Search

CN-121994747-A - Intelligent discrimination method and system for color quality of baked fire of Narcissus wuyi

CN121994747ACN 121994747 ACN121994747 ACN 121994747ACN-121994747-A

Abstract

The invention discloses an intelligent discrimination method and system for the quality of a baked fire color of Narcissus martensii, which are based on a genetic algorithm to optimize a near infrared spectrum. The method comprises the steps of collecting near infrared spectrum of a Wuyi Narcissus crude tea sample in a ‑1 wave band of 400-4000cm, combining CIE-Lab chromaticity parameters, adopting Savitzky-Golay smooth filtering and multiple scattering correction combined pretreatment to eliminate noise interference, utilizing a genetic algorithm to screen key characteristic wavelengths to construct a wavelength-chromaticity correlation matrix, and coupling partial least squares regression (PLS) with a Support Vector Machine (SVM) to construct a GA-PLS-SVM mixed prediction model, so that accurate inversion of baking intensity and time parameters is realized. The model calibration set L/a/b predicts that R2 is more than or equal to 0.90, verifies that R2 is more than or equal to 0.85, can identify 82.3% baking process deviation online, and remarkably improves rock tea standardized production efficiency and process control precision.

Inventors

  • SUN HUI
  • PANG JIE
  • WANG YUXIANG
  • ZHAO CANQIANG

Assignees

  • 武夷学院
  • 赵灿强

Dates

Publication Date
20260508
Application Date
20260127

Claims (8)

  1. 1. The intelligent distinguishing method for the quality of the baked fire color of the Narcissus martensii based on the genetic algorithm optimization near infrared spectrum is characterized by comprising the following steps of: (1) Collecting near infrared spectrum data of the Narcissus martensii raw tea sample in the wave band of 400-4000 cm-1; (2) Measuring CIE-Lab chromaticity space parameters corresponding to the Wuyi narcissus raw tea sample as quality reference values; (3) Carrying out Savitzky-Golay smoothing filtering and multiple scattering correction combined pretreatment on the near infrared spectrum data; (4) Screening key characteristic wavelengths obviously related to the CIE-Lab chromaticity parameters from the preprocessed spectrum data by utilizing a genetic algorithm, and constructing a wavelength-chromaticity correlation matrix; (5) Based on the key characteristic wavelength data and the corresponding CIE-Lab chromaticity parameters, coupling partial least squares regression and a support vector machine to construct a GA-PLS-SVM hybrid prediction model; (6) And (3) intelligently judging the baking color quality of the unknown Narcissus wuyis sample by using the GA-PLS-SVM mixed prediction model, and inverting the baking intensity and time parameters of the unknown Narcissus wuyis sample. .
  2. 2. The intelligent discrimination method for quality of baked fire color of Narcissus martensii as claimed in claim 1, wherein in step 4, parameters of the genetic algorithm are set to be that population scale is 50-100, crossover probability is 0.6-0.8, mutation probability is 0.01-0.05, and evolution algebra is 100-200 generations.
  3. 3. The intelligent discrimination method for the quality of the baked fire color of the Narcissus martensii as claimed in claim 1 or 2, wherein in the step 5), the construction of the GA-PLS-SVM hybrid prediction model is specifically that key characteristic wavelength data is firstly subjected to dimension reduction by using partial least squares regression, a principal component score is extracted as a new characteristic variable, and the new characteristic variable is input into a support vector machine for nonlinear modeling.
  4. 4. The intelligent discrimination method for the quality of the baked fire color of the Narcissus martensii of claim 3, wherein the support vector machine adopts a radial basis function as a kernel function, and the penalty factor C and the kernel function parameter gamma of the support vector machine are optimized through a grid search method.
  5. 5. The intelligent discrimination method for quality of baked fire color of Narcissus martensii of claim 1, wherein in step 6, the intelligent discrimination criteria is that the model is valid when the predictive decision coefficient R2 of the model for the values of sample L, a and b is not lower than 0.90 in the calibration set and not lower than 0.85 in the independent verification set.
  6. 6. The intelligent discrimination method for quality of baked fire color of Narcissus martensii as claimed in claim 1 or 5, further comprising step 7 of deploying the GA-PLS-SVM hybrid prediction model in an online monitoring system, collecting near infrared spectrum of tea sample of production line in real time, and when chromaticity value predicted by the model deviates from a preset process range, automatically triggering deviation alarm by the system, wherein the alarm sensitivity is not lower than 82.3%.
  7. 7. An intelligent discrimination system for the quality of the baked fire color of the daylily narcissus in the arm of any one of claims 1-6, comprising: (1) The near infrared spectrum acquisition module is used for acquiring spectrum data of the tea sample in a 400-4000 cm-1 wave band; (2) The spectrum preprocessing module is used for executing Savitzky-Golay smoothing filtering and multiple scattering correction algorithm; (3) The characteristic wavelength optimization module is internally provided with a genetic algorithm program and is used for screening key characteristic wavelengths; (4) The intelligent judging core module is internally provided with a trained GA-PLS-SVM hybrid prediction model and is used for outputting a baking color quality judging result and a baking process parameter inversion value according to input spectral data; (5) The result output and alarm module is used for displaying the discrimination result and giving an alarm when the result exceeds a threshold value.
  8. 8. The intelligent discrimination system for the quality of the baked fire color of the Narcissus martensii of claim 7, wherein the system is connected with a tea baking production line, and the near infrared spectrum acquisition module is an on-line diffuse reflection probe, so that real-time, nondestructive monitoring and feedback control of the baking process are realized.

