Method and model for predicting chemical component content of motherwort, storage medium and modeling method

Abstract

The invention belongs to the field of model prediction, and particularly relates to a method for predicting the chemical component content of motherwort, a model, a storage medium and a modeling method. The modeling method comprises the following steps: 1) extracting motherwort features; 2) constructing a model; and 3) optimizing the model. Based on the multispectral imaging technology, the partial least squares regression (PLSR) method is combined with Bayesian optimization, the constructed total alkaloid and total flavone content prediction model can rapidly and accurately predict the content of total alkaloid and total flavone in motherwort, the purpose of rapidly evaluating the quality of motherwort is achieved, and the method has practical application value.

Description

Technical Field

[0001] The invention belongs to the field of model prediction, and in particular relates to a method, a model, a storage medium and a modeling method for predicting the content of chemical components of motherwort. Background Art

[0002] Motherwort is the fresh or dry part of Leonurus japonicas Houtt., a plant of the genus Leonurus in the family Lamiaceae. It was originally named Cnidium monnieri and was first recorded in Shennong's Herbal Classic. It is also known as Yimu Ai, Kucao, Kuncao, etc. It has a bitter taste and is mild in nature. It is used in the pericardium and liver meridians, and has the effects of promoting blood circulation, regulating blood stasis, and promoting diuresis and reducing swelling. It is clinically used to treat irregular menstruation, postpartum blood stasis and pain, cardiovascular and cerebrovascular diseases, blood diseases and other diseases.

[0003] The main chemical components of motherwort include total flavonoids and total alkaloids, which have various uses in medicine. The total flavonoids in motherwort have significant anti-inflammatory, antioxidant, and anti-tumor biological activities. Studies have shown that flavonoids in motherwort have significant inhibitory effects on various inflammatory models such as acute lung injury and acute pancreatitis, and can inhibit the production and release of inflammatory mediators, thereby reducing inflammatory responses. In addition, total flavonoids in motherwort have also been found to have a protective effect on myocardial ischemia and can improve myocardial contractile function in rats with coronary artery ligation. The total alkaloids in motherwort are also important active ingredients with multiple biological activities such as anti-inflammatory, antibacterial, and anti-tumor. The total alkaloids of motherwort show good application prospects in the treatment of endometritis, and can play a role by inhibiting inflammatory responses and promoting the repair of epithelial cells. In addition, the total alkaloids of motherwort also have significant anti-dysmenorrhea effects, which can relieve the strong contraction of the rat uterus caused by oxytocin and prostaglandin E2.

[0004] At present, the detection of total flavonoids and total alkaloids in Leonurus involves a variety of chemical analysis techniques, including ultraviolet spectrophotometry, high performance liquid chromatography, supercritical CO2 extraction, etc. However, these methods are not only time-consuming but also consume Leonurus. Therefore, it is necessary to provide a method that can non-destructively, quickly and accurately predict the total flavonoids and total alkaloids content of large quantities of Leonurus, so as to quickly evaluate the quality of Leonurus.

[0005] Multispectral imaging technology is a fast and non-destructive detection method and is widely used in the detection of plant components. However, the differences between different plant species lead to changes in spectral characteristics, which limits the generalization ability of the model. Even if the total amount of components is the same, if the distribution of these components in the plant is different, their spectral characteristics will also change. This makes it difficult to directly apply the plant component prediction model based on multispectral imaging technology to the prediction of the content of different plant components. Therefore, it is necessary to develop new prediction models for specific plants. At present, there is no research on the prediction model of the chemical component content of Leonurus japonicus based on multispectral imaging technology. Summary of the invention

[0006] To solve the above problems, the present invention provides a method for constructing a prediction model for the chemical component content of Leonurus japonicus based on multispectral imaging technology, comprising the following steps:

