Near infrared spectrum rapid identification method and system for different meconopsis plants

Abstract

The invention discloses a near infrared spectrum rapid identification method and system for different meconopsis plants, and belongs to the field of species identification, and the method mainly comprises the following steps: collecting different meconopsis plant samples; collecting near-infrared integrating sphere diffuse reflection spectrums and near-infrared optical fiber probe spectrums of different meconopsis plant samples; analyzing near-infrared integrating sphere diffuse reflection spectrum characteristics and near-infrared optical fiber probe spectrum characteristics between different meconopsis plant samples; establishing an integrating sphere diffuse reflection spectrum discrimination model and an optical fiber probe spectrum discrimination model of different meconopsis plant samples based on the near-infrared integrating sphere diffuse reflection spectrum; comparing the prediction effects of the integrating sphere diffuse reflection spectrum discrimination model and the optical fiber probe spectrum discrimination model of different meconopsis plant samples; and comparing the prediction effects of the integrating sphere diffuse reflection spectrum discrimination model and the optical fiber probe spectrum discrimination model, and selecting the discrimination model with the best prediction effect. According to the method, rapid and accurate identification of the six meconopsis plants can be achieved based on the near infrared spectrum analysis technology, the constructed species identification model is good in effect, and a scientific basis can be provided for guaranteeing the safety and effectiveness of clinical medication of meconopsis medicinal materials.

Description

Technical Field

[0001] The invention relates to the field of species identification, and in particular to a near-infrared spectrum rapid identification method and system for different species of Meconopsis plants. Background Art

[0002] Meconopsis plants are annual or perennial herbs, wild alpine flowers, known as "alpine peony", and praised by Europeans as "world famous flowers". In addition to its ornamental value, Meconopsis plants are also traditional Tibetan medicines, with the effects of clearing away heat and detoxification, relieving cough and asthma, anti-inflammatory and analgesic, and invigorating the middle and replenishing qi. Modern pharmacological studies have shown that Meconopsis has analgesic, antidiarrheal, sedative, anti-fatigue, and liver-protecting effects. Meconopsis is used as medicine in its entirety or in its flowers, and has the effects of promoting blood circulation and dispersing blood stasis, and treating hepatitis and pneumonia.

[0003] Due to the large number of species of Meconopsis and the differences in Tibetan medicine usage habits, there are also differences in their functional efficacy in clinical practice. Among them, Meconopsis integrinus is used as medicine with the whole plant, which has the effect of clearing heat and detoxifying, and can be used to treat hepatitis, pneumonia, lung heat cough, headache, skin diseases, etc. Meconopsis quinqueneri is cool in nature, sweet and astringent in taste, and has the effects of clearing heat and detoxifying, diuresis, anti-inflammatory and analgesic, and can be used to treat hepatitis, pneumonia, headache, edema, skin diseases, liver and lung heat syndrome, etc. Meconopsis safflower can be used to treat diseases such as fever and headache caused by hypertension, and has the effects of dehumidification and diuresis, asthma and cough, and dysentery. The root of Meconopsis racemosa has the effects of tonifying the middle and replenishing qi, relieving cough and asthma, stopping dysentery, and sedation, and the whole plant has analgesic effect. Meconopsis multi-thorn has the effects of clearing heat and relieving pain, promoting blood circulation and removing blood stasis, etc., and can be used to treat fractures, traumatic injuries, chest and back pain, etc., and also has certain anti-tumor, anti-viral, myocardial protection and other effects. Therefore, different species of Meconopsis have different clinical effects, and accurate identification of their species is of great significance for clinical medication according to different diseases. In order to standardize the safety of clinical medication and improve its quality standards, it is necessary to accurately analyze and identify different species of Meconopsis to ensure the accuracy and safety of medication from the source.

[0004] Since Tibetan medicines on the market are often only taken from certain parts of the plant, and the medicinal materials will inevitably have a certain degree of collision and damage during transportation, resulting in the loss of some key classification characteristics, the traditional morphological identification method can no longer meet the identification needs. Existing identification methods for Meconopsis plants, such as high-performance liquid chromatography, thin-layer chromatography, mass spectrometry and DNA barcoding technology, are prone to damage to samples, time-consuming, inefficient and high in measurement cost. Therefore, it is urgent to develop a fast and accurate method for distinguishing species of Meconopsis plants. Summary of the invention

[0005] The purpose of the present invention is to overcome the technical problems existing in the prior art and to provide a near-infrared spectrum rapid identification method and system for different species of Meconopsis plants.

