Model for detecting tanshinone IIA content and / or cryptotanshinone content in salvia miltiorrhiza, and construction method and application thereof
By constructing a Salvia miltiorrhiza image analysis model and utilizing machine learning methods and variable screening technology, the subjectivity and high cost issues of Salvia miltiorrhiza quality assessment were resolved, rapid and non-destructive detection of tanshinone IIA and cryptotanshinone contents was achieved, and reliable data-based evaluation of Salvia miltiorrhiza quality was realized.
Patent Information
- Application Number
- CN202510805722.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-17
- Publication Date
- 2025-09-19
AI Technical Summary
Existing technologies have the problems of strong subjectivity and difficulty in quantification in the quality assessment of Salvia miltiorrhiza, and modern analytical techniques such as high-performance liquid chromatography are expensive and difficult to popularize.
A model for detecting the content of tanshinone IIA and/or cryptotanshinone in Salvia miltiorrhiza was constructed. By obtaining the RGB and HSB values of the Salvia miltiorrhiza sample images, machine learning models such as PLS regression or RIDGE regression were used for prediction, combined with variable screening methods such as VCPA and SPA or MCUVE and SPA, to achieve non-destructive detection.
It realizes reliable and data-based evaluation of the quality of Salvia miltiorrhiza, and quickly and non-destructively detects the content of tanshinone IIA and/or cryptotanshinone, reducing the testing cost and avoiding damage to the medicinal materials.
Smart Images
Figure CN120668652A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of quality detection of traditional Chinese medicines, and in particular to a model for detecting the content of tanshinone IIA and / or cryptotanshinone in salvia miltiorrhiza, as well as a construction method and application thereof. Background Art
[0002] Salvia miltiorrhiza, as a traditional medicinal material with a long history and widely used in clinical Chinese medicine, has unique pharmacological activities that show significant effects in the treatment of cardiovascular and immune systems. The main active ingredients of Salvia miltiorrhiza include tanshinone IIA, salvianolic acid B and cryptotanshinone.
[0003] Traditionally, the color of salvia miltiorrhiza, particularly the depth of its red, is considered a key indicator of its quality. Traditional Chinese Medicine has traditionally evaluated salvia miltiorrhiza quality based on the principle of "identifying appearance and judging quality," considering purplish-red, thick, firm, free of breakage, and authentically sourced salvia miltiorrhiza to be top-quality. However, this evaluation method relies on visual observation, which is highly subjective and difficult to quantify.
[0004] To more scientifically and accurately assess the quality of Danshen, modern technologies have emerged. With advances in science and technology, modern techniques such as high-performance liquid chromatography (HPLC), thin-layer chromatography (TLC), gas chromatography (GC), and capillary electrophoresis (CE) have been widely used in Danshen quality control. While these techniques can provide precise and objective analytical results, they are often destructive and costly, limiting their widespread adoption in practical applications.
[0005] To address these issues, digital imaging technology, with its non-invasive nature, convenient data collection, and low cost, has shown great potential in the field of traditional Chinese medicine quality control. However, the key challenges currently remain: selecting the right digital imaging parameters for different types of traditional Chinese medicine and how to process these parameters for more accurate quality assessment. Summary of the Invention
[0006] The purpose of the present invention is to overcome the above-mentioned deficiencies of the prior art and to provide a model for detecting the content of tanshinone IIA and / or cryptotanshinone in salvia miltiorrhiza and a method for constructing and applying the model.
[0007] The first object of the present invention is to provide a method for constructing a model for detecting the content of tanshinone IIA and / or cryptotanshinone in Danshen.
[0008] The second object of the present invention is to provide a model constructed by the above construction method.
[0009] The third object of the present invention is to provide application of the above model in the quality evaluation of Salvia miltiorrhiza.
[0010] The fourth object of the present invention is to provide a method for detecting the content of tanshinone IIA and / or cryptotanshinone in Salvia miltiorrhiza.
[0011] In order to achieve the above object, the present invention is implemented through the following scheme:
[0012] A method for constructing a model for detecting the content of tanshinone IIA and / or cryptotanshinone in salvia miltiorrhiza comprises the following steps:
[0013] S1. Obtain the tanshinone IIA content and / or cryptotanshinone content of each Salvia miltiorrhiza sample in the Salvia miltiorrhiza sample set, photograph each Salvia miltiorrhiza sample to obtain an image of each Salvia miltiorrhiza sample, obtain the image to be tested of each Salvia miltiorrhiza sample, and obtain the RGB value and HSB value of the image to be tested of each Salvia miltiorrhiza sample, and count the frequency of occurrence of each value in each channel of the RGB value of the image to be tested and the frequency of occurrence of each value in each attribute of the HSB value;
[0014] S2. Processing the frequencies of occurrence of each value in each channel of the RGB values and the frequencies of occurrence of each value in each attribute of the HSB values of the image to be tested obtained in step S1 using the orthogonal signal correction method (OSC method) and then performing normalization processing to obtain the normalized frequencies of occurrence of each value in each channel of the RGB values and the normalized frequencies of occurrence of each value in each attribute of the HSB values;
[0015] S3. Dividing the Salvia miltiorrhiza sample set of step S1 into a training set and a test set, using the frequency of occurrence of each value in each channel of the standardized RGB value and the frequency of occurrence of each value in each attribute of the HSB value of each Salvia miltiorrhiza sample in the training set as input, using the tanshinone IIA content and / or cryptotanshinone content in each Salvia miltiorrhiza sample in the training set as the prediction target, training a machine learning model and testing it using the test set data to obtain a model capable of detecting the content of tanshinone IIA and / or cryptotanshinone in Salvia miltiorrhiza; the machine learning model is a PLS regression model or a RIDGE regression model;
[0016] During the training process, the frequency of occurrence of each value in each channel of the standardized RGB values of the Salvia miltiorrhiza samples and the frequency of occurrence of each value in each attribute of the HSB value were used to screen variables using the method of combining VCPA with SPA or combining MCUVE with SPA.
[0017] Preferably, the number of the Salvia miltiorrhiza samples in the Salvia miltiorrhiza sample set in step S1 is ≥30.
[0018] Preferably, the Salvia miltiorrhiza sample in step S1 is Salvia miltiorrhiza powder.
[0019] Preferably, the tanshinone IIA content and / or cryptotanshinone content of the salvia miltiorrhiza sample in step S1 is determined by high performance liquid chromatography.
[0020] More preferably, the detection wavelength of the HPLC is 280-290 nm; the column temperature is 25-30° C.; and the chromatographic column is a reverse phase chromatographic column Capcell Pak MG-C 18 or ZORBAX SB-Aq C 18 .
[0021] Further preferably, the detection wavelength of the HPLC is 280 nm; the column temperature is 25° C.; and the chromatographic column is Capcell Pak MG-C 18 .
