Wuyi narcissus baking color quality intelligent discrimination method and system
By optimizing the near-infrared spectrum using a genetic algorithm and combining Savitzky-Golay smoothing filtering and multiple scattering correction, a GA-PLS-SVM hybrid model was constructed. This solved the problem of rapid, accurate, and non-destructive discrimination of the roasted color of Wuyi Narcissus tea, enabling real-time monitoring and automatic alarm of the tea roasting process and improving the intelligence level of the production line.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 武夷学院
- Filing Date
- 2026-01-27
- Publication Date
- 2026-05-08
AI Technical Summary
Existing technologies are insufficient for rapid, accurate, and non-destructive identification of the color of roasted Wuyi narcissus flowers. Sensory evaluation is highly subjective, instrumental analysis is complex and costly, and traditional near-infrared spectroscopy models have insufficient predictive capabilities.
A genetic algorithm was used to optimize the near-infrared spectrum, combined with Savitzky-Golay smoothing filtering and multiple scattering correction, key characteristic wavelengths were screened, and a GA-PLS-SVM hybrid prediction model was constructed to achieve intelligent identification of the roasting color of tea leaves, and real-time control was achieved through an online monitoring system.
It enables objective and quantitative discrimination of the color of roasted Wuyi narcissus, improves the accuracy and robustness of the model, supports real-time monitoring and automatic alarm, and enhances the level of automation and intelligence in production.
Smart Images
Figure CN121994747A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of tea processing technology, spectral analysis technology, and artificial intelligence technology, specifically to a method and system for intelligently judging the roasted color quality of Wuyi narcissus tea based on near-infrared spectroscopy optimized by a genetic algorithm. Background Technology
[0002] Wuyi Narcissus is a representative variety of Minbei Oolong tea, renowned for its unique "rocky" flavor and mellow taste. Roasting is the core process for shaping the quality and style of Wuyi Narcissus, using heat to promote the transformation of substances within the tea leaves and the formation of aroma. Color is one of the most direct and important sensory indicators for evaluating roasting quality, highly correlated with the tea's taste, aroma, and biochemical content. Precisely controlling the degree of roasting to achieve standardization and stabilization of color quality is key to enhancing the market competitiveness of Wuyi Narcissus.
[0003] Currently, the determination of the roasting degree of Wuyi Narcissus tea mainly relies on sensory evaluation and physicochemical analysis. Sensory evaluation (as per GB / T 23776-2018) depends on the experience and subjective judgment of tea tasters, and is easily affected by individual differences, environment, and fatigue. It lacks objective quantitative standards, making it difficult to achieve precise control and quality traceability in large-scale production. Instrumental analysis methods, such as high-performance liquid chromatography (HPLC) for determining characteristic pigments or catechin components, while providing accurate results, involve complex sample pretreatment, are time-consuming and labor-intensive, and costly. Furthermore, they are destructive tests and cannot meet the needs of rapid, non-destructive, and real-time monitoring on production lines.
[0004] Near-infrared spectroscopy (FIRS) is a rapid and non-destructive analytical technique that has been widely used in the quality testing of agricultural products. However, directly applying it to the color determination of roasted Wuyi narcissus tea faces significant challenges: First, the near-infrared spectral information of tea is complex, containing a large amount of noise and background interference (such as moisture and particle size scattering) unrelated to color; second, traditional full-band modeling is prone to overfitting and reduced predictive ability; and third, the relationship between color (represented by the CIE-Lab color space) and spectrum is not a simple linear one, and a single linear model cannot accurately capture its complex nonlinear mapping.
[0005] Therefore, there is an urgent need to develop a new method and system that can overcome the above-mentioned defects and achieve rapid, accurate, non-destructive, and intelligent judgment of the color and quality of Wuyi Narcissus tea after roasting, so as to promote the digital and intelligent upgrading of Wuyi Rock Tea processing. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for intelligently judging the roasted color quality of Wuyi Narcissus tea based on near-infrared spectroscopy optimized by genetic algorithm, so as to achieve accurate, rapid and non-destructive judgment of the roasted color of tea leaves, and provide technical support for the optimization and real-time control of the roasting process.