[0006] The objective of the present invention is achieved through the following technical solutions:

[0007] In a first aspect, a method for rapid identification of different species of Meconopsis plants by near infrared spectroscopy is provided, the method comprising the following steps:

[0008] S1. Collect samples of different species of Meconopsis;

[0009] S2, collecting near-infrared integrating sphere diffuse reflectance spectra of different species of Meconopsis plant samples, and collecting near-infrared fiber optic probe spectra of different species of Meconopsis plant samples;

[0010] S3. Analyze the near-infrared integrating sphere diffuse reflectance spectral characteristics and near-infrared fiber optic probe spectral characteristics of different species of Meconopsis plant samples;

[0011] S4. Establish an integrating sphere diffuse reflectance spectrum discrimination model for different species of Meconopsis plant samples based on near-infrared integrating sphere diffuse reflectance spectrum; establish an optical fiber probe spectrum discrimination model for different species of Meconopsis plant samples based on near-infrared optical fiber probe spectrum;

[0012] S5. Compare the prediction effects of the integrating sphere diffuse reflectance spectral discrimination model of different species of Meconopsis plant samples; compare the prediction effects of the optical fiber probe spectral discrimination model of different species of Meconopsis plant samples;

[0013] S6. Compare the prediction effects of the integrating sphere diffuse reflectance spectrum discrimination model and the optical fiber probe spectrum discrimination model, and select the discrimination model with the best prediction effect.

[0014] In some embodiments, the different species of Meconopsis plant samples include Meconopsis spinulosa, Meconopsis erythrophylla, Meconopsis racemosa, Meconopsis hybrid, Meconopsis integrifolia and Meconopsis quinqueneri, wherein the hybrid Meconopsis is a hybrid of Meconopsis quinqueneri and Meconopsis erythrophylla.

[0015] In some embodiments, the method of collecting near-infrared integrating sphere diffuse reflectance spectra of different species of Meconopsis plant samples includes:

[0016] Take a proper amount of the whole herb sample of Meconopsis thunbergii after passing through a 200-mesh sieve and place it in a sample cup, flatten it, and compact it. The sample thickness is 0.1 cm. During the collection process, the background interference of CO2 and water is deducted in real time. The scanning resolution is 8 cm -1 , scan times 64 times.

[0017] In some embodiments, the method of collecting near-infrared fiber optic probe spectra of different species of Meconopsis plant samples includes:

[0018] The near-infrared fiber probe spectrum of the whole herb sample of Meconopsis thunbergii after passing through a 200-mesh sieve was collected using the fiber probe accessory of the Fourier transform infrared spectrometer. The background interference of CO2 and water was deducted in real time during the collection process. The scanning resolution was 4 cm -1 , scan times 64 times.

[0019] In some embodiments, the analysis of near-infrared integrating sphere diffuse reflectance spectral characteristics and near-infrared fiber optic probe spectral characteristics between different species of Meconopsis plant samples includes:

[0020] The near-infrared one-dimensional spectra of the samples were characterized by analysis, and the common peaks were determined with the missing rate less than 5% as an indicator. The characteristic peaks were attributed, and the one-dimensional spectra were processed by first-order derivative and second-order derivative, and the amplification of the differences between the peaks of the first-order derivative spectrum and the second-order derivative spectrum were analyzed. At the same time, the three types of spectra were analyzed for similarity using the correlation coefficient as an indicator. Finally, the absorbance values ​​at the common peaks were subjected to one-way analysis of variance using SPSS software to determine whether there were differences between different species of Meconopsis.

[0021] In some embodiments, the method further comprises:

[0022] Cluster analysis was performed using near-infrared one-dimensional spectra of different species of Meconopsis plant samples to obtain different cluster groups.

[0023] In some embodiments, the method of establishing an integrating sphere diffuse reflectance spectroscopy discrimination model for different species of Meconopsis plant samples based on near-infrared integrating sphere diffuse reflectance spectroscopy includes:

[0024] The best modeling condition is DM+MSC+D1+unsmoothed, and the modeling band is 9881.46-4119.21cm -1 .