[0007] 1) Extracting the characteristics of Leonurus japonicus: taking a leaf of Leonurus japonicus, imaging it using multispectral imaging technology, extracting the spectral data of the leaf, and then using an alkaloid content determination kit and / or a plant flavonoid content detection kit to determine the content of total alkaloids and / or total flavonoids in the leaf;

[0008] 2) constructing a model: using the spectral data obtained in step 1) as a characteristic variable, the content of total alkaloids and / or total flavonoids as a target variable, and using a PLSR algorithm for training to obtain a PLSR model;

[0009] 3) Optimizing the model: The PLSR model obtained in step 2) was optimized using a Bayesian optimization algorithm to obtain a prediction model for the chemical component content of Leonurus japonicus.

[0010] Furthermore, the pixels of the spectral image imaged by the multi-spectral imaging technology in step 1) are 2056×2056, and the resolution is 41 μm / pixel.

[0011] Furthermore, the spectral data is the average reflectivity of the blade in the spectral wavelength range of 365 to 970 nm.

[0012] Furthermore, in step 2), the PLSR algorithm is used for training. 2 , RMSE and RPD are used to evaluate the prediction performance of the model.

[0013]

[0014]

[0015]

[0016] In the formula, is the predicted value of TA and TF content, is the average value of TA and TF content, y i are the measured values ​​of TA and TF contents, n is the number of samples i=1,2,3…n, and SD is the standard deviation.

[0017] Furthermore, the Bayesian optimization algorithm in step 3) is derived from the Hyperopt library.

[0018] Furthermore, the model for predicting the content of chemical components of Leonurus japonicus predicts the content of total alkaloids and / or total flavonoids in the leaves of Leonurus japonicus.

[0019] The present invention also provides a prediction model for the chemical component content of Leonurus japonicus constructed by the aforementioned method.

[0020] The present invention also provides a method for detecting the content of chemical components in Leonurus japonicus, comprising the following steps:

[0021] ① Perform multispectral imaging on the motherwort to be tested, and extract the average reflectivity of the motherwort leaves in the spectral wavelength range of 365 to 970 nm;

[0022] ② Input the average reflectance obtained in step ① into the aforementioned prediction model to obtain the prediction results of the chemical components of Leonurus japonicus;

[0023] The chemical components of Leonurus japonicus are total alkaloids and / or total flavonoids.

[0024] The present invention also provides a system for implementing the above-mentioned method for detecting the content of chemical components in Leonurus japonicus, comprising:

[0025] An input module, used for inputting the characteristics of Leonurus japonicus to be tested;

[0026] A computing module, including the aforementioned prediction model;

[0027] An output module, used for outputting the final prediction results of the chemical components of Leonurus japonicus;

[0028] The chemical components of Leonurus japonicus are total alkaloids and / or total flavonoids.

[0029] Finally, the present invention provides a computer-readable storage medium on which a computer program is stored, wherein the computer program is used to implement the aforementioned method for detecting the content of chemical components in Leonurus japonicus.

[0030] The present invention is based on multispectral imaging technology, adopts partial least squares regression (PLSR) combined with Bayesian optimization, and constructs a total alkaloid and total flavonoid content prediction model, which can quickly and accurately predict the content of total alkaloids and total flavonoids in motherwort, achieve the purpose of quickly evaluating the quality of motherwort, and has practical application value.

[0031] Obviously, according to the above contents of the present invention, in accordance with common technical knowledge and customary means in the art, without departing from the above basic technical ideas of the present invention, other various forms of modification, replacement or change may be made.

[0032] The above contents of the present invention are further described in detail below through specific implementation methods in the form of embodiments. However, this should not be understood as the scope of the above subject matter of the present invention being limited to the following examples. All technologies realized based on the above contents of the present invention belong to the scope of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Spectral curve of motherwort leaves, A: all samples; B: average reflectance of the control group (CK); C: average reflectance of the experimental group (DS)

[0034] Figure 2 Correlation between total alkaloids, total flavonoids and bands; TA in the figure: total alkaloids; TF: total flavonoids.

