reflectance ≥ 99%) was covered, the collected hyperspectral image was calibrated based on the following formula:

[0062]

[0063] Where: I is the corrected hyperspectral data; / 0 is the collected original hyperspectral data; B is the calibration data in a completely dark environment; W is the standard whiteboard data.

[0064] Since the spectrum of tobacco leaves is directly collected in bunches in this application, there are shadow areas caused by the stacking of tobacco leaves. The shadow area will make the spectral reflectance of this part different from normal, thereby affecting the subsequent oil identification. Therefore, if you want to accurately select the region of interest (ROI) of the tobacco leaf surface, you must avoid the main stem of the tobacco leaf while removing the shadow part caused by the stacking of the tobacco leaves, and cover the effective leaf surface of the tobacco leaf as completely as possible.

[0065] This embodiment uses a threshold segmentation method to extract the main body of the tobacco leaf surface as the region of interest. Specifically, in one embodiment, the threshold condition of the threshold segmentation method is: the characteristic band is located within 500.14 to 649.75 nm and the spectral reflectance value is less than 0.3, or the characteristic band is located within 700.52 to 799.26 nm and the spectral reflectance value is greater than or equal to 0.5. That is, the pixel points whose characteristic band is located within 500.14 to 649.75 nm and the spectral reflectance value is less than 0.3, or the characteristic band is located within 700.52 to 799.26 nm and the spectral reflectance value is greater than or equal to 0.5 are extracted as the region of interest. Specifically, Figure 4 Schematic diagram of the region of interest extracted in step S3.

[0066] After extracting the region of interest, calculate the mean spectral reflectance of the pixels in the region of interest that meets all the above conditions as a single spectral curve representing the tobacco leaf sample. Figure 5 .

[0067] S4: preprocessing the spectral curve;

[0068] Preprocessing of spectral curves can improve the quality of spectral data, reduce noise interference, and ensure that spectral analysis is accurate, reliable, and comparable.

[0069] There are many methods for preprocessing spectral curves, such as smoothing, noise reduction, spectral line repair, standardization and spectrum normalization, standard normal variate (SNV), multiplicative scatter correction (MSC), Fourier transform (FT), wavelet transform (WT), orthogonal signal correction (OSC), etc. Specifically, in this embodiment, the spectral curve is preprocessed using a spectral preprocessing method based on SG convolution smoothing combined with standardization (SG-SS).

[0070] It is understood that in other embodiments, a combination of one or more pre-processing methods may be selected according to needs to process the spectral curve.

[0071] Step S5: constructing a flue-cured tobacco part model;

[0072] The data set was divided into 5 independent subsets using 5-fold cross-validation to enhance the generalization of the model: the original data set was randomly divided into 5 independent subsets, one of which was selected as the validation set in each iteration, and the remaining 4 subsets were combined into the training set. The training-validation process was repeated until each group completed the validation. Finally, the average performance parameter of the 5 validation results was used as the benchmark for evaluating the generalizability of the model.

[0073] Specifically, the flue-cured tobacco part classification model uses the Support Vector Classification (SVR) model. Furthermore, grid search was used to optimize the Support Vector Classification (SVC) hyperparameters. Using training set accuracy as the optimization metric, the optimal hyperparameter combination of {'C': 10, 'kernel': 'linear', 'gamma': 'scale'} was determined. Finally, classification models for the upper, middle, and lower parts of the flue-cured tobacco were constructed.

[0074] The qualitative evaluation indicators used include confusion matrix, accuracy, precision, recall, and F1 scores. The results are shown in Table 2.

[0075] Table 2 Verification results of the tobacco leaf position discrimination model of the present invention

[0076]

[0077] Step S6: constructing a scoring model for flue-cured tobacco oil content at different parts;

[0078] Specifically, the spectral curves were classified according to the parts, and based on the spectral curves of different parts and the corresponding oil content values, the partial least squares regression modeling method was used to construct a flue-cured tobacco oil content score prediction model for the corresponding parts.