[0025] In some embodiments, the method of establishing a fiber optic probe spectrum discrimination model for different species of Meconopsis plant samples based on the near-infrared fiber optic probe spectrum includes:

[0026] The best modeling conditions are: DM+SNV+D1+unsmoothed, and the modeling band is 9885.32-4364.12cm -1 .

[0027] In some embodiments, comparing the prediction effects of the integrating sphere diffuse reflectance spectroscopy discrimination model of different species of Meconopsis plant samples comprises:

[0028] The discrimination accuracy of the integrating sphere diffuse reflectance spectroscopy discrimination model of the six species of Meconopsis plants is from high to low: Meconopsis quinquenervii > Meconopsis integrinus > Meconopsis racemosa > Meconopsis hybridus > Meconopsis roxburghii > Meconopsis spicata;

[0029] The method of comparing the prediction effects of the optical fiber probe spectral discrimination model of different species of Meconopsis plant samples includes:

[0030] The discrimination accuracy of the fiber optic probe spectral discrimination model for the six species of Meconopsis plants is as follows from high to low: Meconopsis integrifolia > Meconopsis hybrida > Meconopsis racemosa > Meconopsis quinqueneri > Meconopsis roxburghii > Meconopsis spicata.

[0031] In a second aspect, a near infrared spectroscopy rapid identification system for different species of Meconopsis plants is provided, the system comprising:

[0032] A sample collection module, used to collect samples of different species of Meconopsis;

[0033] A spectrum acquisition module, used for acquiring near-infrared integrating sphere diffuse reflectance spectra of different species of Meconopsis plant samples, and acquiring near-infrared optical fiber probe spectra of different species of Meconopsis plant samples;

[0034] Spectral feature analysis module, used to analyze near-infrared integrating sphere diffuse reflectance spectral features and near-infrared fiber optic probe spectral features between different species of Meconopsis plant samples;

[0035] A discrimination model building module is used to build an integrating sphere diffuse reflectance spectrum discrimination model for different species of Meconopsis plant samples based on near-infrared integrating sphere diffuse reflectance spectrum; and to build a fiber optic probe spectrum discrimination model for different species of Meconopsis plant samples based on near-infrared fiber optic probe spectrum;

[0036] The discrimination model comparison module is used to compare the prediction effects of the integrating sphere diffuse reflectance spectrum discrimination model of different species of Meconopsis plant samples; and to compare the prediction effects of the optical fiber probe spectrum discrimination model of different species of Meconopsis plant samples;

[0037] The discrimination model selection module is used to compare the prediction effects of the integrating sphere diffuse reflectance spectrum discrimination model and the optical fiber probe spectrum discrimination model, and select the discrimination model with the best prediction effect.

[0038] It should be further explained that the technical features corresponding to the above options can be combined or replaced with each other to form a new technical solution if there is no conflict.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The present invention takes six common Meconopsis plants as research objects, utilizes near-infrared spectroscopy technology combined with diversified chemometrics methods, carries out species discrimination research of different Meconopsis plants, provides new research ideas and methods for species discrimination of Meconopsis plants, can effectively guarantee the safety and effectiveness of clinical use of Meconopsis medicinal materials, and is conducive to ensuring the circulation and supervision of Meconopsis medicinal materials in the market, further reduces the cost of medicinal material discrimination, shortens the time of medicinal material discrimination, and also provides reference and thinking for the research of other Tibetan medicinal material resource discrimination methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flowchart of a method for rapid identification of different species of Meconopsis plants by near infrared spectroscopy is shown in an embodiment of the present invention;

[0042] Figure 2 A one-dimensional near-infrared integrating sphere diffuse reflectance spectrum of a Meconopsis spinosa sample shown in an embodiment of the present invention;

[0043] Figure 3 A one-dimensional near-infrared integrating sphere diffuse reflectance spectrum of a Meconopsis erythrorhizome sample shown in an embodiment of the present invention;

[0044] Figure 4 A one-dimensional near-infrared integrating sphere diffuse reflectance spectrum of a Meconopsis integrifolia sample shown in an embodiment of the present invention;

[0045] Figure 5 A one-dimensional near-infrared integrating sphere diffuse reflectance spectrum of a Meconopsis quinqueneri sample shown in an embodiment of the present invention;