[0022] Further preferably, the mobile phase A of the high performance liquid chromatography is acetonitrile, and the mobile phase B is an aqueous phosphoric acid solution with a volume concentration of 0.1 to 0.2%; the gradient elution conditions are (in volume percentage): 0 min, 100% mobile phase A; 5 min, 90% mobile phase A; 6 min, 80% mobile phase A; 12 min, 80% mobile phase A; 25 min, 74% mobile phase A; 33 min, 55% mobile phase A; 58 min, 35% mobile phase A; 59 min, 15% mobile phase A; 65 min, 15% mobile phase A; 70 min, 90% mobile phase A; 80 min, 90% mobile phase A.
[0023] Preferably, in step S1, after obtaining the image of each Salvia miltiorrhiza sample, an area of 250×250 pixels in the center area of the image is acquired.
[0024] Preferably, in step S1, Image J software is used to obtain RGB values and HSB values.
[0025] More preferably, the HSB value is converted to 8-bit format.
[0026] Preferably, before using the OSC data processing method in step S2, the occurrence frequencies of each value in each channel of the RGB values of the image to be tested and the occurrence frequencies of each value in each attribute of the HSB values obtained in step S1 are removed, and the values with a frequency of 0 are removed.
[0027] Preferably, the standardization process in step S2 is performed using a mean center statistical method.
[0028] Preferably, the machine learning model in step S3 is a PLS regression model.
[0029] The present invention also seeks to protect the model constructed by any of the above-mentioned construction methods.
[0030] The model constructed based on any of the above-mentioned construction methods can detect the content of tanshinone IIA and / or cryptotanshinone in Salvia miltiorrhiza according to the HSB value and RGB value of the Salvia miltiorrhiza image; the model can quickly and non-destructively (without adding organic reagents or undergoing ultrasound and heating) detect the content of tanshinone IIA and / or cryptotanshinone in Salvia miltiorrhiza, and can realize reliable and digitized quality evaluation of Salvia miltiorrhiza based on its appearance.
[0031] Therefore, the present invention also claims protection for the application of the above model in the quality evaluation of Salvia miltiorrhiza.
[0032] Preferably, the quality evaluation of Salvia miltiorrhiza is to detect the content of tanshinone IIA and / or cryptotanshinone in Salvia miltiorrhiza.
[0033] The present invention also claims a method for detecting the content of tanshinone IIA and / or cryptotanshinone in salvia miltiorrhiza, comprising the following steps:
[0034] S1. Take a picture of the Salvia miltiorrhiza sample to be tested, obtain a test image of the Salvia miltiorrhiza sample to be tested, and obtain the RGB value and HSB value of the test image, and count the frequency of occurrence of each value in each channel of the RGB value of the test image and the frequency of occurrence of each value in each attribute of the HSB value;
[0035] S2. The frequency of occurrence of each value in each channel of the RGB value of the image to be tested obtained in step S1 and the frequency of occurrence of each value in each attribute of the HSB value are processed by the OSC data processing method, and then standardized to obtain the frequency of occurrence of each value in each channel of the standardized RGB value of the tested Danshen sample and the frequency of occurrence of each value in each attribute of the HSB value;
[0036] S3. The frequency of occurrence of each value in each channel of the standardized RGB value of the tested Danshen sample obtained in step S2 and the frequency of occurrence of each value in each attribute of the HSB value are subjected to variable screening and input into any of the above-mentioned models to obtain the tanshinone IIA content and / or cryptotanshinone content of the tested Danshen sample;
[0037] The variable screening is performed by combining VCPA with SPA or MCUVE with SPA.
[0038] Preferably, in step S1, after obtaining the image to be measured, an area of 250×250 pixels in the center area of the image is acquired.
[0039] Preferably, the standardization process in step S2 is performed using a mean center statistical method.
[0040] Preferably, when obtaining the tanshinone IIA content of the salvia miltiorrhiza sample to be tested in step S3, the variable screening adopts a method combining VCPA and SPA;
[0041] When obtaining the cryptotanshinone content of the Salvia miltiorrhiza sample to be tested in step S3, the variable screening adopts the method combining MCUVE and SPA.
[0042] Compared with the prior art, the present invention has the following beneficial effects:
[0043] The present invention provides a method for constructing a model for detecting the content of tanshinone IIA and / or cryptotanshinone in salvia miltiorrhiza. The method comprises determining the content of tanshinone IIA and / or cryptotanshinone in a salvia miltiorrhiza sample and combining the RGB value and HSB value in a salvia miltiorrhiza sample image with the tanshinone IIA and / or cryptotanshinone content in the salvia miltiorrhiza sample as a prediction target, thereby training a model capable of detecting the content of tanshinone IIA and / or cryptotanshinone in salvia miltiorrhiza. Based on the model constructed above, the content of tanshinone IIA and / or cryptotanshinone in salvia miltiorrhiza can be detected according to the HSB value and RGB value of the salvia miltiorrhiza sample image to be tested, thereby achieving quality evaluation of the medicinal material without damaging the medicinal material. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 Schematic diagram of image data acquisition for the Salvia miltiorrhiza sample in Example 1; (A) shows the Salvia miltiorrhiza sample powder spread flat in a culture dish; (B) shows a photographic diagram; (C) shows the image imported into Image J software;
[0045] Figure 2 This is a data graph of variable points in the image data preprocessing process in Example 1 without being processed by the OSC data processing method;
[0046] Figure 3 This is a data graph of variable points after being processed by the OSC data processing method during the image data preprocessing process in Example 1;
[0047] Figure 4 This is a scatter plot of the VCPA-SPA-PLS model training and verification process in Example 1;
[0048] Figure 5 This is a scatter plot of the VCPA-SPA-RIDGE model training and verification process in Example 2;
[0049] Figure 6 This is a graph showing the variable stability results during the MCUVE variable screening process in Example 3;
[0050] Figure 7 This is a scatter plot of the MCUVE-SPA-PLS model training and verification process in Example 3;
[0051] Figure 8 This is a scatter plot of the MCUVE-SPA-RIDGE model training and verification process in Example 4;
[0052] Figure 9 This is a scatter plot of the VCPA-SPA-PLS model used to detect the tanshinone IIA content in samples 1 to 4 in Example 6;
[0053] Figure 10 This is a scatter plot of the cryptotanshinone content in samples 1 to 4 detected using the MCUVE-SPA-PLS model in Example 6. DETAILED DESCRIPTION
[0054] The present invention is further described in detail below with reference to the accompanying drawings and specific examples. The examples are intended only to illustrate the present invention and are not intended to limit the scope of the present invention. The experimental methods used in the following examples are conventional methods unless otherwise specified; the materials and reagents used are commercially available unless otherwise specified.