[0035] Figure 3 Spectral curves of different pre-treatments, A: original spectrum Constant; B: MSC treatment; C: SG treatment; D: SNV treatment; E: 1st Der treatment; F: 2nd Der treatment

[0036] Figure 4 Changing trends of total alkaloid and total flavonoid content and PLSR prediction model, A: Changing trend of total alkaloid content; B: Changing trend of total flavonoid content; C: PLSR prediction model of total alkaloid content; D: PLSR prediction model of total flavonoid content. DETAILED DESCRIPTION

[0037] Example 1 Study on nondestructive detection method of chemical components content of Leonurus japonicus based on multispectral imaging technology

[0038] 1. Materials and Methods

[0039] 1.1. Multispectral imaging system data acquisition

[0040] The VideometerLab4 instrument (Videometer A / S, Denmark) collected multispectral images of motherwort leaves, with a spectral wavelength range of 365 to 970 nm. The obtained spectral image pixels were 2056×2056, the resolution was 41μm / pixel, and the processing time for each sample was 5 to 10 seconds. Before the test, the VideometerLab4 instrument was preheated for 30 minutes, the working parameters of MSI were set, and a blue background was used as the stage background during the measurement to facilitate the segmentation and identification of the leaves and enhance the color contrast between the leaves and the background. The specific operation is as follows: the motherwort leaves were placed on the loading platform and fixed, and then the hollow integrating sphere was moved downward and removed after 4 to 5 seconds. The 2.12.23 version of VideometerLab software was used for image segmentation. The spectral image of the leaves was segmented according to the typical discriminant analysis and threshold setting, the background was removed, and the Blob segmentation image of the leaves was created. Ten rectangular or circular regions of interest (ROIs) were created on the segmented motherwort leaf image, and then the leaf average reflectance data was extracted. Each reflectance spectrum was calculated as the average value within the ROI. The x-axis in the figure is wavelength, the unit is nanometer (nm), and the y-axis is reflectivity, the unit is percentage (%), that is, the percentage of incident diffuse light that is reflected back.

[0041] 1.2. Plant materials and stress treatments

[0042] Experimental materials: Motherwort seeds were purchased from the standardized planting base of Motherwort in Mianning County, Liangshan Prefecture, built by Chengdu No. 1 Pharmaceutical Co., Ltd. They were sown in January 2024 and planted in nutrient pots in April 2024. Two plants were planted in each pot, and plants with uniform growth were selected in June 2024. The potted plants of Motherwort that had been cultivated normally for 3 months were subjected to stress treatment by continuous natural drought, and the full water supply treatment was used as the control: the soil moisture content was maintained at 75% to 85% of the field water holding capacity. Before the stress treatment, the potted plants were watered normally. The potted seedlings were watered thoroughly on the night before the treatment and then stopped watering. The soil moisture content was kept at 80% in the normal watering group. The other treatment groups were subjected to drought treatment by natural drought method. Rehydration and ABA spraying were carried out on the 8th day. Samples were taken on the 0th, 2nd, 4th, 6th, 8th, and 10th days of stress, and multispectral imaging and chemical component content were measured. The experiment adopted a completely randomized design. Eight plants were taken from the experimental group and the control group every day. The number of samples in each group was three leaves each time. The total sample size of the experiment was 156 leaves. Each leaf was measured in the order of multi-spectral measurement first and then other indicators.