[0046] Figure 6 A one-dimensional near-infrared integrating sphere diffuse reflectance spectrum of a hybrid Meconopsis sample shown in an embodiment of the present invention;

[0047] Figure 7 A one-dimensional near-infrared integrating sphere diffuse reflectance spectrum of a Meconopsis racemosa sample shown in an embodiment of the present invention;

[0048] Figure 8 A one-dimensional near-infrared optical fiber probe spectrum of a Meconopsis spinosa sample shown in an embodiment of the present invention;

[0049] Fig. 9 A one-dimensional near-infrared optical fiber probe spectrum of a Meconopsis erythrorhizome sample shown in an embodiment of the present invention;

[0050] Fig.10 A one-dimensional near-infrared optical fiber probe spectrum of a Meconopsis integrifolia sample shown in an embodiment of the present invention;

[0051] Fig.11 A one-dimensional near-infrared optical fiber probe spectrum of a Meconopsis quinquenervata sample shown in an embodiment of the present invention;

[0052] Fig.12 A one-dimensional near-infrared optical fiber probe spectrum of a hybrid Meconopsis sample shown in an embodiment of the present invention;

[0053] Fig.13 A one-dimensional near-infrared optical fiber probe spectrum of a Meconopsis racemosa sample shown in an embodiment of the present invention;

[0054] Fig.14 A cluster analysis tree diagram shown in an embodiment of the present invention;

[0055] Fig.15 A 3D diagram of a one-dimensional near-infrared spectrum model shown in an embodiment of the present invention;

[0056] Fig.16 A classification diagram of a one-dimensional near-infrared integrating sphere diffuse reflectance spectroscopy model based on Python software shown in an embodiment of the present invention;

[0057] Fig.17 This is a classification diagram of a one-dimensional near-infrared fiber optic probe spectrum model based on Python software according to an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The technical solution of the present invention is clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present application described and shown in the accompanying drawings can be arranged and designed in various configurations. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0059] It should be noted that the defects existing in the solutions in the above-mentioned prior art are the results obtained by the inventor after practice and careful research. Therefore, the discovery process of the above-mentioned problems and the solutions proposed in the embodiments of the present application for the above-mentioned problems below should all be the contributions made by the inventor to the present application in the process of invention and creation, and should not be understood as technical contents known to technical personnel in this field.

[0060] Example 1

[0061] In an exemplary embodiment, a near infrared spectroscopy rapid identification method of different species of Meconopsis plants is provided, referring to Figure 1 , the method comprises the following steps:

[0062] S1. Collect samples of different species of Meconopsis;

[0063] S2, collecting near-infrared integrating sphere diffuse reflectance spectra of different species of Meconopsis plant samples, and collecting near-infrared fiber optic probe spectra of different species of Meconopsis plant samples;

[0064] S3. Analyze the near-infrared integrating sphere diffuse reflectance spectral characteristics and near-infrared fiber optic probe spectral characteristics of different species of Meconopsis plant samples;

[0065] S4. Establish an integrating sphere diffuse reflectance spectrum discrimination model for different species of Meconopsis plant samples based on near-infrared integrating sphere diffuse reflectance spectrum; establish an optical fiber probe spectrum discrimination model for different species of Meconopsis plant samples based on near-infrared optical fiber probe spectrum;

[0066] S5. Compare the prediction effects of the integrating sphere diffuse reflectance spectral discrimination model of different species of Meconopsis plant samples; compare the prediction effects of the optical fiber probe spectral discrimination model of different species of Meconopsis plant samples;

[0067] S6. Compare the prediction effects of the integrating sphere diffuse reflectance spectrum discrimination model and the optical fiber probe spectrum discrimination model, and select the discrimination model with the best prediction effect.

