CN-121995019-A - Method for discriminating honey-fried astragalus membranaceus processing degree based on sensory technology and component analysis

CN121995019ACN 121995019 ACN121995019 ACN 121995019ACN-121995019-A

Abstract

The invention discloses a method for judging the processing degree of honey-fried astragalus based on sensory technology and component analysis. The method comprises the steps of collecting multi-source characteristic data of samples with different processing degrees through an electronic tongue, an electronic eye and high performance liquid chromatography, classifying the samples into three stages of processing failure, moderate and excessive according to RGB color characteristics, extracting taste characteristics which are cooperatively changed with a processing process by adopting local weighted principal component analysis in each stage, establishing a prediction model of sensory characteristics and core active component content by utilizing partial least squares regression, and finally fusing taste information, prediction component information and standardized color information after dimension reduction, and realizing accurate judgment of the processing degree of the samples to be detected by calculating a Markov distance. The invention realizes objectification, digitalization and rapidness of the distinguishing process, effectively overcomes the defects of strong subjectivity and poor reproducibility of the traditional experience method, and provides a reliable technical means for distinguishing the processing quality of honey-fried astragalus.

Inventors

ZHONG LINGYUN
WANG YAN
LU XINGMEI
HUANG YI
Tong hengli
XIE RONGFANG
Ye Xietao

Assignees

江西中医药大学

Dates

Publication Date: 20260508
Application Date: 20260408

Claims (7)

1. A method for judging the processing degree of honey-fried astragalus based on sensory technology and component analysis is characterized by comprising the following steps: s1, multi-source characteristic data acquisition; Preparation Pulverizing radix astragali Preparata with honey to obtain Standard sample of radix astragali processed with honey is prepared by respectively performing HPLC on electronic tongue, electronic eye and high performance liquid chromatography Analyzing taste, surface appearance and core active ingredient content of standard sample of honey-processed radix astragali to obtain honey-processed radix astragali with different processing degree Data on dimensional taste response values Color values of RGB three channels and content data of core active components; Step S2, sample pre-segmentation based on multi-source characteristic data; Extracting the average intensity value of each group of honey-fried astragalus mongholicus standard samples in R, G, B channels according to the RGB three-channel color values, and dividing honey-fried astragalus mongholicus with different processing degrees into three stages of processing failure, moderate processing and processing overdose; step S3, construction stage Is used for enhancing the fusion feature matrix; step S31, phase-specific sensory feature extraction based on local weighted principal component analysis PCA; Step S32, constructing a phase component prediction model based on partial least squares regression PLS; step S33, construction stage Is used for calculating the enhanced fusion feature matrix to a stage Mahalanobis distance of the category center of (c); s4, judging the processing degree of the sample to be tested; and calculating the mahalanobis distance from the reinforced fusion feature matrix of the sample to be detected to the class center of each stage, and judging the processing degree corresponding to the stage with the minimum mahalanobis distance of the sample to be detected.
2. The method for determining the processing degree of honey-fried astragalus membranaceus based on sensory technology and component analysis as claimed in claim 1, wherein in the step S1 Data on dimensional taste response values , , The total number of taste characteristics of the compounds corresponding to all the sensors of the electronic tongue; The vitamin taste characteristic at least comprises sour taste, bitter taste, astringency, bitter aftertaste, astringent aftertaste, fresh taste, richness and salty taste, and the content data of the core active ingredient comprises astragaloside IV Calycosin glucoside And formononetin 。
3. The method for distinguishing the processing degree of honey-fried astragalus based on the sensory technology and the component analysis according to claim 1, wherein the sample pre-segmentation based on the multi-source characteristic data in the step S2 comprises the following steps: according to RGB three-channel color values, extracting the average intensity value of each group of honey-fried astragalus root standard samples in R, G, B channels to form the color feature vector of the standard samples ; According to the empirical rule that the color of honey-fried astragalus is changed from light yellow to dark yellow and then to yellow brown in the processing process of honey-fried astragalus, the color feature vectors of n groups of honey-fried astragalus standard samples are combined The honey-fried astragalus root with different processing degrees is divided into three stages: Stage I: corresponding to pale yellow, the processing degree is not as good as that of the processing; Stage II: the processing degree is moderate corresponding to bright yellow; Stage III: The processing degree is proper, and the corresponding dark yellow or yellow brown color is corresponding; Let the number of samples at each stage be 、、 And (2) and 。
4. A method for discriminating honey-fried astragalus membranaceus processing degree based on sensory technology and component analysis as claimed in claim 3, wherein the stage-specific sensory feature extraction based on the local weighted principal component analysis in step S31 is performed as follows: Step S311, for the stage Assume that this phase comprises Duplicate samples, build stage Is of the original feature matrix of (a) ; ; In the above-mentioned method, the step of, ; Is a stage of Inner first Samples of The data of the response value of the sense of taste is maintained, ; Step S312, calculation stage Inner first Original features And the first stage Average RGB intensity value of the copy samples Taking its absolute value as the phase correlation weight : , ; In the above-mentioned method, the step of, Is a stage of Inner first Sample number of An electronic tongue taste response value; Is that Part of Standard sample No. 1 The average value of the taste response values of the individual electronic tongues, ; Is a stage of Inner first The average RGB intensity value of the copy samples, , Respectively as stages Inner first Intensity values of R, G, B channels of the aliquot; Is a stage of Inner part Average of average RGB intensity values for the standard samples, ; Step S313, for the original feature matrix Weighting to obtain weighted feature matrix After normalization processing, a normalized feature matrix is obtained For standardized feature matrix Performing Principal Component Analysis (PCA); First, a normalized feature matrix is calculated Covariance matrix of (2) , Is that Is a transpose of (2); Then to covariance matrix Performing eigenvalue decomposition and solving eigenvalue problems: ; In the above-mentioned method, the step of, Is a characteristic value; Is a feature vector; obtaining a characteristic value by solving the characteristic value problem And corresponding feature vector ; The main components are ordered according to the magnitude of the characteristic values before selection The accumulated variance contribution rate is more than or equal to 85 percent by the main components; Acquisition stage A sensory principal component scoring matrix of (a): ; In the above-mentioned method, the step of, Is a stage of Front part A matrix of the individual feature vectors, 。
5. The method for distinguishing the processing degree of honey-fried astragalus membranaceus based on the sensory technology and the component analysis according to claim 4, wherein the step S32 is constructed based on a partial least squares regression phase component prediction model, and comprises the following steps: step S321, collecting stage Inner part The content data of core active ingredients of the sample, namely acteoside AC, astragaloside CG and acteoside FO are established Component content matrix of (2) : ; In the above-mentioned method, the step of, 、、 Respectively represent stages Inner first The content of calycosin AC, astragaloside CG and calycosin FO in the sample, ; Because the contents of the three core active ingredients have different dimensions and magnitude orders, the three core active ingredients need to be aligned Performing standardization processing to obtain a standardized component content matrix ; ; In the above-mentioned method, the step of, Is that Average value vectors of all columns; Is that Standard deviation vectors for each column; For all elements 1 A dimension column vector; representing a transpose of the matrix; step S322, establishing a partial least squares regression PLS model to As the amount of the independent variable X, Extracting latent variables by iteration as dependent variables Y Until reaching the preset stopping condition; Order the , For the first The number of latent variables is a function of, , For the best latent variable number, calculating weight vector : ; In the above-mentioned method, the step of, Is an independent variable residual error matrix; Is that Is a transpose of (2); For the Y-score vector, initially fetch Is formed by the first column of (1) in an iterative process Updating to be converged is carried out, As a matrix of dependent variable residuals, Is a Y load vector; ; Calculating an X score vector The score vector is the projection of the argument residual in the direction of weight, expressed as: ; obtaining an X load vector by least square estimation And Y load vector ; ; ; Removing the extracted latent variable information from the current independent variable residual matrix and the dependent variable residual matrix to obtain the first The residual matrices after the multiple iterations are respectively expressed as: , ; Determination of optimal latent variable number by cross-validation Latent variable number for each candidate Calculating the sum of squares of prediction residues : ; In the above-mentioned method, the step of, Is the first Predicted values for the individual samples; To exclude the first Component prediction model pair established after each sample Predicted values for the individual samples; Further calculating the prediction root mean square error according to the calculated prediction residual error square sum ; ; Selecting and making With the smallest value As the optimal latent variable number Or when Time selection As the optimal latent variable number ; Extraction of Obtaining final regression coefficient matrix after each latent variable : ; In the above-mentioned method, the step of, As a matrix of weights, the weight matrix, ; In the case of an X-load matrix, ; In the case of a Y-load matrix, ; Acquisition stage The component prediction model of (2) is as follows 。
6. The method for determining the processing degree of honey-fried astragalus membranaceus based on sensory technology and component analysis as claimed in claim 5, wherein the steps of step S33 are as follows Is a sensory principal component scoring matrix Sum component prediction model Splicing according to columns to obtain a fusion feature matrix Enhancing fusion feature matrices to stages The mahalanobis distance calculation process for the category center of (c) is as follows: Calculation stage Mean and standard deviation of each of the three channels in R, G, B, for the stage Color feature vector of each honey-roasted astragalus root standard sample Normalizing to obtain stage Internal normalized RGB matrix : ; In the above-mentioned method, the step of, 、、 Respectively as stages Inner first Intensity values of R, G, B channels of the individual samples; 、、 And 、、 Respectively as stages Mean and standard deviation of each of the three channels in the interior R, G, B; Will be Adding the new column into the fusion feature matrix to obtain an enhanced fusion feature matrix ; Calculation stage Category center of (f) : ; In the above-mentioned method, the step of, Is a stage of The number of all samples in a sample; To enhance the fusion feature matrix Is the first of (2) The number of row vectors, ; Calculation stage Is a global combining covariance matrix of (1) : ; Computing an enhanced fusion feature matrix To stage Category center of (f) Mahalanobis distance of (v) : ; In the above-mentioned method, the step of, For the fusion feature vector of the sample, , At the stage of the sample Is characterized by a sensory principal component molecular vector, The predicted content sub-vectors of the three core active ingredients of the sample, The sample normalized RGB color subvectors.
7. The method for determining the processing degree of honey-fried astragalus membranaceus based on the sensory technology and the component analysis according to claim 6, wherein the determining process of the processing degree of the sample to be detected in the step S4 is as follows: Step S41, collecting a honey-fried astragalus sample to be detected The data of the taste response value and the color values of RGB three channels are maintained; step S42, determining the stage of the honey-fried astragalus sample to be detected according to the obtained RGB three-channel color value ; S43, calculating the sensory principal component score and the predicted component content of the honey-fried astragalus sample to be detected, and splicing with the normalized RGB color characteristics to obtain the enhanced fusion feature vector of the honey-fried astragalus sample to be detected ; Step S44, calculating enhanced fusion feature vectors of the honey-fried astragalus samples to be detected Class center to stages Mahalanobis distance of (v) ; And step S45, judging that the honey-fried astragalus sample to be detected belongs to the processing degree corresponding to the stage with the minimum Mahalanobis distance.

Description

Method for discriminating honey-fried astragalus membranaceus processing degree based on sensory technology and component analysis Technical Field The invention belongs to the technical field of traditional Chinese medicine processing degree discrimination, and particularly relates to a discrimination method for honey-roasted astragalus membranaceus processing degree based on a sensory technology and component analysis. Background Astragalus membranaceus (ASTRAGALI RADIX) is used as a common traditional Chinese medicine in clinic and has the effects of tonifying qi and yang, strengthening exterior and arresting sweating, inducing diuresis and detumescence, promoting fluid production and nourishing blood and the like. The honey-processing method is one of the main processing methods, and can enhance the qi-tonifying, lung-moistening, spleen-strengthening and middle-jiao-regulating effects of astragalus through the sweet and slow middle-jiao-tonifying property of honey. Modern researches prove that the honey-moxibustion process can influence the content and conversion of core active ingredients such as astragaloside IV, calycosin glucoside, formononetin and the like in astragalus, thereby directly influencing the efficacy and clinical safety of the astragalus. Therefore, the accurate and objective judgment of the processing degree (such as the fire) of the honey-fried astragalus is a key link for guaranteeing the uniform quality of decoction pieces and stable and controllable curative effect. Currently, the judgment of the honey-fried astragalus root processing end point in the industry mainly depends on the traditional sensory experience of processing staff, namely, the comprehensive judgment is realized by observing the color change (such as light yellow to bright yellow, dark yellow to even yellow brown) of the surface of decoction pieces, smelling the burnt aroma of honey and tasting the sweet and bitter balance degree of the honey-fried astragalus root. The empirical method of 'seeing, smelling and tasting' has the advantages of high subjectivity and poor reproducibility, is highly dependent on the accumulation of individual experience and sensory state, is easy to generate judgment difference among different operators, and is difficult to form uniform and quantized quality standard although precious practical wisdom is included. In recent years, some modern analysis techniques are tried to be used for evaluating the quality of traditional Chinese medicines, for example, high Performance Liquid Chromatography (HPLC) can accurately measure the content of active ingredients, but the method is time-consuming and high in cost, belongs to destructive detection, and cannot realize rapid and nondestructive discrimination of decoction pieces on a production line, the electronic tongue technique is used as an artificial taste analysis system, can convert the whole taste information (such as sour, sweet, bitter, salty, fresh and the like) of a sample into objective and quantifiable sensor response signals, and shows advantages in evaluating the flavor of foods and medicines, and the electronic eye (or high-resolution image analysis technique) can objectively and accurately quantify the color characteristics of the sample and avoid subjective deviation of human eye discrimination. However, these methods are mostly used in isolation or simply in data tabulation comparison, and fail to deeply mine the deep correlation between "sensory attributes (color, taste)" and "intrinsic substance basis (chemical components)", and in particular, the following drawbacks remain in the prior art: 1. the identification is based on singleness, namely the identification is only based on color or only based on a few components, the process that the traditional Chinese medicine preparation is the overall change of 'nature, taste and effect' is ignored, and a comprehensive evaluation system for multi-dimensional information fusion is lacked. 2. The model universality is poor, the processing is a continuous dynamic process, the physicochemical property change rules of samples in different processing stages (such as early stage, medium stage and later stage) are different, and the conventional method for processing the data in the whole processing process by adopting a single global mathematical model (such as a single regression or classification model) is difficult to accurately describe the nonlinear and stepwise change characteristics, so that the discrimination precision is limited. 3. The judgment logic of the existing analysis method is difficult to understand and trust by the technicians in the traditional Chinese medicine processing field, and the traditional experience essence of 'color judgment degree and taste differentiation' can not be effectively butted and inherited. Disclosure of Invention The invention aims to overcome the defects of the prior art, provides a method for judging the processing degree of honey-fried astraga