[0055] The tanshinone IIA standard reference substance used in the examples of the present invention was purchased from the China Food and Drug Inspection Institute with batch number 110766-201721; the cryptotanshinone standard reference substance was purchased from the China Food and Drug Inspection Institute with batch number 110852-201807.
[0056] Example 1 A model for detecting the content of tanshinone IIA in salvia miltiorrhiza
[0057] 1. Establishment of a model for detecting the content of tanshinone IIA in Danshen
[0058] 1. Experimental methods
[0059] (1) Data Acquisition
[0060] Forty-six Salvia miltiorrhiza samples (numbered S1 to S46) from 18 different production areas, including Weifang City, Shandong Province, Liaocheng City, Shandong Province, Bozhou City, Anhui Province, Nanyang City, Henan Province, Luoyang City, Henan Province, and Ganzhou City, Jiangxi Province, were used as the Salvia miltiorrhiza sample set. The tanshinone IIA content and cryptotanshinone content of the 46 Salvia miltiorrhiza samples were obtained, as well as the occurrence frequency of each value in each channel of the standardized RGB values and the occurrence frequency of each value in each attribute of the HSB values of the 46 Salvia miltiorrhiza samples, as shown in the following figure:
[0061] 1) Tanshinone IIA and cryptotanshinone content acquisition
[0062] The tanshinone IIA and cryptotanshinone contents of the salvia miltiorrhiza samples were determined by high performance liquid chromatography, as follows:
[0063] The Salvia miltiorrhiza sample was crushed, passed through a No. 3 sieve (50 mesh) and the sieved Salvia miltiorrhiza sample powder was collected. 0.3 g of the Salvia miltiorrhiza sample powder was placed in a stoppered conical flask, followed by adding 50 mL of methanol and weighing. After ultrasonication at 40 KHz and 500 W for 30 minutes, the weight was supplemented with methanol to the weighed weight to obtain a mixed solution of the Salvia miltiorrhiza sample.
[0064] The mixed solution of the Salvia miltiorrhiza sample was filtered through a 0.22 μm microporous filter membrane, and the filtrate was collected to obtain the Salvia miltiorrhiza sample liquid to be tested. The Salvia miltiorrhiza sample liquid to be tested was subjected to high performance liquid chromatography detection by a high performance liquid chromatograph, and the tanshinone IIA standard reference substance and the cryptotanshinone standard reference substance were used as controls to determine the concentration C1 of tanshinone IIA and the concentration C2 of cryptotanshinone in the Salvia miltiorrhiza sample liquid to be tested. Then, the tanshinone IIA content and the cryptotanshinone content in 0.3 g of the Salvia miltiorrhiza sample were calculated according to formula 1;
[0065] Formula I: Content = (50 × 100 × C × V) / m;
[0066] When calculating the content of tanshinone IIA according to formula I, C is the concentration of tanshinone IIA in the salvia miltiorrhiza sample liquid (C1); V is the injection volume of the HPLC test (10 μL); m is the mass of the salvia miltiorrhiza sample (0.5 g);
[0067] When calculating the cryptotanshinone content according to formula I, C is the concentration of cryptotanshinone C2 in the Salvia miltiorrhiza sample liquid; V is the injection volume of the HPLC test (10 μL); m is the mass of the Salvia miltiorrhiza sample (0.5 g);
[0068] The HPLC conditions were as follows: mobile phase A was acetonitrile, mobile phase B was 0.1% (v / v) phosphoric acid aqueous solution; the gradient elution conditions were as shown in Table 1; the column temperature was 35°C; the detection wavelength was 280 nm; the flow rate was 1.0 mL / min; the injection volume was 10 μL; and the chromatographic column was a reversed-phase column Capcell Pak MG-C 18 .
[0069] Table 1 Gradient elution conditions
[0070] Time / min Mobile phase A / % (v%) Mobile phase B / % (v%) 0 100 0 5 90 10 6 80 20 12 80 20 25 74 26 33 55 45 58 35 65 59 15 85 65 15 85 70 90 10 80 90 10
[0071] 2) Image data collection of Salvia miltiorrhiza samples
[0072] The schematic diagram of image data acquisition of Salvia miltiorrhiza samples is as follows Figure 1 As shown, Figure 1 (A) is the powder of the Salvia miltiorrhiza sample spread flatly in a culture dish. Figure 1 (B) is a photo-taking diagram. Figure 1 (C) is a schematic diagram of importing images into Image J software, as follows:
[0073] The camera was fixed on a bracket of a photography box (60cm×60cm×60cm), the camera lens was fixed in a sealed black box, the gaps in the black box were light-proofed, and an LED light tube was placed on top of the photography box to produce uniform lighting.
[0074] The powder of the Salvia miltiorrhiza sample in step (1) is evenly spread in a culture dish and fixed directly below the camera lens, and photographed three times continuously to obtain three images of the Salvia miltiorrhiza sample; wherein the parameters of the camera during shooting are: aperture value f / 5.6; exposure time is 1 / 250 second; ISO speed is 140; maximum aperture is 4.4; metering mode is spot; and flash mode is no flash.
[0075] Each image of the Salvia miltiorrhiza sample was imported into Image J software, and a 250×250 pixel area in the center of each image was selected as the test image of the Salvia miltiorrhiza sample. Then, data extraction was performed on the test image of each Salvia miltiorrhiza sample to obtain the RGB value and HSB value of the test image of each Salvia miltiorrhiza sample. The frequency of occurrence of each value in each channel of the RGB value of the test image of each Salvia miltiorrhiza sample {including three channels R, G and B (blue channel), the value range of each channel is 0-255} and the frequency of occurrence of each value in each attribute of the HSB value {including three attributes H, S and B (brightness), the value range of H is 0-360, the value range of S is 0-100, the value range of B is 0-100, and the HSB values are converted to 8-bit format} were counted;
[0076] The HSB value is converted to 8-bit format as follows:
[0077] The H value in the 8-bit format = the H value / 360×255; the S value in the 8-bit format = the S value / 100×255; and the B value in the 8-bit format = the B value / 100×255.
[0078] 3) Image data preprocessing
[0079] The occurrence frequencies of each value in each channel of the RGB value of each sample to be tested and the occurrence frequencies of each value in each attribute of the HSB value obtained in step 2) are merged into one data set (containing 1536 variable points, 256×6), and the values with a frequency of 0 are removed to obtain 439 variable points;
[0080] The 439 variable points were processed using the orthogonal signal correction method (OSC) and then standardized using the mean center statistical method to obtain the frequency of occurrence of each value in each channel of the standardized RGB value and the frequency of occurrence of each value in each attribute of the HSB value of each Salvia miltiorrhiza sample.