[0068] Specifically, the present invention uses the whole grass of six common Meconopsis plants as the research object, uses near-infrared spectroscopy technology combined with multivariate chemometrics methods to carry out species discrimination research on different species of Meconopsis plants, and analyzes their main near-infrared spectral characteristics by collecting their near-infrared spectra. On this basis, combined with multivariate chemometrics methods, Meconopsis species discrimination models based on two different spectral collection methods, namely, near-infrared integrating sphere diffuse reflectance-sample cup and near-infrared fiber optic probe, are constructed, respectively, which provides new research ideas and methods for species discrimination of Meconopsis plants, can effectively ensure the safety and effectiveness of clinical use of Meconopsis medicinal materials, and is conducive to ensuring the circulation and supervision of Meconopsis medicinal materials in the market, further reducing the cost of medicinal material discrimination, shortening the time of medicinal material discrimination, and also providing reference and thinking for the research of other Tibetan medicinal material resource discrimination methods.

[0069] Example 2

[0070] Based on the inventive concept of Example 1, this example provides a specific identification process, which mainly includes the following parts:

[0071] 1. Instruments and Materials

[0072] 1.1 Sample source

[0073] In the Qinghai Plateau, according to different habitat types and altitude gradients, 1,735 samples of 6 species of Meconopsis were collected from 48 populations / origins during the Meconopsis flowering period (generally from June to July). The samples collected in the field were washed and dried, dried in an oven at 45°C, crushed, sieved through a 200-mesh sieve, and stored in a desiccator for later use. The original plant specimens were identified by researchers from the Northwest Plateau Institute of Biology, Chinese Academy of Sciences as Meconopsis horridula Hook.f.et Thoms., Meconopsis punicea Maxim., Meconopsis racemosa Maxim., Meconopsis×cookei G.Taylor., Meconopsis integrifolia (Maxim.) Franch., and Meconopsis quintuplinervia Regel., among which Meconopsis quintuplinervia Regel is a hybrid of Meconopsis quintuplinervia Regel and Meconopsis punicea Maxim. The specific information of the samples is shown in Table 1.

[0074] Table 1 Sample collection information of six species of Meconopsis

[0075]

[0076]

[0077] 1.2 Instruments

[0078] Thermo Fisher iS50 Fourier transform infrared spectrometer (Thermo Fisher, USA) equipped with near-infrared integrating sphere diffuse reflectance and near-infrared fiber optic probe accessories; drying oven (DHG9245A, Shanghai Yiheng Technology Co., Ltd., China); 200-mesh sample sieve; pulverizer (Tianjin Test Co., Ltd., China); desiccator.

[0079] 2. Experimental methods

[0080] 2.1 Near infrared spectrum acquisition

[0081] 2.1.1 Near-infrared integrating sphere diffuse reflectance spectrum collection

[0082] Take an appropriate amount of the six species of Meconopsis whole herb samples that have been crushed and passed through a 200-mesh sieve and place them in a sample cup, flatten them, and compact them. The sample thickness is 0.1 cm. During the collection process, the background interference of CO2 and water is deducted in real time, and the scanning resolution is 8 cm -1, 64 scans. Preheat the instrument for at least 30 minutes before spectrum acquisition, and use it after the instrument is stable. Keep the room temperature at around 25°C and the humidity below 35%. Use the air as a reference at 10000-4000cm -1 The NIR spectrum was scanned within a range (n=3), and the average spectrum was taken for analysis.

[0083] 2.1.2 Near-infrared fiber optic probe spectrum acquisition

[0084] The whole herb sample of Meconopsis thunbergii was sieved through a 200-mesh sieve and the near-infrared fiber optic probe was used to collect the spectrum. The background interference of CO2 and water was deducted in real time during the collection process. The scanning resolution was 4 cm -1 , 64 scans. The instrument was preheated for at least 30 minutes before spectrum acquisition and used after stabilization. The room temperature was maintained at around 25°C and the humidity was controlled below 35%. With air as the reference, the instrument was set at 10000-4000cm -1 Scan the NIR spectrum within the range, scan the front and back sides 3 times each, and take the average spectrum for analysis.

[0085] 2.2 Cluster analysis

[0086] IBM SPSS Statistics25 statistical analysis software was used to perform cluster analysis on the near-infrared one-dimensional spectra of the samples, and the similarities and differences between the clustering results and the actual classification were compared.

[0087] 2.3 Sample set division

[0088] 80% (1388 samples) of the 1735 samples were randomly selected as the modeling set for model establishment, and the other 20% (347 samples) were used as a completely independent external validation set to verify the applicability of the model. The 1388 modeling samples were randomly divided into a calibration set and a prediction set, of which 2 / 3 were the calibration set and 1 / 3 were the prediction set.