[0007] To achieve the above objectives, the present invention adopts the following technical solution: Firstly, the present invention provides an intelligent method for judging the roasted color quality of Wuyi daffodils based on near-infrared spectroscopy optimized by a genetic algorithm, comprising the following steps: (1) Data acquisition: Near-infrared spectral data of Wuyi Narcissus tea samples in the 400-4000 cm⁻¹ band were acquired; at the same time, the CIE-Lab color space parameters corresponding to the tea samples were measured as quality benchmark values. (2) Spectral preprocessing: The near-infrared spectral data are subjected to joint preprocessing of Savitzky-Golay smoothing filtering and multiple scattering correction; (3) Feature wavelength optimization: Use a genetic algorithm to screen key feature wavelengths that are significantly related to the CIE-Lab chromaticity parameters from the preprocessed spectral data; 4. Construction of Hybrid Prediction Model: Based on the key feature wavelength data and their corresponding CIE-Lab chromaticity parameters, a GA-PLS-SVM hybrid prediction model is constructed by coupling partial least squares regression and support vector machine. (5) Intelligent discrimination and application: Using the trained GA-PLS-SVM hybrid prediction model, the roasting color quality of unknown Wuyi daffodil samples can be intelligently discerned, and their roasting process parameters can be inverted.
[0008] Further, in step S1, the near-infrared spectral data is acquired using a Fourier transform near-infrared spectrometer with a scanning resolution of 4-8 cm⁻¹ and 32 scans. Each sample is measured three times to obtain the average spectrum. The CIE-Lab colorimetric parameters are measured using a spectrophotometer.
[0009] Furthermore, in step S2, the window size of the Savitzky-Golay smoothing filter is 11, and the polynomial order is 2.
[0010] Further, in step S3, the parameters of the genetic algorithm are set as follows: population size of 50-100, crossover probability of 0.6-0.8, mutation probability of 0.01-0.05, and number of generations of evolution of 100-200.
[0011] Furthermore, in step S4, the construction of the GA-PLS-SVM hybrid prediction model specifically includes: first, using partial least squares regression to reduce the dimensionality of the key feature wavelength data and extracting the principal component scores as new feature variables; Then, the new feature variables are input into a support vector machine for nonlinear modeling.
[0012] Furthermore, the support vector machine uses a radial basis function as its kernel function, and its penalty factor C and kernel function parameter γ are optimized using a grid search method.
[0013] Furthermore, in step S5, the validity criterion for the intelligent discrimination is: the prediction determination coefficient R² of the model for the sample L*, a*, and b* values is not less than 0.90 in the calibration set and not less than 0.85 in the independent validation set.
[0014] Furthermore, the method also includes step S6: online monitoring and feedback control: the trained GA-PLS-SVM hybrid prediction model is deployed in the online monitoring system to collect the near-infrared spectrum of tea samples from the production line in real time and make predictions; when the color value predicted by the model deviates from the preset process range, the system automatically triggers a deviation alarm.
[0015] Secondly, the present invention provides an intelligent system for judging the color and quality of Wuyi narcissus roasting for implementing the above method, comprising: (1) Spectral acquisition module, used to acquire near-infrared spectral data of tea samples in the 400-4000 cm⁻¹ band; (2) Spectral preprocessing module, used to perform Savitzky-Golay smoothing filtering and multiple scattering correction algorithm on spectral data; (3) Feature wavelength optimization module, with built-in genetic algorithm program, used to screen key feature wavelengths from the preprocessed spectrum; (4). The intelligent discrimination core module has a built-in trained GA-PLS-SVM hybrid prediction model, which is used to output the discrimination results of roasting color quality and the inversion value of process parameters based on the input key feature wavelength data. (5) Output and alarm module, used to display the judgment result and issue an alarm when the result exceeds the preset threshold.
[0016] Furthermore, the system is connected to the tea roasting production line, and the spectral acquisition module is an online diffuse reflectance probe, enabling real-time, non-destructive monitoring and feedback control of the roasting process.
[0017] The beneficial effects of the present invention are: (1) Objective and accurate: By combining near-infrared spectroscopy with CIE-Lab colorimetry, the objective and quantitative judgment of the roasted color of Wuyi narcissus is realized, overcoming the subjectivity of sensory evaluation.