[0043] 1.3. Determination of chemical composition content

[0044] Alkaloids can react with bromocresol green indicator to generate green compounds, and the product has a characteristic absorption peak at 416nm. The content of alkaloids can be quantitatively detected by the change of absorbance value; flavonoids and aluminum ions can form red complexes in alkaline nitrite solution, and the product has a characteristic absorption peak at 510nm. The content of plant flavonoids can be quantitatively detected by the change of absorbance value. The leaves that have been scanned by too much spectrum are placed in an oven and dried at a temperature of 65℃ to 70℃ for 24 to 48 hours. After drying, the dried leaves are ground into powder using a tissue grinder, and about 0.1g of motherwort powder is weighed respectively for total alkaloids (Total alkaloids, TA) and total flavonoids (Total flavonoids, TF) content determination. The total alkaloid content is determined using an alkaloid content determination kit (AKOT007M, Beijing Box Biotechnology Co., Ltd., Beijing), and the total flavonoid content is determined using a plant flavonoid content detection kit (AKPL015M, Beijing Box Biotechnology Co., Ltd., Beijing).

[0045] Data processing and analysis

[0046] Data preprocessing

[0047] The correlation analysis between spectral data and total alkaloid and total flavonoid contents was performed. Considering the high dimensionality and complexity of spectral data and possible noise and scattering problems, standard normal variable transformation (SNV), Savitzky-Golay convolution smoothing (SG), multiplication scatter correction (MSC), first derivative method (1 st Der), second-order derivative method (Second derivative,2 nd Der) five preprocessing methods to optimize the data, and the best preprocessing method was selected. The data set was randomly divided into a calibration set and a validation set in a ratio of 7:3 for model training and evaluation. The distribution of total alkaloids and total flavonoids in the calibration set and validation set samples is shown in Table 1.

[0048] Table 1 Statistical description of the content of total alkaloids and total flavonoids

[0049]

[0050] 1.4.2. Model construction and evaluation

[0051] Partial Least Squares Regression (PLSR) is an efficient multivariate statistical analysis method in the field of chemometrics, especially suitable for dealing with complex spectral data and its correlation analysis with chemical properties. The core of this technology lies in the joint factor analysis of the spectral data matrix X and the target chemical property matrix Y, decomposing these two matrices into several latent variables. Through the cross-validation method, the optimal number of latent variables can be determined to construct a regression model. The advantage of PLSR is that it not only considers the variables in the spectral data X but also the variables in the chemical index data Y, enabling the model to more accurately capture the relationship between the two. In addition, the cross-validation method can effectively evaluate the prediction accuracy of the model and prevent overfitting, thus ensuring the robustness of the model. To further optimize the PLSR model, the method of hyperparameter grid search is adopted. This method systematically traverses various combinations of hyperparameters to find the optimal model parameters. When constructing a quantitative prediction model for the chemical components of Leonurus japonicus, the original spectral data and the preprocessed spectral data (reflectance in the full spectral band of 365 - 970 nm) are used as input feature variables, and the total alkaloid and total flavonoid contents in the Leonurus japonicus leaves are used as target variables. The PLSR algorithm is used to predict the chemical component contents in the Leonurus japonicus leaves under drought stress conditions.

[0052] To evaluate the quality of the constructed prediction model, three parameters, namely the Coefficient of Determination (R 2 ), Root Mean Square Error (RMSE), and Relative Percentage Difference (RPD), are used to evaluate the stability and prediction accuracy of the model. For the calibration set, R 2 c and RMSEc are used as evaluation indicators, and for the validation set, R 2 v, RMSEv, and RPD are used for evaluation. The higher R 2 c and R 2 v are, and the smaller RMSEc and RMSEv are, the higher the inversion accuracy and the better the stability of the model. When RPD > 3, it indicates that the model can make accurate predictions; when 2 < RPD ≤ 3, it indicates that the inversion model can make good predictions; when 1.5 ≤ RPD ≤ 2.0, it means that the model can make rough predictions. Generally speaking, a good model should have high R 2 c, R 2 v, and RPD values, low RMSEc and RMSEv values, and a small difference between RMSEc and RMSEv.

[0053]

[0054]

[0055]

[0056] In the formula, is the predicted value of TA and TF content, is the average value of TA and TF content, y i are the measured values ​​of TA and TF contents, n is the number of samples (i=1,2,3…n), and SD is the standard deviation.