[0079] Specifically, the 5-fold cross validation was also used to train the flue-cured tobacco oil score prediction model. Specifically, the dataset was randomly divided into 5 subsets;

[0080] During the model training process, the grid search method (GridSearch) was used to globally optimize the core parameters of the PLSR model, and the cross-validation root mean square error (CV-RMSE) was used as the optimization indicator. Finally, it was determined that the number of latent variables n_components = 45 was the optimal hyperparameter combination, and prediction models for the oil content scores of the upper, middle and lower parts of the flue-cured tobacco were constructed respectively.

[0081] For details, see Figure 6-Figure 9 , Figure 6 is the confusion matrix of the training set and test set of the part model (fifth fold), Figure 7 This is a scatter plot of the true value and predicted value of the upper leaf oil score model (fifth fold). Figure 8 This is a scatter plot of the true value and predicted value of the oil content score model for the middle leaves (fifth fold). Figure 9 This is a scatter plot of the true and predicted values ​​of the lower leaf oil score model (fifth fold).

[0082] The quantitative model performance evaluation indicators are root mean square error (RMSE), coefficient of determination (R 2 ) and relative analysis error (Relative Percentage Deviation, RPD), the evaluation index results are shown in Table 3.

[0083] Table 3 Verification results of the tobacco leaf oil score model of the present invention

[0084]

[0085] Step S7: predicting the oil content score of external sample sets;

[0086] In this example, 21 unknown flue-cured tobacco leaf samples were selected as an external validation set. After sequentially passing through steps S2, S3, and S4, their feature data was input into the part classification model constructed in step S5 for classification. After obtaining the tobacco leaf part information, the system automatically invoked the corresponding part oil score prediction model established in step S6, ultimately outputting the oil content prediction results shown in Table 4.

[0087] Tobacco leaf oil content, a key indicator of appearance quality, is essentially a qualitative description quantified by professional graders through visual and haptic evaluation. This evaluation process inevitably involves subjective judgment. In tobacco industry grading standards, a relative deviation of less than 10% between the predicted and measured values ​​is considered accurate. The relative deviation achieved in this application was 7.06%. 72.72% of the total tobacco samples fell within this range, demonstrating that the visible light spectrum prediction method of the present invention has good accuracy and practical value, providing effective technical support for intelligent tobacco quality testing.

[0088] Table 4 Prediction results of flue-cured tobacco oil content score

[0089]

[0090]

[0091]

[0092] It should be understood that, although the steps in the flowcharts of the above embodiments are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the flowcharts of the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0093] Example 2

[0094] Based on the same inventive concept, embodiments of the present application also provide a device for non-destructive quantitative characterization of oil content in different parts of flue-cured tobacco, for implementing the aforementioned method for non-destructive quantitative characterization of oil content in different parts of flue-cured tobacco. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the device for non-destructive quantitative characterization of oil content in different parts of flue-cured tobacco provided below can be found in the aforementioned method for non-destructive quantitative characterization of oil content in different parts of flue-cured tobacco, and will not be further elaborated here.

[0095] Specifically, the non-destructive quantitative characterization device for oil content in different parts of flue-cured tobacco based on visible light spectroscopy includes:

[0096] The oil content value acquisition module is used to obtain the oil content value of tobacco leaf samples through expert scoring;

[0097] A spectrum acquisition module is used to obtain visible light spectrum information of a tobacco leaf sample or a tobacco leaf to be tested using spectral technology;

[0098] The interest extraction module is used to perform black and white plate radiation correction on the visible light spectrum information, and extract pixels with characteristic bands within 500.14 to 649.75 nm and spectral reflectance values ​​less than 0.3, or with characteristic bands within 700.52 to 799.26 nm and spectral reflectance values ​​greater than or equal to 0.5 as regions of interest based on the threshold segmentation method. The mean spectral reflectance of all pixels in the region of interest is calculated as a single spectral curve representing the tobacco leaf sample.