[0081] (2) Establishment of a model for detecting the content of tanshinone IIA in Salvia miltiorrhiza
[0082] The 34 Salvia miltiorrhiza samples in the Salvia miltiorrhiza sample set in step (1) were used as the model training set, and the remaining 12 Salvia miltiorrhiza samples were used as the model validation set.
[0083] The frequency of occurrence of each value in each channel of the standardized RGB values and the frequency of occurrence of each value in each attribute of the HSB values of each Salvia miltiorrhiza sample in the model training set were respectively screened by combining the variable combination cluster analysis method (VCPA) with the successive projection algorithm (SPA), and the data after variable screening were imported into Matlab software. The tanshinone IIA content in each Salvia miltiorrhiza sample in the model training set was used as the prediction target. Combined with the actual tanshinone IIA content of each Salvia miltiorrhiza sample in the model training set (obtained in step 1), a PLS regression model was trained to obtain a model for detecting the tanshinone IIA content in Salvia miltiorrhiza (VCPA-SPA-PLS model); at the same time, a scatter plot of the predicted value and the true value of the tanshinone IIA content in the Salvia miltiorrhiza samples during the PLS regression model training process was recorded.
[0084] (3) Verification of the VCPA-SPA-PLS model
[0085] The VCPA-SPA-PLS model obtained in step (2) was validated using the model validation set shown in step (2). The scatter plot of the predicted value and the true value (actual tanshinone IIA content in the salvia miltiorrhiza sample) of the model output during the validation process was recorded, and R 2 cv, Q2, RMSECV (root mean square error of cross calibration), and RMSEP (root mean square error of prediction) values.
[0086] 2. Experimental results
[0087] The results of the determination of tanshinone IIA and cryptotanshinone contents of 46 Salvia miltiorrhiza samples obtained by HPLC are shown in Table 2.
[0088] Table 2 Determination results of tanshinone IIA content and cryptotanshinone content in Salvia miltiorrhiza samples
[0089]
[0090] The data graph of the variable points in the image data preprocessing process without OSC data processing method is as follows: Figure 2 As shown, the data graph of the variable points after being processed by the OSC data processing method is as follows Figure 3 shown.
[0091] The results showed that the OSC data processing method can effectively eliminate the noise in the data, so that the variable points can be used more accurately to predict the content of tanshinone IIA in Salvia miltiorrhiza samples.
[0092] The scatter plots during the training and validation of the VCPA-SPA-PLS model are shown in Figure 2. Figure 4 As shown in the figure, the hollow circles are the samples of the model training set, and the red circles represent the samples of the model validation set; R 2 cv=0.701,Q 2 =0.915, RMSECV=0.0006, RMSEP=0.0003.
[0093] The results showed that when the constructed VCPA-SPA-PLS model was used to predict the content of tanshinone IIA in Salvia miltiorrhiza samples, all data points were clustered near the diagonal, indicating that the VCPA-SPA-PLS model can effectively predict the content of tanshinone IIA in Salvia miltiorrhiza samples and has excellent prediction performance.
[0094] 2. A model for detecting the content of tanshinone IIA in salvia miltiorrhiza
[0095] A model for detecting the content of tanshinone IIA in salvia miltiorrhiza, comprising a data acquisition module, a data processing module, a detection module and a result output module;
[0096] The data acquisition module is used to obtain the RGB value and HSB value in the image of the salvia miltiorrhiza sample powder to be tested, and to count the occurrence frequency of each value in each channel of the RGB value of the image and the occurrence frequency of each value in each attribute of the HSB value;
[0097] The data processing module processes the occurrence frequency of each value in each channel of the RGB value of the image and the occurrence frequency of each value in each attribute of the HSB value obtained by the data acquisition module according to the method described in step 3) to obtain the occurrence frequency of each value in each channel of the RGB value and the occurrence frequency of each value in each attribute of the HSB value after standardization;
[0098] Then, the frequency of occurrence of each value in each channel of the standardized RGB values and the frequency of occurrence of each value in each attribute of the HSB values were screened using a combination of variable combination cluster analysis (VCPA) and successive projection algorithm (SPA). The frequency of occurrence of each value in each channel of the screened RGB values and the frequency of occurrence of each value in each attribute of the screened HSB values were obtained.
[0099] The judgment module uses the occurrence frequency of each value in each channel of the RGB value and the occurrence frequency of each value in each attribute of the HSB value obtained by the data processing module as input to the VCPA-SPA-PLS model to obtain the tanshinone IIA content in the powder of the salvia miltiorrhiza sample to be tested;
[0100] The result output module is used to output the tanshinone IIA content in the powder of the salvia miltiorrhiza sample to be tested obtained by the judgment module.
[0101] Example 2 A model for detecting the content of tanshinone IIA in salvia miltiorrhiza
[0102] 1. Establishment of a model for detecting the content of tanshinone IIA in Danshen
[0103] 1. Experimental methods
[0104] Using the tanshinone IIA content and cryptotanshinone content of the 46 Salvia miltiorrhiza samples obtained in step (1) of Example 1 and the frequency of occurrence of each value in each channel of the standardized RGB values and the frequency of occurrence of each value in each attribute of the HSB values of the 46 Salvia miltiorrhiza samples, the PLS regression model in step (2) of Example 1 was replaced by the RIDGE regression model, and the other processing methods remained unchanged to obtain the VCPA-SPA-RIDGE model.
[0105] The VCPA-SPA-RIDGE model was validated according to the method shown in step (3) of Example 1, and the scatter plot of the predicted value and the true value (actual tanshinone IIA content in the salvia miltiorrhiza sample) of the model output during the model training and validation process was recorded, and R 2 cv, Q2, RMSECV and RMSEP values.
[0106] 2. Experimental results
[0107] The scatter plots during the training and validation of the VCPA-SPA-RIDGE model are as follows: Figure 5 As shown in the figure, the hollow circles are the samples of the model training set, and the red circles represent the samples of the model validation set; R 2 cv=0.696,Q 2 =0.906, RMSECV=0.0006, RMSEP=0.0003.
[0108] The results showed that when the constructed VCPA-SPA-PLS model was used to predict the content of tanshinone IIA in Salvia miltiorrhiza samples, all data points were concentrated near the diagonal of the scatter plot, indicating that the VCPA-SPA-PLS model can effectively predict the content of tanshinone IIA in Salvia miltiorrhiza samples.