[0057] Model Optimization

[0058] The Bayesian optimization algorithm is a commonly used parameter optimization method that can maintain a relatively fast search speed while taking into account prior knowledge, and this method is not prone to "overfitting". Bayesian optimization is a global optimization algorithm that corrects parameter priors through samples, and uses the Bayesian method when optimizing parameters, taking into account the prior distribution of parameters, and obtaining parameters by averaging. Gaussian process is a common strategy of Bayesian optimization algorithm, which can be used for nonlinear classification, parameter optimization, nonlinear regression, etc.

[0059] In this study, the Bayesian optimization algorithm was used to optimize the PLSR model. The Bayesian optimization method in the Hyperopt library was used to automatically tune the hyperparameters of the PLSR model. The goal was to minimize the mean square error (MSE) of the model on the test set. First, the hyperparameter search space was defined for the PLSR model, which only included the number of latent variables (n_components), and its value was selected from 1 to the minimum value of the number of spectral data features minus 1 and 19. Then, the objective function was written, which built the PLSR model using a given parameter combination, fitted it on the training set, predicted it on the test set, and finally returned the test error MSE. The fmin function of Hyperopt was used to find the best parameter combination that minimized the MSE within 100 evaluations. After finding the best parameters, the code rebuilt and trained the PLSR model to ensure that the final model can achieve the best prediction performance under the best parameters.

[0060] R 4.4.2 software was used for correlation analysis, Unscrambler X10.4 software was used for spectral data preprocessing, and Python 3.12 software was used for PLSR model construction and Bayesian optimization algorithm implementation.

[0061] 2. Results

[0062] 2.1. Spectral characteristics of Leonurus japonicus leaves

[0063] Extract 19 bands of spectral information of motherwort leaf samples in the range of 365-970nm from multispectral images. Figure 1 As shown in the figure, the spectral trends of 156 samples of motherwort leaves are basically consistent, with no significant differences. The average spectral reflectance of leaves under drought stress is lower than that of normal leaves and the difference is larger. This may be due to the change of the internal chemical composition of motherwort leaves after drought stress, which enhances its ability to absorb light. Therefore, the two are separable in the visible multi-spectral range and can be used for the detection of chemical components in motherwort leaves.

[0064] according to Figure 2 The correlation analysis between the total alkaloid (TA) and total flavonoid (TF) content and the spectral data showed that there were significant differences in the correlation between the reflectance of different bands and the TA and TF content (* indicates p<0.05, ** indicates p<0.01, *** indicates p<0.001). The total alkaloid (TA) and total flavonoid (TF) content in certain bands such as 405-690nm, the spectral data and the component content showed a significant correlation, with the highest correlation coefficients of 0.53 and 0.58. The correlation between the reflectance of these bands and the TA and TF content was statistically significant. Therefore, by analyzing the reflectance of specific bands, the content of total alkaloids and total flavonoids in motherwort leaf samples can be effectively predicted.

[0065] 2.2. Analysis of chemical components in Leonurus japonicus

[0066] Drought, rehydration and ABA treatment of motherwortThe changes in the content of total alkaloids and total flavonoids in the leaves of motherwort under different treatment days were determined, and the differences in the content of total alkaloids and total flavonoids between the control group and the experimental group were compared. During the entire treatment process, the content of total alkaloids and total flavonoids in motherwort showed an increasing trend with the increase of stress time. Specifically, the total alkaloid content of the control group (CK) increased to a certain extent during the stress period of 0 to 10 days, and the total flavonoid content showed a trend of first increasing and then decreasing, but remained roughly stable; while the total alkaloids and total flavonoids content of the experimental group (DS) gradually increased with the extension of stress time. After 4 days of stress, this increasing trend was more obvious, indicating that motherwort may adapt to drought conditions by increasing the synthesis of total alkaloids. In general, the total alkaloids and total flavonoids content of the experimental group were always higher than those of the control group, which further confirmed the promoting effect of drought stress on the content of total alkaloids and total flavonoids in motherwort. This response may help plants survive and adapt to drought conditions, especially when the stress time is longer, this adaptive change is more significant.