[0089] 2.4 Analysis of near infrared spectral characteristics

[0090] Normally, peak position, peak intensity and peak shape are the three major elements of spectrum analysis. Therefore, the near-infrared one-dimensional spectra of the samples were first analyzed to determine the characteristic peaks of the six Meconopsis samples, and then the missing rate was less than 5% as an indicator to determine the common peaks and attribute the characteristic peaks. Then the one-dimensional spectrum was processed by first-order derivative and second-order derivative, and the difference between the peaks of the first-order derivative spectrum and the second-order derivative spectrum was analyzed. At the same time, the correlation coefficient was used as an indicator to perform similarity analysis on the three types of spectra, and finally the absorbance values ​​at the common peaks were analyzed by one-way variance analysis using SPSS software to determine whether there were differences between different species of Meconopsis plants.

[0091] 2.5 Model building

[0092] 2.5.1 Modeling based on TQ Analyst software

[0093] The preprocessing and modeling methods provided by TQ Analyst software were used to model the data, and a 4-factor 3-level table (Table 2) was designed to select the best modeling conditions. Among them, the classification model used two modeling methods: distance matching (DM) and discriminant analysis (DA). In order to correct factors such as random noise in the spectrum, length changes in the light direction, and light scattering generated by the instrument, a variety of spectral preprocessing methods were used in modeling, including first derivative (FD), second derivative (SD), multiplicative scattering correction (MSC), standard normalvariate correction (SNV), Savitzky-Golay filter smoothing (SG) and Norris smoothing.

[0094] Table 2 TQ Analyst software modeling 4 factors 3 levels table

[0095]

[0096] 2.5.2 Modeling based on Python software

[0097] Python software was used to combine four classification methods, namely SVM, DT, RF and ET, to establish models for the spectra of different species of Meconopsis samples. The spectrum preprocessing adopted the combination of “Correction” + “Derivation” + “Smoothing”.

[0098] 2.6 Model Performance Evaluation

[0099] The recognition rate and prediction rate were used as indicators to evaluate the performance of the model. The calculation formulas for the accuracy of the training set and the calibration set were: recognition rate = (total number of calibration sets - number of calibration set misjudgments) / total number of calibration sets × 100%, prediction rate = (total number of prediction sets - number of prediction set misjudgments) / total number of prediction sets × 100%. The applicability of the model was verified using the near-infrared spectroscopy data of the samples in the external validation set. The external validation set recognition rate was used as an indicator for evaluation, and its calculation formula was: external validation set recognition rate = total number of verified identifications / total number of verified × 100%.

[0100] In addition to the above recognition rate and prediction rate as indicators, the model performance recognition evaluation also uses the error rate (ER), true positive rate (TPR), true negative rate (TNR) and F1 value for evaluation, where ER refers to the proportion of samples with classification prediction errors; TPR, also known as sensitivity, refers to the proportion of correctly identified positive samples; TNR refers to the proportion of correctly identified negative samples. The higher the TPR, TNR and F1 value, the lower the ER, and the better the performance of the classification model. The calculation formulas for each parameter are as follows:

[0101] ER=(FP+FN) / n

[0102] TPR=TP / (TP+FN)

[0103] TNR=TN / (FP+TN)

[0104] F1=2×TP / (2×TP+FP+FN)

[0105] In the formula, the True Positive (TP) class indicates that the true sample is identified as the true sample; the False Positive (FP) class indicates that the true sample is identified as the pseudo sample; the True Negative (TN) class indicates that the pseudo sample is identified as the pseudo sample; the False Negative (FN) class indicates that the pseudo sample is identified as the true sample, and n is the total sample size.

[0106] 2.7 External Validation

[0107] After the species discrimination model was established using the optimized modeling conditions, the near-infrared spectra of 347 randomly selected external validation set samples were substituted into the model for verification, and the external validation set recognition rate was used as an indicator for evaluation. The external validation set recognition rate of the model was calculated as follows:

[0108] External validation set recognition rate = total number of validation recognitions / total number of validations

[0109] In the formula, the total number of verification identification is the number of samples that have passed the verification and the model prediction results are consistent with the actual classification results; the total number of verification is the total number of samples involved in the model verification.

