CN-122024898-A - Honey refining process characterization method and system based on near infrared spectrum and color characteristics

CN122024898ACN 122024898 ACN122024898 ACN 122024898ACN-122024898-A

Abstract

The invention belongs to the technical field of detection and analysis of a honey refining process of traditional Chinese medicines. A method for representing honey refining process based on near infrared spectrum and color features includes such steps as obtaining near infrared spectrum data and corresponding color image of sample in flow cell by external circulation pipeline, extracting low-dimensional color features from image, coupling and fusing with preprocessed spectrum data to construct state representing vector of honey refining process, creating partial least square regression model as main model, introducing extreme gradient lifting model as nonlinear supplement to obtain comprehensive representing result, analyzing variable contribution of dual model by hierarchical interpretation frame, and verifying the consistency of model prediction mechanism and physical-chemical reaction path in honey refining process. The method integrates spectrum and visual information, combines linear and nonlinear modeling advantages, ensures reliable model mechanism through interpretability analysis, and provides a new way for accurate monitoring of the honey refining process.

Inventors

HUANG PANLING
LIU WENJIE
LIU DAOLONG
LI MINGMING

Assignees

山东大学

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (10)

1. The honey refining process characterization method based on the near infrared spectrum and the color characteristics is characterized by comprising the following steps of: acquiring near infrared spectrum data of a sample in a flow cell and a color image corresponding to the sample, wherein the sample in the flow cell is a sample in a honey refining process input through an external circulation pipeline; extracting low-dimensional color features from the color image, and coupling and fusing the preprocessed near infrared spectrum data with the low-dimensional color features to construct a state representation vector of the honey refining process; based on the state characterization vector of the honey refining process, a partial least square regression model is established as a main characterization model, and an extreme gradient lifting model is introduced as a nonlinear supplementary analysis model so as to obtain the comprehensive characterization result of the honey refining process; And according to the comprehensive characterization result of the honey refining process, carrying out variable contribution analysis on the main characterization model and the nonlinear supplementary analysis model by using a layered interpretation framework, and judging the consistency of a model prediction mechanism and a physical and chemical reaction path in the honey refining process.
2. The method for characterizing a honey-refining process based on near infrared spectrum and color features as claimed in claim 1, Performing variable contribution analysis on the main characterization model by using a hierarchical interpretation framework, wherein the variable contribution analysis comprises the following steps: calculating variable projection importance values of variables in the main characterization model, and identifying key wavelength variables and key color variables which have great influence on the prediction of key quality attributes in the whole sample range; The calculation process of the variable projection importance value is that the sum of squares of response variable interpretation of each latent variable component is counted, the weights of each variable in each component are combined, and the weighted contribution sum of each variable to all the latent variable components is obtained through accumulation, so that the importance of each variable in the construction of a unified latent variable space is quantified.
3. The method for characterizing a honey-refining process based on near infrared spectrum and color features as claimed in claim 1, Performing variable contribution analysis on the nonlinear supplemental analysis model by using a hierarchical interpretation framework, including: Quantifying the contribution value of each feature in the nonlinear supplemental analysis model to a single prediction result by adopting a SHAP method, and generating a SHAP summary graph; analyzing the contribution direction and the contribution strength of the key wavelength variable and the key color variable to the key quality attribute prediction result according to the generated SHAP summary graph; Calculating marginal contribution of the model output when each feature exists or not by taking expected output of the model as a reference, and obtaining independent contribution of each feature to a final predicted value by arranging and combining average marginal contribution of all feature subsets.
4. The method for characterizing a honey-refining process based on near infrared spectrum and color features as claimed in claim 1, Key quality attributes include glucose content, fructose content, moisture content, pH, and 5-hydroxymethylfurfural content; the partial least squares regression model is used for representing the linear variation trend of the glucose content, the fructose content, the moisture content, the pH value and the 5-hydroxymethylfurfural content; The extreme gradient lifting model is used for detecting whether deviations caused by nonlinear coupling of internal composition state variables and external thermal reaction state variables exist in glucose content, fructose content, moisture content, pH value and 5-hydroxymethylfurfural content.
5. The method for characterizing a honey-refining process based on near infrared spectrum and color features as claimed in claim 4, Judging the consistency of a model prediction mechanism and a physical-chemical reaction path in the honey refining process, comprising the following steps: determining a key variable set predicted by a driving model based on the variable projection importance value and the SHAP summary map; analyzing whether the contribution direction of the key color variable in the key variable set accords with the deepening trend of the browning degree caused by the sugar dehydration reaction and the Maillard reaction in the honey refining process; If the characteristic representing the decrease in brightness in the key color variable has a positive contribution to the prediction of the 5-hydroxymethylfurfural content and the characteristic representing the increase in redness has a positive contribution to the prediction of the 5-hydroxymethylfurfural content, the prediction mechanism of the judgment model is consistent with the physicochemical reaction path.
6. The method for characterizing a honey-refining process based on near infrared spectrum and color features as claimed in claim 1, The hierarchical interpretation framework further includes a single sample decision path analysis, the single sample decision path analysis including: Selecting a representative sample, and utilizing a single sample to try to show how a predicted result of the representative sample is formed by pushing together a key wavelength variable and a key color variable; and judging whether the low-dimensional color features are in the absence of information of the supplementary near infrared spectrum data or exist as redundant variables by comparing variable contribution distribution of the single spectrum model and the multi-mode fusion model on the representative sample.
7. The method for characterizing a honey-refining process based on near infrared spectrum and color features as claimed in claim 1, The output of the comprehensive characterization result comprises the following steps: quantitative predicted values of key quality attributes output based on the main characterization model and the nonlinear supplemental analysis model; Based on a key variable contribution analysis report output by the hierarchical interpretation framework, the report comprises global importance ordering and local contribution directions of key wavelength variables and key color variables to key quality attributes; and the process state reliability evaluation generated based on the consistency judgment result is used for guiding state judgment, process trend analysis and quality control assistance of the honey refining process.
8. A honey refining process characterization system based on near infrared spectrum and color characteristics, comprising: The spectrum image acquisition unit is configured to acquire near infrared spectrum data of a sample in the flow cell and a color image corresponding to the sample, wherein the sample in the flow cell is a sample in a honey refining process input through an external circulation pipeline; The feature fusion construction unit is configured to extract low-dimensional color features from the color image, and perform coupling fusion on the preprocessed near infrared spectrum data and the low-dimensional color features to construct a honey refining process state representation vector; The dual-mode modeling and characterization unit is configured to establish a partial least square regression model as a main characterization model based on a state characterization vector of the honey refining process, and introduce an extreme gradient lifting model as a nonlinear supplementary analysis model to obtain a comprehensive characterization result of the honey refining process; And the mechanism consistency criterion unit is configured to utilize a layered interpretation framework to conduct variable contribution analysis on the main characterization model and the nonlinear supplementary analysis model according to the comprehensive characterization result of the honey refining process and judge the consistency of a model prediction mechanism and a physical and chemical reaction path in the honey refining process.
9. A honey refining process characterization system based on near infrared spectrum and color characteristics, comprising: A honey refining device; the external circulation pipeline is connected with the honey refining device and is used for conveying samples in the honey refining process; the flow cell is arranged on the outer circulation pipeline and is used for accommodating a sample to be tested; And a processor configured to perform the process of the honey process characterization method based on near infrared spectroscopy and color characteristics of any one of claims 1-7.
10. A computer device comprises a processor and a computer-readable storage medium; a processor adapted to execute a computer program; A computer readable storage medium having stored therein a computer program which, when executed by the processor, implements the honey refining process characterization method based on near infrared spectrum and color features as claimed in any one of claims 1 to 7.