[0067] 2.3. Model construction and evaluation

[0068] Different spectral preprocessing methods have different effects on the model ( Figures 3-4), this experiment uses partial least squares regression PLSR as the modeling algorithm, and combines the above-mentioned spectral preprocessing method to evaluate the model quality and select the optimal modeling parameters. According to the data analysis of Table 2 and Table 3, it can be seen that different preprocessing methods have a significant impact on the performance of the PLSR model in the quantitative model of total alkaloids and total flavonoids. In the model of total alkaloid components, 1 st The Der preprocessing method showed the highest R 2 The R value (0.8009) of other preprocessing methods on the validation set showed that it had the strongest explanatory power for the model, and also had the lowest RMSEv (0.0009) and a higher RPD value (2.2414), which further confirmed the prediction accuracy and reliability of the model. 2 The lower the value, the weaker the explanatory power of the model. In the model of total flavonoid components, the R 2 The value (0.7575) was the highest and performed best, and the RMSEv (0.0014) and RPD (2.0308) values ​​showed the predictive stability of the model. Other pretreatment methods such as SG and SNV performed poorly in the model of total flavonoid components, especially the R value of SNV pretreatment. 2 The value (0.5757) and RPD value (1.5353) both indicate that the prediction ability of the model is weak.

[0069] In general, due to the advantages of preprocessing methods in reducing data noise and improving signal purity, they can be used as an effective means to optimize the performance of PLSR models. However, for the quantitative analysis of different components, the choice of preprocessing methods should be customized according to the specific data characteristics and model objectives. In this experiment, there was no significant improvement in the performance of the model after preprocessing. In order to further improve the performance of the model, further optimization measures need to be considered, such as adjusting the number of latent variables in the PLSR model, trying different preprocessing methods, or combining with other machine learning algorithms for ensemble learning. Through these methods, the possibility of improving the prediction accuracy of the model can be explored in order to achieve a higher R 2 value, so as to better meet the needs of practical applications.

[0070] Table 2 Quantitative model results of total alkaloid components after different pretreatments

[0071]

[0072] Table 3 Quantitative model results of total flavonoid components after different pretreatments

[0073]

[0074]

[0075] Model Optimization

[0076] Bayesian optimization shows significant advantages in hyperparameter tuning and is a global optimization method. In this study, the Bayesian optimization algorithm was used to optimize the PLSR prediction model. From Tables 4 and 5, it can be seen that after Bayesian optimization, the performance of each treatment on the training set and the validation set has been improved to a certain extent. In the model of total alkaloid components, the modeling using the original spectral data performed well, with the highest R 2 c and R 2 v values ​​were 0.8484 and 0.8369, respectively, which indicates that the model fits the data very well. At the same time, the RMSEc and RMSEv values ​​of this model are the lowest, 0.0008, respectively, indicating that the model prediction error is small. In addition, the RPD value of the constant model reached 2.4764, exceeding 2, indicating that the model has good prediction ability. For the model of total flavonoid components, the original spectral data also performed well, with R 2 c and R 2 The v values ​​are 0.7991 and 0.8060, respectively, showing a high degree of fit. The RMSEc and RMSEv values ​​of the model are also low, 0.0013, respectively, indicating a small prediction error. The RPD value of the constant model is 2.2704, which also exceeds 2, indicating that its prediction ability is good.

[0077] Table 4 Results of the quantitative model of total alkaloid components after Bayesian optimization

[0078]

[0079] Table 5 Results of the quantitative model of total flavonoid components after Bayesian optimization

[0080]

[0081] Overall, for both the total alkaloid and total flavonoid component models, the raw spectral data showed high inversion accuracy and stability, and strong predictive ability. These results show that raw spectral data is a reliable choice as an input variable in the quantitative analysis of both components. Bayesian optimization performs better than traditional hyperparameter optimization methods in hyperparameter tuning because it can more effectively explore the parameter space and find the optimal parameter combination, thereby significantly improving the predictive ability and stability of the model. These results show that Bayesian optimization is an effective means to improve the performance of PLSR models. When processing complex spectral data, it can provide more reliable prediction results for quantitative analysis in practical applications.

[0082] 3. Conclusion

[0083] Based on multispectral imaging technology, this study used partial least squares regression (PLSR) combined with Bayesian optimization to conduct non-destructive detection of total alkaloids and total flavonoids in Leonurus japonicus. The experimental results showed that the original spectral data performed best in model construction, with high fit and low prediction error, and Bayesian optimization further enhanced the prediction ability of the model. The research results provide an effective method for the rapid and accurate detection of chemical components in Leonurus japonicus, which has important application value.

Claims

1. A method for constructing a prediction model for the content of chemical components of Leonurus japonicus based on multispectral imaging technology, characterized in that: The following steps are involved: 1) Extracting the characteristics of Leonurus japonicus: taking a leaf of Leonurus japonicus, imaging it using multispectral imaging technology, extracting the spectral data of the leaf, and then using an alkaloid content determination kit and / or a plant flavonoid content detection kit to determine the content of total alkaloids and / or total flavonoids in the leaf; 2) constructing a model: using the spectral data obtained in step 1) as a characteristic variable, the content of total alkaloids and / or total flavonoids as a target variable, and using a PLSR algorithm for training to obtain a PLSR model; 3) Optimizing the model: The PLSR model obtained in step 2) was optimized using a Bayesian optimization algorithm to obtain a prediction model for the chemical component content of Leonurus japonicus.

2. The method according to claim 1, characterized in that: Step 1) The spectral image imaged by the multi-spectral imaging technology has a pixel size of 2056×2056 and a resolution of 41 μm / pixel.

3. The method according to claim 1, characterized in that: The spectral data is the average reflectivity of the blade in the spectral wavelength range of 365 to 970 nm.

4. The method according to claim 1, characterized in that: Step 2) During the training process using the PLSR algorithm, R 2 , RMSE and RPD are used to evaluate the prediction performance of the model. In the formula, is the predicted value of TA and TF content, is the average value of TA and TF content, y i are the measured values ​​of TA and TF contents, n is the number of samples i=1,2,3…n, and SD is the standard deviation.

5. The method according to claim 1, characterized in that: Step 3) The Bayesian optimization algorithm is derived from the Hyperopt library.

6. The method according to any one of claims 1 to 5, characterized in that: The model for predicting the content of chemical components of Leonurus japonicus predicts the content of total alkaloids and total flavonoids in the leaves of Leonurus japonicus.

7. The prediction model for the chemical component content of Leonurus japonicus constructed by the method described in any one of claims 1 to 6.

8. A method for detecting the content of chemical components in Leonurus japonicus, characterized in that: ① Perform multispectral imaging on the motherwort to be tested, and extract the average reflectivity of the motherwort leaves in the spectral wavelength range of 365 to 970 nm; ② Input the average reflectance obtained in step ① into the prediction model described in claim 7 to obtain the prediction result of the chemical composition of Leonurus japonicus; The chemical components of Leonurus japonicus are total alkaloids and / or total flavonoids.

9. A system for implementing the method for detecting the content of chemical components in Leonurus japonicus according to claim 8, characterized in that: include An input module, used for inputting the characteristics of Leonurus japonicus to be tested; A computing module, comprising the prediction model according to claim 7; An output module, used for outputting the final prediction results of the chemical components of Leonurus japonicus; The chemical components of Leonurus japonicus are total alkaloids and / or total flavonoids.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and the computer program is used to implement the method for detecting the content of chemical components in motherwort as described in claim 8.

Images