[0110] 3. Results and analysis

[0111] 3.1 Near infrared spectral feature analysis

[0112] 3.1.1 Near-infrared integrating sphere diffuse reflection

[0113] Figure 2-Figure 7The one-dimensional near-infrared spectra of six species of Meconopsis plants under the conditions of integrating sphere diffuse reflectance collection. It can be seen that the one-dimensional near-infrared spectra of the same species of Meconopsis samples have the same trend and there are no abnormal samples, indicating that although the origins of the same plants are not exactly the same, the main compound components contained are basically the same and can be used for modeling. The one-dimensional near-infrared spectra of different species of Meconopsis samples are also similar and difficult to distinguish from a macroscopic perspective, so further analysis is needed. Table 3 gives the one-dimensional near-infrared integrating sphere diffuse reflectance average spectral data of six Meconopsis plants. In the range of 9090-6250cm -1 Mid-wave near infrared and 6250~4000cm -1 The peak positions of different species of Meconopsis in the two long-wave near-infrared bands are different. In addition, only Meconopsis spinulosa and Meconopsis racemosa have peaks at 8401 cm -1 There is absorption at the same position, while the other four species of Meconopsis lack this absorption. It is speculated that the two species are distantly related to the other species, and Meconopsis racemosa is a variant of Meconopsis spinosa, which can also be used to distinguish the other four species of Meconopsis. The six species of Meconopsis have three common peaks, namely 5778cm -1 、5171cm -1 and 4255cm -1 ; There are 6 characteristic peaks, among which 8401cm -1 It is the second harmonic absorption band of unsaturated CH stretching vibration, indicating that the sample may contain olefin or benzene ring structure, 6790cm -1 It is the second harmonic absorption band of the stretching vibration of alcohol hydroxyl group OH, 5778cm -1 It is the first harmonic absorption band of the asymmetric stretching vibration of methyl CH, 5171cm -1 It is the combined absorption band of water hydroxyl group OH, 4680cm -1 It is the first harmonic absorption band of the stretching vibration of alcohol hydroxyl group OH, 4255cm -1 It is the combined absorption band of methyl CH and methylene CH, indicating that the sample contains alkyl or cycloalkyl groups.

[0114] Table 3 One-dimensional near-infrared integrating sphere diffuse reflectance average spectral data of six species of Meconopsis

[0115]

[0116] 3.1.2 Near-infrared fiber optic probe

[0117] The one-dimensional near-infrared spectra of six species of Meconopsis obtained under the conditions of optical fiber probe collection are shown in the figure below. Figure 8-13As shown, it is consistent with the spectrum shape result obtained under the integrating sphere diffuse reflectance acquisition conditions under item "3.1.1", indicating that the same species of Meconopsis samples have one-dimensional near-infrared spectra with consistent trends, and there are no abnormal samples. Comparing the average spectral data of the one-dimensional near-infrared spectra of the six Meconopsis samples (Table 4), the results show that, similar to the average one-dimensional near-infrared integrating sphere diffuse reflectance spectra of the six Meconopsis species, there are differences in the peak positions of the one-dimensional near-infrared spectra between different varieties of Meconopsis samples. In addition, as shown in Table 4, compared with the one-dimensional near-infrared integrating sphere diffuse reflectance one-dimensional spectrum under item "3.1.1", the one-dimensional spectrum of the near-infrared fiber optic probe has more absorption peaks, and the spectra of the six Meconopsis plants all have absorption at the same wave number, indicating that the spectrum obtained under the collection conditions of the near-infrared fiber optic probe expresses more comprehensive information. 5776cm -1 5664cm -1 、5172cm -1 The three peaks are common peaks of the six species of Meconopsis plants. The sample has seven characteristic peaks, among which 8250cm -1 It is the double frequency absorption band of unsaturated CH stretching vibration, indicating that the sample contains olefin or benzene ring structure. 6829cm-1 is the first harmonic absorption band of alcohol OH stretching vibration, 5776cm -1 and 5664cm -1 It is the first harmonic absorption band of the asymmetric stretching vibration of methyl CH, 5172cm -1 It is the combined absorption band of water hydroxyl group OH, 4748cm -1 It is the first harmonic absorption band of the stretching vibration of alcohol hydroxyl group OH, 4322cm -1 It is the combined frequency absorption band of methyl CH.

[0118] