Method for rapidly identifying lossless semen ziziphi spinosae decoction pieces based on electronic eye technology

Abstract

According to the lossless spina date seed decoction piece rapid identification method based on the electronic eye technology, the electronic eye technology is adopted to detect certified products, counterfeit products and adulterated products of lossless spina date seeds, and a spina date seed authenticity and adulteration identification model is established based on obtained visual information; comprising five steps of sample collection, data acquisition, model construction, model optimization and result identification. According to the method, principal component analysis-discriminant analysis, partial least squares-discriminant analysis, K nearest neighbor, a support vector machine and a back propagation neural network are respectively utilized to establish an authenticity and adulteration qualitative identification model, and the highest positive judgment rate of the shutter interactive verification model is taken as an index. The modern artificial intelligence sense-electronic eye technology and the machine learning algorithm are combined, the authenticity and adulteration of the spina date seeds can be well distinguished without processing the spina date seed decoction pieces, and the social and economic benefits are remarkable.

Description

Technical Field

[0001] The present invention relates to a method using electronic eye technology and a machine learning algorithm, which is used to quickly identify the authenticity and adulteration of lossless Chinese jujube seed decoction pieces. The invention relates to a method for quickly identifying lossless Chinese jujube seed decoction pieces based on electronic eye technology. Background Art

[0002] As a treasure of traditional Chinese medicine, Chinese medicine occupies an important position in the field of healthcare. In recent years, with the increasing attention paid to health and the advancement of the internationalization of traditional Chinese medicine, the scale of the Chinese medicine market has continued to expand. However, the problem of uneven quality of Chinese medicine has become increasingly prominent, seriously affecting the efficacy and safety of Chinese medicine and hindering the healthy development of the Chinese medicine industry. Among them, as one of the important dosage forms of Chinese medicine, the quality control of Chinese herbal medicine slices is particularly critical. Chinese jujube seed slices are widely used in clinical practice and have the effects of nourishing the heart and liver, calming the mind and soothing the nerves. They are often used to treat symptoms such as insomnia, palpitations and nightmares. However, due to the large market demand and high price of Chinese jujube seeds, there are many counterfeit and inferior products on the market, which have brought great harm to patients' health and market order. Therefore, there is an urgent need for an efficient and accurate method to identify the authenticity and adulteration of Chinese jujube seed slices.

[0003] At present, the traditional methods for identifying the authenticity and adulteration of Chinese herbal medicine slices of Ziziphus jujuba seeds mainly include property identification, microscopic identification, and physical and chemical identification. Property identification is highly subjective, requires extremely high experience from the identification personnel, and it is difficult to accurately identify counterfeit products that have undergone special treatment. Microscopic identification is complicated to operate, requiring professional microscope equipment and technicians, and it is difficult to accurately judge the microscopic characteristics of crushed slices or samples with a small amount of adulteration. Although physical and chemical identification can detect ingredients through chemical reactions or instrumental analysis, it has disadvantages such as long detection cycle, high cost, and destructiveness to samples, and cannot meet the actual needs of rapid and non-destructive identification. When faced with market sampling of a large number of Ziziphus jujuba seeds slices, it is difficult to accurately judge the quality of a large number of samples in a short period of time.

[0004] With the advancement of science and technology, modern techniques such as spectral analysis and chromatographic analysis have been applied to the identification of traditional Chinese medicines. Spectral analysis can verify the authenticity of samples by detecting their spectral characteristics, but it struggles to effectively identify complex adulteration situations. While chromatographic analysis can accurately separate and detect components, it also suffers from expensive equipment, complex operation, and sample loss. Furthermore, these techniques often require complex and time-consuming sample preparation steps, and lack intelligent data analysis tools, making it difficult to quickly and accurately extract meaningful information from large amounts of data.

[0005] Electronic eye technology, also known as machine vision technology, uses an optical imaging system to capture target images and then uses computers to process, analyze, and interpret the images, enabling identification, detection, and measurement of the target. In traditional Chinese medicine (TCM) quality evaluation, electronic eye technology, with its non-contact detection and rapid image data acquisition, can efficiently capture multi-dimensional visual information such as the surface texture, color, and shape of jujube seed slices. Compared to traditional manual observation, electronic eye technology is unaffected by subjective factors and can capture subtle differences with micron-level accuracy. Compared to modern techniques such as spectroscopy and chromatography, it requires no complex pre-processing and does not damage the sample, allowing for non-destructive batch testing. Combining machine learning algorithms with image data for feature extraction and model training has the potential to address the shortcomings of existing technologies, enabling rapid, non-destructive, and accurate authentication and adulteration identification, meeting the urgent need for efficient quality testing technologies in the TCM market. This provides a new path for rapid, accurate, and non-destructive authentication and adulteration of jujube seed slices, meeting the urgent need for efficient quality testing technologies in the TCM market. However, no relevant public reports have been published to date. Summary of the Invention

[0006] In view of the above situation, in order to overcome the defects of the existing technology, the purpose of the present invention is to provide a non-destructive rapid identification method of Chinese jujube seed slices based on electronic eye technology, which can quickly and accurately identify the quality of Chinese jujube seed slices, and effectively solve the problems of strong subjectivity and poor repeatability of manual identification and time-consuming and labor-intensive detection by modern precision instrument analysis methods.

[0007] To achieve the above object, the technical solution provided by the present invention is a non-destructive rapid identification method of jujube seed slices based on electronic eye technology, comprising the following steps:

[0008] (1) Sample collection: n batches of authentic Chinese medicine Suanzaoren and m batches of counterfeit Chinese medicine Suanzaoren were collected, and samples of one batch of authentic and counterfeit products were randomly mixed to obtain a number of adulterated samples with different adulteration ratios;

[0009] (2) Data collection: The electronic visual data of the Chinese herbal medicine slices were collected using an electronic eye (EE) instrument for authentic and counterfeit Chinese jujube seeds and adulterated samples with different adulteration ratios;

[0010] (3) Model construction: The electronic visual data is transmitted to the pattern recognition system to construct the authenticity model and the adulteration model respectively. The construction method is as follows:

[0011] ① After collecting electronic eye sensory data from authentic and counterfeit Suanzaoren seeds, as well as samples with different adulteration ratios, three data matrices were obtained for each type of sample: authentic matrix A: n×p; counterfeit matrix B: m×p; and adulterated matrix C: a×p, where p represents the number of sensors in the electronic eye sensory instrument.

[0012] ② Merge the above matrices AB and AC respectively to obtain the authenticity identification data matrix X D , adulteration identification data matrix X E ;

[0013] ③ Confirm X based on known information D 、X E The sample authenticity qualitative information of the matrix (Y D ), whether there is adulteration information (Y E );

[0014] ④ Using computer MATLAB software and classification toolbox, transfer data (X) to the pattern recognition system and build models of Y = f(X), including PCA-DA, PLS-DA, SVM, KNN and BPNN models;

[0015] ⑤ Optimize the model and use blinds interactive verification to finally obtain an identification model with good prediction performance, and finally select the optimal model based on the positive judgment rate;

[0016] ⑥ Identification results: Substitute the data X of the decoction piece to be tested into the model and give the quality identification result y of the decoction piece, thus realizing the rapid identification of non-destructive Chinese jujube seed decoction pieces.

[0017] The Chinese jujube seed slices collected in the step (1) have passed the content determination under the Chinese Pharmacopoeia 2025 Chinese Pharmacopoeia of Chinese jujube seed, and are confirmed to be authentic Chinese jujube seed.

[0018] The electronic eye in step (2) is composed of a standard light source, backlight, and an infrared camera, and is used to detect the appearance data of the shape and characteristic patterns of the medicinal slices, as well as the optical data of surface reflected light and diffuse reflected light.

[0019] The collection method in step (2) is as follows: accurately weigh 5g of medicinal slices, place them in a transparent culture dish and shake them gently to ensure that the medicinal slices do not overlap; the test procedure is as follows: turn on the IRISVA-400 electronic eye and its computer program, preheat the machine for 5-10 minutes, wait until the machine lighting is stable and the instrument indicates the green light is on, then place the 24-color color correction plate into the detection area for correction, and perform sample testing after correction. Each sample needs to collect images 3 times, set the image background elimination processing method, select the feature information of the central area of ​​the image, perform unified information extraction processing on the collected image, and finally take the average of the three image information to obtain the chromaticity value of the sample.

[0020] In the step (4), the authenticity identification data X D Match the authenticity information corresponding to the medicinal material (Y D ), whether there is adulteration identification data X E , match the corresponding adulteration information of the decoction piece (Y E) Using computer MATLAB software and classification toolbox, combined with PCA-DA, PLS-DA, KNN, SVM, BPNN, and machine learning algorithms, a qualitative identification model for authenticity and adulteration was established.

[0021] The method of the present invention is novel and unique, easy to operate, and establishes a non-destructive method for rapid identification of the authenticity and adulteration of Chinese jujube seed decoction pieces. It is fast, efficient, and accurate in testing. It can effectively solve the problem of identifying the authenticity and adulteration of Chinese medicine Chinese jujube seed decoction pieces, and has significant social and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a radar diagram of electronic eye visual information of the present invention (A is authentic Chinese jujube seed slices, B is counterfeit Chinese jujube seed slices, C is adulterated Chinese jujube seed slices, and D is adulterated Hovenia dulcis fruit slices).

[0023] Figure 2 These are the results diagrams of the PCA-DA, PLS-DA, and SVM models for authenticity identification of the present invention (a: PCA-DA model score diagram, b: PLS-DA model score diagram, c: SVM model classification result scatter diagram, class 1 is authentic and class 2 is counterfeit).

[0024] Figure 3 These are the PCA-DA, PLS-DA, and SVM model result diagrams for identifying whether or not the product is adulterated (a: PCA-DA model score diagram, b: PLS-DA model score diagram, c: SVM model classification result scatter diagram, class 1 is authentic and class 2 is adulterated). DETAILED DESCRIPTION

[0025] The specific implementation of the present invention is further described in detail below with reference to the accompanying drawings and examples.

[0026] The present invention is specifically implemented by the following examples.

[0027] Example 1

[0028] A non-destructive rapid identification method for Chinese date seed slices based on electronic eye technology.

[0029] (1) Sample collection

[0030] Sixty-six batches of Chinese jujube kernels were collected from various locations across China. Experts conducted traditional identification and assayed their contents according to the Chinese Pharmacopoeia, confirming that all samples qualified as Chinese jujube kernel slices. The counterfeit Chinese jujube kernels included 18 batches each of jujube kernels and Hovenia dulcis fruit. Furthermore, the adulterated Chinese jujube kernels were artificially adulterated at varying weight / weight ratios. Two adulterants were designed, each containing 50 adulteration ratios ranging from 1% to 50%, with an interval of 1%.

[0031] Table 1 Source information of authentic jujube seeds

[0032]

[0033] Table 2 Source information of counterfeit jujube seeds

[0034]

[0035]

[0036] (2) Data Collection

[0037] The test was performed using the IRIS VA400 electronic eye. 5g of each medicinal piece sample was precisely weighed, placed in a petri dish, gently smoothed, and then placed in the EE testing chamber for testing. After the EE was turned on and stabilized for a few minutes, and the lighting stabilized (after the instrument indicated a green light), the operating software was opened. Calibration was first performed using a 24-color color calibration plate, followed by testing. Each sample was captured three times, and the image information of the central area was selected for analysis. The average value was used to obtain the sample's chromaticity value.

[0038] (3) Construction and verification of authenticity and adulteration identification models

[0039] Using (X) as the independent variable, qualitative information about the authenticity and adulteration of jujube seeds was confirmed based on known information, and used as benchmark information (Y). A model (Y = f(X)) was established. The jujube seed authenticity and adulteration identification model included PCA-DA, PLS-DA, and SVM. The model was optimized based on the positive detection rate. Finally, the sample was substituted into the optimal model (Y = f(X)) to obtain the identification results.

[0040] 3.1 Construction of PCA-DA discriminant model

[0041] ① Load the fused data matrix (X) and the authenticity and adulteration information (Y) obtained from the Classification toolbox-version 6.0. Enter the "PCA-DA Optimal Principal Components" tab to optimize the number of principal components for this matrix. Parameter settings: select Automatic scaling for data scaling; Linear for recognition mode; Venetian blind cross-validation for validation method; and All for number of classification groups.

[0042] ② According to the "Number of principal components - Classification error rate" graph obtained in step ①, determine the number of principal components to be selected and enter them in the PCA-DA calculation parameter window. Set other parameters as follows: select automatic scaling for data scaling, linear for recognition mode, blind cross validation for validation method, and all for number of classification groups. Start the calculation and check which samples are correctly or incorrectly classified in the classification result window and calculate the positive rate.

[0043] ③ In the viewing window, call the Wilks' lambda graph, remove the variables with the largest Wilks' lambda values ​​one by one, and perform the calculation in step ② again to improve the model's positive detection rate by removing redundant variables;

[0044] ④ Repeat step ③ until the positive rate decreases after deleting the variable. The previous variable is selected as the final variable, and the parameter model is the optimal positive rate model.

[0045] 3.2 Construction of PLS-DA discriminant model

[0046] ① Load the fused data matrix (X) and the known authenticity and adulteration information (Y) obtained through the Classification toolbox-version 6.0. Enter the "PLS-DA Optimal Latent Variables" tab to optimize the number of latent variables for this matrix. Parameter settings: select Automatic scaling for data scaling; Linear for recognition mode; Venetian blind cross-validation for validation method; and All n for number of classification groups.

[0047] ② Based on the "Number of Latent Variables - Classification Error Rate" graph obtained in step ①, determine the number of latent variables to be selected and enter them in the PLS-DA calculation parameter window. Set other parameters as follows: select automatic scaling for data scaling, linear for recognition mode, blind cross-validation for validation method, and all for number of classification groups. Start the calculation and check which samples are correctly or incorrectly classified in the classification result window and calculate the positive rate.

[0048] ③ In the viewing window, call the Wilks' lambda graph, remove the variables with the largest Wilks' lambda values ​​one by one, and perform the calculation in step ② again to improve the model's positive detection rate by removing redundant variables;

[0049] ④ Repeat step ③ until the positive rate decreases after deleting the variable. The previous variable is selected as the final variable, and the parameter model is the optimal positive rate model.

[0050] 3.3 Construction of SVM discriminant model

[0051] The information matrix (X) after data fusion and the known authenticity and presence of adulteration information (Y) were loaded from the Classification toolbox-version 6.0. Before modeling, the data were preprocessed with "Standard Normal Variation (SNV)" row preprocessing and "autoscaling" column preprocessing method. During program execution, the model was optimized by selecting functions (linear kernel function, radial basis kernel function, polynomial kernel function) and the number of variables until the model with the highest positive judgment rate was determined.

[0052] 3.4 Construction of KNN discriminant model

[0053] Use the Classification toolbox-version 6.0 to load the fused data information matrix (X), known authenticity and adulteration information (Y), and enter the "Optimal Number of K-values" tab to optimize the number of K nearest neighbors. Parameter settings: select Euclidean distance for distance, automatic scaling for data scaling, interactive blinds for verification, and all for number of classification groups. Obtain the "K-values ​​Classification Error Rate" graph, which serves as the basis for the number of K nearest neighbors. Then, perform fitting and verification of the optimal K-value, and view the classification status in the classification window.

[0054] 3.5 Construction of BPNN discriminant model

[0055] The BPNN discriminant model requires optimization of the hidden layer, using the fused data matrix (X) and known authenticity and adulteration information (Y) loaded from the Classification toolbox-version 6.0. This toolbox designs 1-5 hidden layers. The number of hidden layers is adjusted sequentially, prioritizing the optimal number of hidden layers based on the accuracy of classification predictions. Parameters are optimized within the default range based on the highest positive rate. Venetian blind interactive validation is used for verification, and the maximum sample size is selected for the number of classification groups.

[0056] 3.6 Model Validation

[0057] The accuracy and discrimination efficiency of the established model were verified using the interactive verification method in MATLAB2022a software and classification_toolbox_6.0 toolbox.

[0058] 4 Model selection

[0059] In the model for identifying the authenticity of jujube seeds, the correct judgment rate was above 96.00%, among which the optimal model was PLS-DA with a correct judgment rate of 100.00%; in the model for identifying the presence of adulteration in jujube seeds, the correct judgment rate was above 98.00%, among which the optimal models were PCA-DA, PLS-DA, SVM, and KNN, with correct judgment rates of 100.00%.

[0060] Table 3. The accuracy of the identification model for the authenticity and adulteration of Chinese medicine Suanzaoren.

[0061]

[0062] Note: “*” indicates the highest correct rate

[0063] The present invention aims to combine electronic eye technology with machine learning algorithms to establish a non-destructive method for rapid identification of the authenticity and adulteration of Chinese jujube seed slices, so as to achieve efficient and accurate detection of the samples to be tested. As shown in Table 3, the correct judgment rate of the authenticity identification model is in the range of 96.08% to 100.00%, and the PLS-DA model has the highest correct judgment rate; the correct judgment rate of the adulteration identification model is in the range of 98.19% to 100.00%, and the PCA-DA, PLS-DA, SVM, and KNN models have the highest correct judgment rates. Their performance meets the actual needs of classification. The present invention can efficiently solve the problem of identifying the authenticity and adulteration of Chinese jujube seed slices, and at the same time provide new ideas for the authenticity and adulteration identification and application research of "new" Chinese medicines, "new" medicinal parts, rare drug substitutes, etc. It has strong practical application value and significant economic and social benefits.

[0064] It should be pointed out that the above is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any technician familiar with this profession can make changes or modify the technical content disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention, and all of them fall within the scope of protection of the present invention.

Claims

1. A non-destructive rapid identification method for Chinese date seed slices based on electronic eye technology, characterized in that: The following steps are involved: (1) Sample collection: Collect n batches of authentic Chinese medicine Suanzaoren and m batches of counterfeit Chinese medicine Suanzaoren. Take one batch of authentic and counterfeit samples and randomly mix them to obtain a number of adulterated samples with different adulteration ratios. (2) Data collection: The electronic visual data of the Chinese herbal medicine slices were collected using an electronic eye instrument for the authentic and counterfeit Suanzaoren seeds and adulterated samples with different adulteration ratios; (3) Model construction: The electronic visual data is transmitted to the pattern recognition system to construct the authenticity model and the adulteration model respectively. The construction method is as follows: ① After collecting electronic eye sensory data of authentic and counterfeit Suanzaoren seeds and adulterated samples with different adulteration ratios, three data matrices were obtained for each type of sample, namely: authentic matrix A: n×p; Counterfeit matrix B: m×p, adulterated matrix C: a×p, where p represents the number of sensors in the electronic eye sensory instrument; ② Merge the above matrices AB and AC respectively to obtain the authenticity identification data matrix X D , adulteration identification data matrix X E ; ③ Confirm X based on known information D 、X E The sample authenticity qualitative information of the matrix (Y D ), whether there is adulteration information (Y E ); ④Use computer MATLAB software and classification toolbox to classify the data ( X ) is transferred to the pattern recognition system to build Y = f ( X ) models, including PCA-DA, PLS-DA, SVM, KNN and BPNN models; ⑤ Optimize the model and use blinds interactive verification to finally obtain an identification model with good prediction performance, and finally select the optimal model based on the positive judgment rate; ⑥ Identification results: the data of the decoction to be tested X Substitute into the model and give the quality identification result of the decoction piece y， Achieve rapid identification of non-destructive Chinese jujube seed slices.

2. The non-destructive rapid identification method of spinach seed slices based on electronic eye technology according to claim 1 is characterized in that: The Chinese jujube seed slices collected in step (1) have passed the content determination under the Chinese Pharmacopoeia 2025 Chinese Pharmacopoeia, and are confirmed to be authentic Chinese jujube seed.

3. The non-destructive rapid identification method of spinach seed slices based on electronic eye technology according to claim 1 is characterized in that: The electronic eye in step (2) is composed of a standard light source, backlight, and infrared camera, and is used to detect the appearance data of the shape and characteristic patterns of the medicinal slices, as well as the optical data of surface reflected light and diffuse reflected light.

4. The non-destructive rapid identification method of Suanzaorensis seeds based on electronic eye technology according to claim 1, characterized in that: The collection method in step (2) is as follows: accurately weigh 5g of the medicinal slices, place them in a transparent culture dish and shake them gently to ensure that the medicinal slices do not overlap; the test procedure is as follows: turn on the IRISVA-400 electronic eye and its computer program, preheat the machine for 5-10 minutes, wait for the machine lighting to stabilize, and after the green light of the instrument indicator is on, place the 24-color color calibration plate into the detection area for calibration, and perform sample testing after correction. Each sample needs to collect images three times, set the image background elimination processing method, select the feature information of the central area of ​​the image, perform unified information extraction processing on the collected image, and finally take the average value of the three image information to obtain the chromaticity value of the sample.

5. The non-destructive rapid identification method of spinach seed slices based on electronic eye technology according to claim 1 is characterized in that: In the step (4), the authenticity identification data X D Match the authenticity information corresponding to the medicinal material (Y D ), whether there is adulteration identification data X E , match the corresponding adulteration information of the medicinal material (Y E ), using computer MATLAB software and classification toolbox, combined with PCA-DA, PLS-DA, KNN, SVM, BPNN, and machine learning algorithms to establish a qualitative identification model for authenticity and adulteration.

Images