Description

Intelligent discrimination method and system for color quality of baked fire of Narcissus wuyi Technical Field The invention relates to the technical field of tea processing, the technical field of spectral analysis and the technical field of artificial intelligence, in particular to an intelligent distinguishing method and system for the quality of a baked fire color of Narcissus martensii on the basis of a genetic algorithm optimized near infrared spectrum. Background The Narcissus martensii is a representative variety of the North Fujian oolong tea, and is known as the peculiar rock charm and mellow taste. Baking fire is a core process for shaping the quality style of Narcissus martensii, and promotes the conversion of substances in tea and the formation of aroma through the action of heat. Wherein, the color is one of the most visual and important sensory indexes for evaluating the baking quality, and is highly related to the taste, aroma and biochemical component content of the tea. The accurate control of the baking degree realizes the standardization and stabilization of the color quality, and is a key for improving the market competitiveness of Narcissus wuyi. Currently, the judgment of the baking degree of the Narcissus marcescens is mainly based on a sensory evaluation method and a physicochemical analysis method. Sensory evaluation methods (see GB/T23776-2018, for example) rely on experience and subjective judgment of tea evaluation operators, are easily affected by individual differences, environments and fatigue degrees, lack objective quantification standards, and are difficult to realize accurate control and quality tracing of mass production. The analytical method of the instrument, such as High Performance Liquid Chromatography (HPLC) for measuring characteristic pigment or catechin components, has the advantages of accurate result, complex sample pretreatment, time and labor consumption, high cost, destructive detection and incapability of meeting the requirements of quick, nondestructive and real-time monitoring on a production line. Near infrared spectrum (FIRS) technology is widely used as a rapid and nondestructive analysis means in the quality detection of agricultural products. However, the direct application of the method to the light color discrimination of the Narcissus wuyi baked fire presents significant challenges, namely firstly, the near infrared spectrum information of the tea is complex and contains a large amount of noise and background interference (such as moisture and granularity scattering) which are irrelevant to colors, secondly, the traditional full-band modeling is easy to cause the over fitting and the prediction capability of the model to be reduced, and furthermore, the color (characterized by CIE-Lab chromaticity space) and the spectrum are not simple linear relations, so that a single linear model is difficult to accurately capture the complex nonlinear mapping of the model. Therefore, there is an urgent need to develop a new method and system for quickly, accurately, nondestructively and intelligently judging the quality of the baked fire color of the Narcissus wuyi, so as to promote the digitization and intelligent upgrading of the process of the Narcissus wuyi tea. Disclosure of Invention The invention aims to overcome the defects of the prior art, and provides an intelligent distinguishing method and system for the quality of the baking fire color of the Narcissus wuyi based on a genetic algorithm optimized near infrared spectrum, so as to realize accurate, rapid and nondestructive distinguishing of the baking fire color of tea leaves, and provide technical support for optimizing and controlling the baking process in real time. In order to achieve the purpose, the invention adopts the following technical scheme that in the first aspect, the invention provides an intelligent distinguishing method for the quality of the baking fire color of the Narcissus martensii on the basis of a genetic algorithm to optimize a near infrared spectrum, which comprises the following steps: (1) Collecting near infrared spectrum data of the Wuyi Narcissus tea sample in 400-4000 cm-1 wave band, and simultaneously measuring CIE-Lab chromaticity space parameters corresponding to the tea sample as quality reference values; (2) Spectral preprocessing, namely carrying out joint preprocessing of Savitzky-Golay smoothing filtering and multiple scattering correction on the near infrared spectrum data; (3) The characteristic wavelength optimization, namely screening key characteristic wavelengths obviously related to the CIE-Lab chromaticity parameters from the pretreated spectrum data by utilizing a genetic algorithm; 4, constructing a mixed prediction model, namely constructing a GA-PLS-SVM mixed prediction model based on the key characteristic wavelength data and the CIE-Lab chromaticity parameters corresponding to the key characteristic wavelength data and coupling partial least squ