[0018] (2) High efficiency and non-destructive: The method is fast, requires no complicated pretreatment, does not damage the sample, and is suitable for online monitoring of the production line.
[0019] (3) Superior model: The use of genetic algorithm to optimize feature wavelength effectively removes irrelevant information and improves model accuracy and robustness; the dimensionality reduction capability of PLSR coupled with the nonlinear fitting advantage of SVM results in a GA-PLS-SVM hybrid model whose prediction performance is significantly better than that of a single model or a full-spectrum model.
[0020] (4) Functional integration: It can not only judge the color quality, but also reverse the key process parameters such as roasting intensity and time, providing a direct basis for process optimization.
[0021] (5) Intelligent control: The system can realize real-time monitoring and automatic alarm of the roasting process, which helps to stabilize product quality and improve the level of production automation and intelligence. Attached Figure Description
[0022] Figure 1 The overall technical flowchart of the intelligent method for judging the color and quality of Wuyi narcissus roasting provided in the embodiments of the present invention is shown.
[0023] Figure 2 The images show the near-infrared spectra of Wuyi daffodil samples with different roasting degrees in the embodiments of the present invention.
[0024] Figure 3 This is a schematic diagram of the evolution curve and the final selected feature wavelength distribution of the genetic algorithm screening process for key feature wavelengths in an embodiment of the present invention.
[0025] Figure 4 The scatter plot shows the prediction results of L*, a*, and b* values of the GA-PLS-SVM hybrid model constructed for embodiments of the present invention on an independent validation set.
[0026] Figure 5 This is a schematic diagram illustrating the online application of the intelligent color quality judgment system for Wuyi daffodils provided in an embodiment of the present invention. Detailed Implementation
[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0028] The following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.
[0029] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0030] In the description of this invention, it should be noted that the terms "upper," "inner," "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of this invention is usually placed when in use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.
[0031] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," and "connect" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0032] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.
[0033] like Figure 1 As shown, a method for intelligently judging the color quality of roasted Wuyi narcissus based on near-infrared spectroscopy optimized by genetic algorithm is disclosed. The method includes the following steps.
[0034] S1: Data Acquisition: Near-infrared spectral data of Wuyi Narcissus tea samples in the 400-4000 cm⁻¹ band were acquired. Specifically, a near-infrared spectrometer (such as the Thermo Fisher Antaris II) was used to scan the Wuyi Narcissus tea samples, with a scanning resolution of 8 cm⁻¹ and 32 scans. Each sample was scanned three times, and the average spectrum was taken as the spectral data for that sample. The CIE-Lab color space parameters corresponding to the Wuyi Narcissus tea samples were determined as quality benchmark values. Specifically, a YS6010 colorimeter was used to measure the Wuyi Narcissus tea samples, measuring three times and taking the average value as the L, a, and b values for that sample.
[0035] S2: Data Preprocessing: The near-infrared spectral data are preprocessed using a combination of Savitzky-Golay smoothing filtering and multiple scattering correction. Specifically, the PLS Toolbox in Matlab can be used for preprocessing. The window size for Savitzky-Golay smoothing is set to 11, and the polynomial order is set to 2. Multiple scattering correction (MSC) can eliminate baseline drift caused by scattering.
[0036] Example 1: Sample preparation and data acquisition, spectral database such as Figure 1 As shown.
[0037] Wuyi Narcissus tea leaves were selected as raw materials and roasted at different temperature (700-110℃) and time (30-300 min, 30 min intervals) gradients on the same equipment to prepare 90 samples covering different roasting degrees, including light, medium, and full roasting. The tea infusion was brewed according to GB / T 23776-2018 standard, and the CIE-Lab values were determined using a YS6010 benchtop spectrophotometer. A Thermo Fisher Nicolet iS50 FT-IR spectrometer with an integrating sphere detector was used, with a scanning range of 4000-400 cm⁻¹, a resolution of 4 cm⁻¹, and 32 scans. Each sample was measured three times, and the average spectrum was obtained.
[0038] Example 2: Model building and optimization.