Description

Honey refining process characterization method and system based on near infrared spectrum and color characteristics Technical Field The invention relates to the technical field of detection and analysis of traditional Chinese medicine honey refining processes, in particular to a honey refining process characterization method and system based on near infrared spectrum and color characteristics. Background The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. Honey refining is a key process in the processing of traditional Chinese medicines, and the quality of the honey refining directly determines the curative effect of the patent medicine. Traditional monitoring relies on manual experience to observe macroscopic indexes such as color, viscosity and the like, and is strong in subjectivity and lagging. In recent years, near infrared spectrum technology is introduced into process analysis due to its rapid and nondestructive characteristics, and can reflect internal chemical composition changes such as sugar, moisture and the like. Meanwhile, the machine vision technology can quantitatively extract the color characteristics of materials and capture the apparent color evolution caused by Maillard reaction and the like. The spectrum and the image data are combined, so that a more comprehensive digital representation of the honey refining process can be constructed theoretically, and a data base is provided for intelligent control. However, when the spectrum and the color features are fused, the prior art scheme only focuses on the improvement of prediction precision, and lacks the interpretation analysis of a model decision mechanism. The constructed comprehensive characterization model becomes a black box, and the prediction of how the spectral wavelength or color variable specifically drives the quality attribute cannot be clarified, so that whether the prediction logic of the model accords with the real physicochemical reaction path in the honey refining process is more difficult to verify. The opacity of the mechanism makes it impossible for technicians to confirm whether the model truly captures key process features, so that reliable process control and quality tracing based on model results are difficult. Disclosure of Invention In order to solve the defects of the prior art, the invention provides a honey refining process characterization method and system based on near infrared spectrum and color characteristics, which are used for solving the problems of incomplete honey refining process characterization, insufficient correspondence between a detection result and a process state, excessively high visual mode dimension and lack of mechanism explanation of a model in the prior art. In order to achieve the above purpose, the present invention adopts the following technical scheme: in a first aspect, the invention provides a method for characterizing a honey refining process based on near infrared spectrum and color characteristics. A honey refining process characterization method based on near infrared spectrum and color characteristics comprises the following steps: acquiring near infrared spectrum data of a sample in a flow cell and a color image corresponding to the sample, wherein the sample in the flow cell is a sample in a honey refining process input through an external circulation pipeline; extracting low-dimensional color features from the color image, and coupling and fusing the preprocessed near infrared spectrum data with the low-dimensional color features to construct a state representation vector of the honey refining process; based on the state characterization vector of the honey refining process, a partial least square regression model is established as a main characterization model, and an extreme gradient lifting model is introduced as a nonlinear supplementary analysis model so as to obtain the comprehensive characterization result of the honey refining process; And according to the comprehensive characterization result of the honey refining process, carrying out variable contribution analysis on the main characterization model and the nonlinear supplementary analysis model by using a layered interpretation framework, and judging the consistency of a model prediction mechanism and a physical and chemical reaction path in the honey refining process. In one implementation of the first aspect of the present invention, extracting low-dimensional color features from a color image includes: performing center clipping treatment on the color image, and converting the image into CIE Lab color space; calculating CIE Lab color space 、AndAnd a low dimensional color characteristic is obtained. In one implementation manner of the first aspect of the present invention, constructing a state characterization vector of a honey refining process includes: And (3) carrying out standardization processing on an inter