[0099] Spectrum preprocessing module, used to preprocess the spectrum curve;

[0100] A part classification model building module is used to build a part classification model based on the preprocessed spectral curve using a support vector classification modeling method;

[0101] The flue-cured tobacco oil content score prediction model construction module is used to classify the spectral curves according to the parts, and based on the spectral curves of different parts and the corresponding oil content values, use the partial least squares regression modeling method to construct the flue-cured tobacco oil content score prediction model for the corresponding parts;

[0102] The tobacco leaf oil content prediction module is used to call the spectrum acquisition module, the interest extraction module and the spectrum preprocessing module to process the tobacco leaves to be tested, obtain the spectral curve and call the part classification model to identify the parts of the tobacco leaves to be tested; according to the part identification results, the flue-cured tobacco oil content score prediction model of the corresponding part is called to process the spectral curve to obtain the oil content score prediction value of the tobacco leaves to be tested.

[0103] Example 3

[0104] This embodiment provides a computer device, which may 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.

[0105] The processor of the computer device is used to provide computing and control capabilities.

[0106] The computer device's memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium.

[0107] The input / output interface of the computer device is used to exchange information between the processor and external devices.

[0108] The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner. The wireless manner can be achieved through WIFI, mobile cellular network, NFC (near field communication) or other technologies.

[0109] When the computer program is executed by a processor, it implements the non-destructive quantitative characterization method for oil content in different parts of flue-cured tobacco described in Example 1.

[0110] The display unit of the computer device is used to produce a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, a keypad, a trackball, or a touchpad provided on the computer device housing, or an external keyboard, touchpad, or mouse.

[0111] Example 4

[0112] Based on the above embodiments, this embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for non-destructive quantitative characterization of oil content in different parts of flue-cured tobacco described in Example 1 is implemented.

[0113] Example 5

[0114] Based on the above embodiments, this embodiment provides a computer program product, including a computer program. When the computer program is executed by a processor, the method for non-destructive quantitative characterization of oil content in different parts of flue-cured tobacco described in Example 1 is implemented.

[0115] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may 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 may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0116] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, 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.

[0117] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solutions of the present invention. They should all be included in the scope of the technical solutions claimed for protection by the present invention.

Claims

1. A non-destructive quantitative characterization method for oil content in different parts of flue-cured tobacco based on visible light spectroscopy, characterized in that: The following steps are involved: Obtain tobacco leaf samples from different parts of the tobacco leaf, obtain the oil content of the tobacco leaf samples through expert scoring, and use spectral technology to obtain the visible light spectrum information of the tobacco leaf samples; Visible light spectrum information was corrected for black and white plate radiometry. Pixels with characteristic bands between 500.14 and 649.75 nm and spectral reflectance values ​​less than 0.3, or with characteristic bands between 700.52 and 799.26 nm and spectral reflectance values ​​greater than or equal to 0.5, were extracted as regions of interest (ROIs) based on a threshold segmentation method. The mean spectral reflectance of all pixels within the ROI was calculated as a single spectral curve representing the tobacco leaf sample. The spectral curve is preprocessed, and based on the preprocessed spectral curve, a part classification model is constructed using a support vector classification modeling method; The spectral curves were classified according to the parts of the tobacco leaves. Based on the spectral curves of different parts and the corresponding oil content values, a partial least squares regression model was used to construct a prediction model for the oil content of the corresponding parts of the tobacco leaves. For the tobacco leaves to be tested, spectral technology is used to obtain visible light spectral information, and the visible light spectral information is corrected for black and white plate radiation. Based on the threshold segmentation method, the main body of the tobacco leaf surface is extracted as the region of interest, and the mean spectral reflectance of all pixels in the region of interest is calculated as the spectral curve representing the tobacco leaves to be tested; the spectral curve is imported into the part classification model to identify the part of the tobacco leaves to be tested; according to the part identification result, the flue-cured tobacco oil content score prediction model of the corresponding part is called to process the spectral curve to obtain the oil content score prediction value of the tobacco leaves to be tested.

2. The non-destructive quantitative characterization method for oil content in different parts of flue-cured tobacco based on visible light spectroscopy according to claim 1, characterized in that: The specific steps of using spectroscopy technology to obtain visible light spectrum information include: Spread the tobacco leaves naturally on the stage, and use a hyperspectral camera to collect spectral data of the first measurement surface of the tobacco leaves at a wavelength of 400-1000 nm. The tobacco leaves were flipped upside down 180° and laid flat again. The spectral data of the second measurement surface of the tobacco leaves were collected again using a hyperspectral camera at a wavelength of 400-1000 nm. The spectrum data of the first measuring surface and the spectrum data of the second measuring surface are averaged to serve as the visible light spectrum information of the tobacco leaves.