[0039] Ninety samples were randomly divided into a training set (60 samples) and an independent validation set (30 samples) at a 2:1 ratio. Data processing and modeling were performed using MATLAB R2021a software. Preprocessing: The original spectra, SG smoothing, SG + first derivative, SG + first derivative + SNV, SG + first derivative + MSC, and the combined SG + first derivative + SNV + MSC method used in this invention were compared. The results show that the combined preprocessing method resulted in the highest correlation between the spectra and chromaticity values, and the smoothest baseline.
[0040] Table 1. Comparison of the predictive performance of different pretreatment methods on the color of Wuyi Narcissus tea soup.
[0041]
[0042] S3: Characteristic wavelength selection, such as Figure 3 As shown.
[0043] A genetic algorithm was used to screen key feature wavelengths significantly correlated with the CIE-Lab chromaticity parameters from the preprocessed spectral data, constructing a wavelength-chromaticity correlation matrix. Specifically, the genetic algorithm program could be written using Matlab software. The parameters of the genetic algorithm were set as follows: population size of 80, crossover probability of 0.7, mutation probability of 0.03, and number of generations of evolution of 150. The fitness function could be the root mean square error (RMSE) of the prediction model.
[0044] S4: Model building.
[0045] Based on the key characteristic wavelength data and the corresponding CIE-Lab chromaticity parameters, a GA-PLS-SVM hybrid prediction model is constructed by coupling partial least squares regression and support vector machine. In practice, the PLSToolbox and LIBSVM toolbox in Matlab software can be used for modeling. The model performance is shown in Table 2.
[0046] Table 2. Comparison of the predictive performance of different model methods on the color of Wuyi Narcissus tea soup.
[0047]
[0048] First, partial least squares regression is used to reduce the dimensionality of the key feature wavelength data, and the principal component scores are extracted as new feature variables. The number of principal components can be determined through cross-validation. Then, the new feature variables are input into a support vector machine for nonlinear modeling. The support vector machine uses a radial basis function (RBF) as its kernel function, and its penalty factor C and kernel function parameter γ are optimized using a grid search method. The grid search range can be set as C = [0.1, 1, 10, 100], γ = [0.01, 0.1, 1, 10].
[0049] S5: Model validation and application, such as Figure 4 As shown.
[0050] The GA-PLS-SVM hybrid prediction model is used to intelligently determine the roasting color quality of unknown Wuyi daffodil samples and inversely determine their roasting intensity and time parameters. Specifically, the near-infrared spectral data of the collected unknown samples can be input into the trained GA-PLS-SVM model to obtain the predicted L, a, and b values. The intelligent determination criterion is: the model is considered effective when the prediction determination coefficients R² for the L, a, and b values of the samples are all not less than 0.90 on the calibration set and not less than 0.85 on the independent validation set.
[0051] The roasting process parameters can be inverted by establishing a regression model between the L, a, and b values and the roasting intensity and time parameters.
[0052] S6: Online monitoring and alarm.
[0053] The GA-PLS-SVM hybrid prediction model is deployed in an online monitoring system to collect near-infrared spectra of tea samples from the production line in real time. When the colorimetric values predicted by the model deviate from the preset process range, the system automatically triggers a deviation alarm with an alarm sensitivity of no less than 82.3%. In specific implementation, an online diffuse reflectance probe (such as Brimrose Luminar 2030) can be used to collect the near-infrared spectra of tea samples, and the spectral data can be transmitted to a computer for processing and prediction.
[0054] like Figure 5 As shown, a smart system for judging the color and quality of Wuyi narcissus roasting for implementing the above method includes: (1) Near-infrared spectral acquisition module, used to acquire spectral data of tea samples in the 400-4000 cm⁻¹ band; (2) Spectral preprocessing module, used to execute Savitzky-Golay smoothing filtering and multiple scattering correction algorithm; (3) Feature wavelength optimization module, with built-in genetic algorithm program, used to screen key feature wavelengths; (4) Intelligent discrimination core module, with a built-in trained GA-PLS-SVM hybrid prediction model, is used to output the roasting color quality discrimination result and the roasting process parameter inversion value based on the input spectral data; (5) Result output and alarm module, used to display the judgment result and issue an alarm when the result exceeds the threshold.