[0109] 2. A model for detecting the content of tanshinone IIA in salvia miltiorrhiza
[0110] A model for detecting the content of tanshinone IIA in salvia miltiorrhiza, comprising a data acquisition module, a data processing module, a detection module and a result output module;
[0111] The data acquisition module is used to obtain the RGB value and HSB value in the image of the salvia miltiorrhiza sample powder to be tested, and to count the occurrence frequency of each value in each channel of the RGB value of the image and the occurrence frequency of each value in each attribute of the HSB value;
[0112] The data processing module processes the occurrence frequency of each value in each channel of the RGB value of the image and the occurrence frequency of each value in each attribute of the HSB value obtained by the data acquisition module according to the method described in step 3) of Example 1 to obtain the occurrence frequency of each value in each channel of the RGB value and the occurrence frequency of each value in each attribute of the HSB value after standardization;
[0113] Then, the frequency of occurrence of each value in each channel of the standardized RGB values and the frequency of occurrence of each value in each attribute of the HSB values were screened using a combination of variable combination cluster analysis (VCPA) and successive projection algorithm (SPA). The frequency of occurrence of each value in each channel of the screened RGB values and the frequency of occurrence of each value in each attribute of the screened HSB values were obtained.
[0114] The judgment module uses the occurrence frequency of each value in each channel of the RGB value and the occurrence frequency of each value in each attribute of the HSB value obtained by the data processing module as input to the VCPA-SPA-RIDGE model to obtain the tanshinone IIA content in the powder of the salvia miltiorrhiza sample to be tested;
[0115] The result output module is used to output the tanshinone IIA content in the powder of the salvia miltiorrhiza sample to be tested obtained by the judgment module.
[0116] Example 3 A model for detecting the content of cryptotanshinone in Salvia miltiorrhiza
[0117] 1. Establishment of a model for detecting the content of cryptotanshinone
[0118] 1. Experimental methods
[0119] (1) Data Acquisition
[0120] According to the method shown in step (1) of Example 1, the tanshinone IIA content and cryptotanshinone content of the 46 Salvia miltiorrhiza samples in the Salvia miltiorrhiza sample set shown in Example 1 and the frequency of occurrence of each value in each channel of the standardized RGB values of the 46 Salvia miltiorrhiza samples and the frequency of occurrence of each value in each attribute of the HSB value were obtained.
[0121] (2) Establishment of a model for detecting the content of tanshinone IIA in Salvia miltiorrhiza
[0122] Thirty-four of the 46 Salvia miltiorrhiza samples were used as the model training set, and the remaining 12 Salvia miltiorrhiza samples were used as the model validation set.
[0123] The frequency of occurrence of each value in each channel of the standardized RGB values and the frequency of occurrence of each value in each attribute of the HSB values of each Salvia miltiorrhiza sample in the model training set were respectively screened by the Monte Carlo signalless information variable elimination method (MCUVE) combined with the SPA method, and the data after variable screening were imported into Matlab software. The cryptotanshinone content in each Salvia miltiorrhiza sample in the model training set was used as the prediction target, and the PLS regression model was trained in combination with the actual cryptotanshinone content of each Salvia miltiorrhiza sample in the model training set (obtained in step (1)), and a model for detecting the cryptotanshinone content in Salvia miltiorrhiza (MCUVE-SPA-PLS model) was obtained; and a scatter plot of the predicted value and the true value of the tanshinone IIA content in the Salvia miltiorrhiza sample during the PLS regression model training process was recorded.
[0124] (3) Verification of the MCUVE-SPA-PLS model
[0125] The MCUVE-SPA-PLS model obtained in step (2) was verified using the model verification set shown in step (2), and a scatter plot of the predicted value and true value (actual cryptotanshinone content in the Salvia miltiorrhiza sample) of the cryptotanshinone content in the Salvia miltiorrhiza sample output by the model during the verification process was recorded.
[0126] 2. Experimental results
[0127] The variable stability results during the MCUVE variable screening process are shown in the figure below: Figure 6 As shown, Figure 6 The red dashed line in the figure is the cutoff value (threshold) for MCUVE variable screening. During the variable screening process, variables with an absolute value of stability greater than the absolute value of the threshold are selected.
[0128] The scatter plots during the training and validation of the MCUVE-SPA-PLS model are as follows: Figure 7 As shown in the figure, the hollow circles are the samples of the model training set, and the red circles represent the samples of the model validation set. The results show that when the constructed MCUVE-SPA-PLS model is used to predict the cryptotanshinone content in Salvia miltiorrhiza samples, all the data points are concentrated near the diagonal line of the scatter plot, indicating that the MCUVE-SPA-PLS model can effectively predict the cryptotanshinone content in Salvia miltiorrhiza samples and has good robustness.
[0129] 2. A model for detecting cryptotanshinone content in Salvia miltiorrhiza
[0130] A model for detecting the content of cryptotanshinone in salvia miltiorrhiza, comprising a data acquisition module, a data processing module, a detection module and a result output module;
[0131] The data acquisition module is used to obtain the RGB value and HSB value in the image of the salvia miltiorrhiza sample powder to be tested, and to count the occurrence frequency of each value in each channel of the RGB value of the image and the occurrence frequency of each value in each attribute of the HSB value;
[0132] The data processing module processes the occurrence frequency of each value in each channel of the RGB value of the image and the occurrence frequency of each value in each attribute of the HSB value obtained by the data acquisition module according to the method shown in step 3) of Example 1 to obtain the occurrence frequency of each value in each channel of the RGB value and the occurrence frequency of each value in each attribute of the HSB value after standardization;
[0133] Then, the frequency of occurrence of each value in each channel of the standardized RGB values and the frequency of occurrence of each value in each attribute of the HSB values were screened using the Monte Carlo signal-free variable elimination method (MCUVE) combined with SPA to obtain the frequency of occurrence of each value in each channel of the filtered RGB values and the frequency of occurrence of each value in each attribute of the HSB values.
[0134] The judgment module uses the occurrence frequency of each value in each channel of the RGB value after screening and the occurrence frequency of each value in each attribute of the HSB value obtained by the data processing module as input to the MCUVE-SPA-PLS model to obtain the cryptotanshinone content in the powder of the Salvia miltiorrhiza sample to be tested;
[0135] The result output module is used to output the cryptotanshinone content in the powder of the Salvia miltiorrhiza sample to be tested obtained by the judgment module.
[0136] Example 4 A model for detecting the content of cryptotanshinone in Salvia miltiorrhiza
[0137] 1. Establishment of a model for detecting the content of cryptotanshinone
[0138] 1. Experimental methods
[0139] Using the tanshinone IIA content and cryptotanshinone content of the 46 Salvia miltiorrhiza samples obtained in step (1) of Example 1 and the frequency of occurrence of each value in each channel of the standardized RGB values and the frequency of occurrence of each value in each attribute of the HSB values of the 46 Salvia miltiorrhiza samples, the PLS regression model in step (2) of Example 3 was replaced with the RIDGE regression model, and the other processing methods remained unchanged to obtain the MCUVE-SPA-RIDGE model.