3. The non-destructive quantitative characterization method for oil content in different parts of flue-cured tobacco based on visible light spectroscopy according to claim 1 or 2, characterized in that: The spectral curves were preprocessed using convolution smoothing and normalization.

4. The non-destructive quantitative characterization method for oil content in different parts of flue-cured tobacco based on visible light spectroscopy according to claim 1 or 2, characterized in that: Obtaining the oil content value of the bamboo tobacco leaf sample includes: balancing the moisture content of the bamboo tobacco leaf sample to 16% to 18% in a constant humidity environment, pre-treating the bamboo tobacco leaf sample according to the "Flue-cured Tobacco" (GB2635-1992) grading standard, and having multiple professional grading technicians score the oil content of the tobacco leaves according to a 10-point system, and using the average of the scores as the oil content value of the bamboo tobacco leaf sample.

5. The non-destructive quantitative characterization method for oil content in different parts of flue-cured tobacco based on visible light spectroscopy according to claim 1, characterized in that: Use 5-fold cross-validation to train the part classification model or the flue-cured tobacco oil score prediction model; During the part classification model training process, the validation set accuracy was used as the optimization indicator, and the optimal hyperparameter combination of the model was determined through the grid search method. The confusion matrix, accuracy, precision, recall rate, and F1 score evaluation indicators were used to qualitatively evaluate the model performance. During the training process of the flue-cured tobacco oil score prediction model, the cross-validation root mean square error was used as the optimization indicator, and the grid search method was used to globally optimize the core parameters of the model to determine the optimal hyperparameter combination. The root mean square error, determination coefficient and relative analytical error evaluation indicators were used to quantitatively evaluate the model performance.

6. A non-destructive quantitative characterization device for oil content in different parts of flue-cured tobacco based on visible light spectroscopy, characterized in that: include: The oil content value acquisition module is used to obtain the oil content value of tobacco leaf samples through expert scoring; A spectrum acquisition module is used to obtain visible light spectrum information of a tobacco leaf sample or a tobacco leaf to be tested using spectral technology; The interest extraction module is used to perform black and white plate radiometric correction on the visible light spectrum information and, based on the threshold segmentation method, extract pixels with characteristic bands between 500.14 and 649.75 nm and spectral reflectance values ​​less than 0.3, or with characteristic bands between 700.52 and 799.26 nm and spectral reflectance values ​​greater than or equal to 0.5 as regions of interest. The mean spectral reflectance of all pixels within the region of interest is calculated as a single spectral curve representing the tobacco leaf sample. Spectrum preprocessing module, used to preprocess the spectrum curve; A part classification model building module is used to build a part classification model based on the preprocessed spectral curve using a support vector classification modeling method; The flue-cured tobacco oil content score prediction model construction module is used to classify the spectral curves according to the parts, and based on the spectral curves of different parts and the corresponding oil content values, use the partial least squares regression modeling method to construct the flue-cured tobacco oil content score prediction model for the corresponding parts; The tobacco leaf oil content prediction module is used to call the spectrum acquisition module, the interest extraction module and the spectrum preprocessing module to process the tobacco leaves to be tested, obtain the spectral curve and call the part classification model to identify the parts of the tobacco leaves to be tested; according to the part identification results, the flue-cured tobacco oil content score prediction model of the corresponding part is called to process the spectral curve to obtain the oil content score prediction value of the tobacco leaves to be tested.

7. A computer device, characterized in that: The processor, the communication interface, the memory and the communication bus are connected to each other via the communication bus. Memory for storing computer programs; The processor is configured to implement the non-destructive quantitative characterization method for oil content in different parts of flue-cured tobacco as claimed in any one of claims 1 to 5 when executing the program stored in the memory.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for non-destructive quantitative characterization of oil content in different parts of flue-cured tobacco according to any one of claims 1 to 5 is implemented.

9. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for non-destructive quantitative characterization of oil content in different parts of flue-cured tobacco according to any one of claims 1 to 5 is implemented.

Images