[0055] The system is connected to the tea roasting production line, and the near-infrared spectral acquisition module is an online diffuse reflection probe, which realizes real-time, non-destructive monitoring and feedback control of the roasting process.
[0056] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.
Claims
1. A method for intelligently judging the roasted color quality of Wuyi narcissus based on near-infrared spectroscopy optimized by genetic algorithm, characterized in that, Includes the following steps: (1) Near-infrared spectral data of Wuyi Narcissus tea samples in the 400-4000 cm⁻¹ band were collected; (2) The CIE-Lab color space parameters corresponding to the Wuyi Narcissus tea sample were determined and used as the quality benchmark value; (3) Perform Savitzky-Golay smoothing filtering and multiple scattering correction on the near-infrared spectral data; (4) Use a genetic algorithm to screen key feature wavelengths that are significantly related to the CIE-Lab chromaticity parameters from the preprocessed spectral data and construct a wavelength-chromaticity correlation matrix; (5) Based on the key feature wavelength data and the corresponding CIE-Lab chromaticity parameters, a GA-PLS-SVM hybrid prediction model is constructed by coupling partial least squares regression and support vector machine; (6) Using the GA-PLS-SVM hybrid prediction model, the roasting color quality of unknown Wuyi daffodil samples is intelligently determined, and their roasting intensity and time parameters are inverted.
2. The intelligent method for judging the color and quality of Wuyi narcissus roasting according to claim 1, characterized in that, In step 4, the parameters of the genetic algorithm are set as follows: population size of 50-100, crossover probability of 0.6-0.8, mutation probability of 0.01-0.05, and number of generations of evolution of 100-200.
3. The intelligent method for judging the color and quality of Wuyi narcissus after roasting according to claim 1 or 2, characterized in that, In step 5), the construction of the GA-PLS-SVM hybrid prediction model is as follows: First, partial least squares regression is used to reduce the dimensionality of the key feature wavelength data and extract the principal component scores as new feature variables; then, the new feature variables are input into the support vector machine for nonlinear modeling.
4. The intelligent method for judging the color and quality of Wuyi narcissus roasting according to claim 3, characterized in that, The support vector machine uses a radial basis function as its kernel function, and its penalty factor C and kernel function parameter γ are optimized using a grid search method.
5. The intelligent method for judging the color and quality of Wuyi narcissus roasting according to claim 1, characterized in that, In step 6, the intelligent judgment criterion is: when the model's prediction determination coefficient R² for the values of samples L, a, and b is not less than 0.90 in the calibration set and not less than 0.85 in the independent validation set, the model is judged to be effective.
6. The intelligent method for judging the color and quality of Wuyi narcissus roasting according to claim 1 or 5, characterized in that, The method further includes step 7: deploying the GA-PLS-SVM hybrid prediction model on the online monitoring system, collecting the near-infrared spectrum of tea samples from the production line in real time, and automatically triggering a deviation alarm when the color value predicted by the model deviates from the preset process range, with an alarm sensitivity of not less than 82.3%.
7. A smart system for judging the color and quality of roasted Wuyi narcissus flowers for implementing the method according to any one of claims 1-6, characterized in that, include: (1) Near-infrared spectral acquisition module, used to acquire spectral data of tea samples in the 400-4000 cm⁻¹ band; (2) Spectral preprocessing module, used to execute Savitzky-Golay smoothing filtering and multiple scattering correction algorithm; (3) Feature wavelength optimization module, with built-in genetic algorithm program, used to screen key feature wavelengths; (4) Intelligent discrimination core module, with a built-in trained GA-PLS-SVM hybrid prediction model, is used to output the roasting color quality discrimination result and the roasting process parameter inversion value based on the input spectral data; (5) Result output and alarm module, used to display the judgment result and issue an alarm when the result exceeds the threshold.
8. The intelligent judgment system for the roasting color and quality of Wuyi narcissus according to claim 7, characterized in that, The system is connected to the tea roasting production line, and the near-infrared spectral acquisition module is an online diffuse reflection probe, which realizes real-time, non-destructive monitoring and feedback control of the roasting process.