[0140] The MCUVE-SPA-RIDGE model was verified according to the method shown in step (3) of Example 3, and a scatter plot of the predicted value and true value (actual cryptotanshinone content in the Salvia miltiorrhiza sample) of the cryptotanshinone content of the Salvia miltiorrhiza sample output by the model during the model training and verification process was recorded.
[0141] 2. Experimental results
[0142] The scatter plots during the training and validation of the MCUVE-SPA-RIDGE model are as follows: Figure 8 As shown in the figure, the hollow circles are the samples of the model training set, and the red circles represent the samples of the model validation set. The results show that when the constructed MCUVE-SPA-RIDGE model is used to predict the cryptotanshinone content in Salvia miltiorrhiza samples, all the data points are concentrated near the diagonal line of the scatter plot, indicating that the MCUVE-SPA-RIDGE model can effectively predict the cryptotanshinone content in Salvia miltiorrhiza samples and has good robustness.
[0143] 2. A model for detecting cryptotanshinone content in Salvia miltiorrhiza
[0144] A model for detecting the content of cryptotanshinone in salvia miltiorrhiza, comprising a data acquisition module, a data processing module, a detection module and a result output module;
[0145] The data acquisition module is used to obtain the RGB value and HSB value in the image of the salvia miltiorrhiza sample powder to be tested, and to count the occurrence frequency of each value in each channel of the RGB value of the image and the occurrence frequency of each value in each attribute of the HSB value;
[0146] The data processing module processes the occurrence frequency of each value in each channel of the RGB value of the image obtained by the data acquisition module and the occurrence frequency of each value in each attribute of the HSB value according to the method described in step (3) of Example 1 to obtain the occurrence frequency of each value in each channel of the standardized RGB value and the occurrence frequency of each value in each attribute of the HSB value;
[0147] Then, the frequency of occurrence of each value in each channel of the standardized RGB values and the frequency of occurrence of each value in each attribute of the HSB values were screened using the Monte Carlo signal-free variable elimination method (MCUVE) combined with SPA to obtain the frequency of occurrence of each value in each channel of the filtered RGB values and the frequency of occurrence of each value in each attribute of the HSB values.
[0148] The judgment module uses the occurrence frequency of each value in each channel of the RGB value after screening and the occurrence frequency of each value in each attribute of the HSB value obtained by the data processing module as input, and inputs them into the MCUVE-SPA-RIDGE model to obtain the cryptotanshinone content in the powder of the Salvia miltiorrhiza sample to be tested;
[0149] The result output module is used to output the cryptotanshinone content in the powder of the Salvia miltiorrhiza sample to be tested obtained by the judgment module.
[0150] Example 5 Methodological Investigation of Model Construction
[0151] 1. Investigation of Data Preprocessing Methods
[0152] 1. Experimental methods
[0153] The 46 Salvia miltiorrhiza samples (numbered S1 to S46) in Example 1 were divided into groups using the Kennard-Stone algorithm, and the calibration set (70%) and the prediction set (30%) were obtained according to a 7:3 ratio.
[0154] According to the method shown in Example 1, the Salvia miltiorrhiza samples in the calibration set and the prediction set were processed separately to obtain the tanshinone IIA content and cryptotanshinone content of each Salvia miltiorrhiza sample in the calibration set prediction set, as well as the frequency of occurrence of each value in each channel of the standardized RGB value of each Salvia miltiorrhiza sample and the frequency of occurrence of each value in each attribute of the HSB value.
[0155] Experimental group 1: According to the method shown in Example 1, the VCPA-SPA-PLS model was constructed using the calibration set according to the five-fold cross-validation method, and the effect of the model was calculated in combination with the prediction set, including R 2 ,RMSE,R 2 cv、RMSECV、Q 2 and RMSEP, while calculating RPD and Nlv (optimal number of latent variables).
[0156] Experimental group 2: According to the method shown in Example 2, the VCPA-SPA-RIDGE model was constructed using the calibration set according to the five-fold cross-validation method, and the effect of the model was calculated in combination with the prediction set, including R 2 ,RMSE,R 2 cv、RMSECV、Q 2 and RMSEP, while RPD and N1v are calculated.
[0157] Experimental group 3: According to the method shown in Example 3, the MCUVE-SPA-PLS model was constructed using the calibration set according to the five-fold cross-validation method, and the effect of the model was calculated in combination with the prediction set, including R 2 ,RMSE,R 2 cv、RMSECV、Q 2 and RMSEP, while RPD and N1v are calculated.
[0158] Experimental group 4: According to the method shown in Example 4, the MCUVE-SPA-RIDGE model was constructed using the calibration set according to the five-fold cross-validation method, and the effect of the model was calculated in combination with the prediction set, including R 2 ,RMSE,R 2 cv、RMSECV、Q 2 and RMSEP, while RPD and N1v are calculated.
[0159] The difference between control group 1 and experimental group 1 is that in the process of model construction, the image data preprocessing step does not use the OSC data processing method for processing, and the rest of the processing is exactly the same. The R of the model constructed by control group 1 is 2 ,RMSE,R 2 cv、RMSECV、Q 2 and RMSEP, while RPD and N1v are calculated.
[0160] The difference between control group 2 and experimental group 2 is that in the process of model construction, the image data preprocessing step does not use the OSC data processing method for processing, and the rest of the processing is exactly the same. The R of the model constructed by control group 2 is 2 ,RMSE,R 2 cv、RMSECV、Q 2 and RMSEP, while RPD and N1v are calculated.
[0161] The difference between control group 3 and experimental group 3 is that in the process of model construction, the image data preprocessing step does not use the OSC data processing method for processing, and the rest of the processing is exactly the same. The R of the model constructed by control group 3 is 2 ,RMSE,R 2 cv、RMSECV、Q 2 and RMSEP, while RPD and N1v are calculated.
[0162] The difference between control group 4 and experimental group 4 is that in the process of model construction, the image data preprocessing step does not use the OSC data processing method for processing, and the rest of the processing is exactly the same. The R of the model constructed by control group 4 is 2 ,RMSE,R 2 cv、RMSECV、Q 2 and RMSEP, while RPD and N1v are calculated.
[0163] 2. Experimental results
[0164] The model effects of experimental groups 1 to 4 and control groups 1 to 4 are shown in Table 3.
[0165] Table 3 Model results
[0166]
[0167] The results showed that compared with the experimental group 1, the R 2 From 0.867 to 0.571, Q 2 It decreased from 0.915 to 0.738, RMSE increased from 0.0004 to 0.0007, and RMSEP increased from 0.0003 to 0.0005;
[0168] Compared with experimental group 2, the R 2 From 0.856 to 0.571, Q 2 The data decreased from 0.906 to 0.73; RMSE increased from 0.0004 to 0.0007, and RMSEP increased from 0.0003 to 0.0005;
[0169] Compared with experimental group 3, the R 2 From 0.907 to 0.4, Q 2 It decreased from 0.916 to 0.744, RMSE increased from 0.0003 to 0.0007, and RMSEP increased from 0.0002 to 0.0004;
[0170] R of the model constructed for control group 4 2 It decreased from 0.907 to 0.672, and RMSE increased from 0.0003 to 0.0005.
[0171] The above results all indicate that in the process of constructing the model shown in the present invention, combining the OSC data processing method to process the variables and then constructing the model can make the model performance more excellent, and thus more accurately predict the tanshinone IIA content or cryptotanshinone content in Salvia miltiorrhiza.
[0172] 2. Investigation of variable screening methods
[0173] 1. Experimental methods
[0174] The variable screening methods in the process of establishing the model of Examples 1 to 4 were replaced in sequence with the method of combining MCUVE and GA, the method of combining MCUVE and IRIV, the method of combining MCUVE and SPA, the method of combining VCPA and GA, the method of combining VCPA and IRIV, the method of combining VCPA and SPA, the IRIV method, the MCUVE method, the SPA method, the VCPA method and the GA method, respectively, to construct a model for detecting the content of tanshinone IIA in Danshen and a model for detecting the content of cryptotanshinone in Danshen, and calculate the effect of the model, including R 2 ,RMSE,R 2 cv、RMSECV、Q 2and RMSEP, while RPD and N1v are calculated.
[0175] At the same time, without variable screening, the model was established according to Examples 1 to 4 and the effects of the model were calculated, including R 2 ,RMSE,R 2 cv、RMSECV、Q 2 and RMSEP, while RPD and N1v are calculated.
[0176] 2. Experimental results
[0177] The model effects during the variable screening method investigation are shown in Table 4.
[0178] Table 4 Model effects during the investigation of variable screening methods
[0179]
[0180]
[0181] The results showed that compared with the model constructed without variable screening, the model constructed after variable screening had better performance in predicting the content of tanshinone IIA in Danshen or predicting the content of cryptotanshinone in Danshen; and when the model was used to predict the content of tanshinone IIA in Danshen, the Q of the model constructed by the variable screening method combining VCPA and SPA or the variable screening method combining MCUVE and SPA was significantly improved. 2 It is significantly better than the models constructed using other variable screening methods, and the model constructed by the variable screening method combining VCPA and SPA can most accurately predict the content of tanshinone IIA in Salvia miltiorrhiza.
[0182] When the model was used to predict the cryptotanshinone content in Salvia miltiorrhiza, the performance of the model constructed by the variable screening method combining VCPA and SPA or the variable screening method combining MCUVE and SPA was significantly better than that of the model constructed using other variable screening methods, and the model constructed by the variable screening method combining MCUVE and SPA was able to most accurately predict the cryptotanshinone content in Salvia miltiorrhiza.
[0183] Example 6: Verification of the detection effect of the model
[0184] 1. Experimental Methods
[0185] Take 4 Salvia miltiorrhiza samples (sample numbers are sample 1 to sample 4, excluding any one of the 46 Salvia miltiorrhiza samples in Example 1), and according to the "Obtaining the content of tanshinone IIA and cryptotanshinone" steps described in Example 1, use high performance liquid chromatography to determine the content of cryptotanshinone and tanshinone IIA in samples 1 to 4, respectively, and record them as the measured values of cryptotanshinone and tanshinone IIA in samples 1 to 4.
[0186] The VCPA-SPA-PLS model shown in Example 1 was used to determine the tanshinone IIA content (predicted value) of each sample in Samples 1 to 4, respectively. Then, the MCUVE-SPA-PLS model shown in Example 3 was used to determine the cryptotanshinone content (predicted value) of each sample in Samples 1 to 4, respectively.
[0187] Then, SPSS26 software was used to perform independent sample t-tests on the predicted and measured values of tanshinone IIA content and cryptotanshinone content, and the t-test results were recorded.
[0188] Then, the scatter plots of the tanshinone IIA content and cryptotanshinone content in samples 1 to 4 were recorded respectively using the VCPA-SPA-PLS model and the MCUVE-SPA-PLS model.
[0189] 2. Experimental Results
[0190] The determination results of tanshinone IIA content and cryptotanshinone content of samples 1 to 4 are shown in Table 5.
[0191] Table 5 Determination results of tanshinone IIA content and cryptotanshinone content
[0192]
[0193] The P value of the independent sample t-test for tanshinone IIA content was 0.947, and the P value of the independent sample t-test for cryptotanshinone content was 0.948.
[0194] The scatter plot of the VCPA-SPA-PLS model for detecting the content of tanshinone IIA in samples 1 to 4 is shown in the figure below. Figure 9 As shown, Figure 9 The red dots in the figure represent samples 1 to 4. The scatter plots of the cryptotanshinone content in samples 1 to 4 detected using the MCUVE-SPA-PLS model are shown in Figure 2. Figure 10 As shown, Figure 10 The red dots in the figure represent samples 1 to 4.
[0195] The results showed that for Samples 1 to 4, the tanshinone IIA content measured using the VCPA-SPA-PLS model described in Example 1 was not significantly different from the actual tanshinone IIA content in the samples (independent t-test P value > 0.05); the cryptotanshinone content measured using the MCUVE-SPA-PLS model described in Example 3 was also not significantly different from the tanshinone content in the samples; and the scatter plot also showed that all data points were concentrated near the diagonal line. These results indicate that the VCPA-SPA-PLS model described in Example 1 and the MCUVE-SPA-PLS model described in Example 3 can accurately predict the tanshinone IIA and cryptotanshinone content in the Salvia miltiorrhiza samples.
[0196] Example 7 A method for detecting the content of tanshinone IIA in salvia miltiorrhiza
[0197] A method for detecting the content of tanshinone IIA in salvia miltiorrhiza comprises the following steps:
[0198] S1. The powder of the Salvia miltiorrhiza sample to be tested is photographed to obtain a test image of the Salvia miltiorrhiza sample to be tested, the RGB value and HSB value of the test image are obtained, and the frequency of occurrence of each value in each channel of the RGB value of the test image and the frequency of occurrence of each value in each attribute of the HSB value are counted;
[0199] S2. According to step 3) of Example 1, the frequency of occurrence of each value in each channel of the RGB value of the image to be tested and the frequency of occurrence of each value in each attribute of the HSB value are processed to obtain the frequency of occurrence of each value in each channel of the RGB value and the frequency of occurrence of each value in each attribute of the HSB value;
[0200] S3. Perform variable screening on the frequency of occurrence of each value in each channel of the standardized RGB values and the frequency of occurrence of each value in each attribute of the HSB values obtained in step S2 using a combination of variable combination cluster analysis (VCPA) and a successive projection algorithm (SPA) to obtain the frequency of occurrence of each value in each channel of the filtered RGB values and the frequency of occurrence of each value in each attribute of the filtered HSB values;
[0201] The occurrence frequency of each value in each channel of the screened RGB values and the occurrence frequency of each value in each attribute of the HSB values were input into the VCPA-SPA-PLS model shown in Example 1 to obtain the tanshinone IIA content in the salvia miltiorrhiza sample powder to be tested.
[0202] Example 8 A method for detecting the content of cryptotanshinone in Salvia miltiorrhiza
[0203] A method for detecting the content of cryptotanshinone in Salvia miltiorrhiza comprises the following steps:
[0204] S1. The powder of the Salvia miltiorrhiza sample to be tested is photographed to obtain a test image of the Salvia miltiorrhiza sample to be tested, the RGB value and HSB value of the test image are obtained, and the frequency of occurrence of each value in each channel of the RGB value of the test image and the frequency of occurrence of each value in each attribute of the HSB value are counted;
[0205] S2. According to step 3) of Example 1, the frequency of occurrence of each value in each channel of the RGB value of the image to be tested and the frequency of occurrence of each value in each attribute of the HSB value are processed to obtain the frequency of occurrence of each value in each channel of the RGB value and the frequency of occurrence of each value in each channel of the HSB value;
[0206] S3. Filter the occurrence frequencies of each value in each channel of the standardized RGB values obtained in step S2 and the occurrence frequencies of each value in each attribute of the HSB values using a Monte Carlo signal-free variable elimination method (MCUVE) combined with SPA to obtain the occurrence frequencies of each value in each channel of the filtered RGB values and the occurrence frequencies of each value in each attribute of the HSB values;
[0207] The occurrence frequency of each value in each channel of the screened RGB values and the occurrence frequency of each value in each attribute of the HSB values are input into the MCUVE-SPA-PLS model shown in Example 3 to obtain the cryptotanshinone content in the Salvia miltiorrhiza sample powder to be tested.
[0208] Finally, it should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Those skilled in the art will readily appreciate that other variations or modifications may be made based on the above descriptions and concepts. It is not necessary and impossible to provide an exhaustive list of all possible implementations. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the claims of the present invention.
Claims
1. A method for constructing a model for detecting the content of tanshinone IIA and / or cryptotanshinone in Danshen, characterized in that: The following steps are involved: S1. Obtain the tanshinone IIA content and / or cryptotanshinone content of each Salvia miltiorrhiza sample in the Salvia miltiorrhiza sample set, photograph each Salvia miltiorrhiza sample to obtain an image of each Salvia miltiorrhiza sample, obtain the image to be tested of each Salvia miltiorrhiza sample, and obtain the RGB value and HSB value of the image to be tested of each Salvia miltiorrhiza sample, and count the frequency of occurrence of each value in each channel of the RGB value of the image to be tested and the frequency of occurrence of each value in each attribute of the HSB value; S2. The frequencies of occurrence of each value in each channel of the RGB values and each value in each attribute of the HSB values of the image to be tested obtained in step S1 are processed using the orthogonal signal correction method and then normalized to obtain the frequencies of occurrence of each channel value in each channel of the RGB values and the frequencies of occurrence of each value in each attribute of the HSB values; S3. Dividing the Salvia miltiorrhiza sample set of step S1 into a training set and a test set, using the frequency of occurrence of each value in each channel of the standardized RGB value and the frequency of occurrence of each value in each attribute of the HSB value of each Salvia miltiorrhiza sample in the training set as input, using the tanshinone IIA content and / or cryptotanshinone content in each Salvia miltiorrhiza sample in the training set as the prediction target, training a machine learning model and testing it using the test set data to obtain a model capable of detecting the content of tanshinone IIA and / or cryptotanshinone in Salvia miltiorrhiza; the machine learning model is a PLS regression model or a RIDGE regression model; During the training process, the frequency of occurrence of each value in each channel of the standardized RGB values of the Salvia miltiorrhiza samples and the frequency of occurrence of each value in each attribute of the HSB value were used to screen variables using the method of combining VCPA with SPA or combining MCUVE with SPA.
2. The construction method according to claim 1, characterized in that The tanshinone IIA content and / or cryptotanshinone content of each salvia miltiorrhiza sample in step S1 is determined by high performance liquid chromatography.
3. The construction method according to claim 1, characterized in that Before the OSC data processing method is used in step S2 , the occurrence frequencies of each value in each channel of the RGB values of the image to be tested and the occurrence frequencies of each value in each attribute of the HSB values obtained in step S1 are removed, and the values with a frequency of 0 are removed.
4. The construction method according to claim 1, characterized in that The standardization process in step S2 is performed using a mean center statistical method.
5. The construction method according to claim 1, characterized in that The machine learning model described in step S3 is a PLS regression model.
6. The model constructed by the construction method according to any one of claims 1 to 5.
7. Application of the model according to claim 6 in the quality evaluation of Salvia miltiorrhiza.
8. The use according to claim 7, characterized in that The quality evaluation of Salvia miltiorrhiza is to detect the content of tanshinone IIA and / or cryptotanshinone in Salvia miltiorrhiza.
9. A method for detecting the content of tanshinone IIA and / or cryptotanshinone in Danshen, characterized in that: The following steps are involved: S1. The tested Salvia miltiorrhiza samples were photographed to obtain a test image of the tested Salvia miltiorrhiza sample, and the RGB and HSB values of the tested image were obtained. The frequency of occurrence of each value in each channel of the RGB value of the tested image and the frequency of occurrence of each value in each attribute of the HSB value were counted. S2. The frequency of occurrence of each value in each channel of the RGB value of the image to be tested obtained in step S1 and the frequency of occurrence of each value in each attribute of the HSB value are processed by the OSC data processing method, and then standardized to obtain the frequency of occurrence of each value in each channel of the standardized RGB value of the tested Danshen sample and the frequency of occurrence of each value in each attribute of the HSB value; S3. The frequency of occurrence of each value in each channel of the standardized RGB value of the tested Danshen sample obtained in step S2 and the frequency of occurrence of each value in each attribute of the HSB value are subjected to variable screening and input into the model described in Example 6 to obtain the tanshinone IIA content and / or cryptotanshinone content of the tested Danshen sample; The variable screening is performed by combining VCPA with SPA or MCUVE with SPA.
10. The method according to claim 9, characterized in that When obtaining the tanshinone IIA content of the salvia miltiorrhiza sample to be tested in step S3, the variable screening adopts the method combining VCPA and SPA; When obtaining the cryptotanshinone content of the Salvia miltiorrhiza sample to be tested in step S3, the variable screening adopts the method combining MCUVE and SPA.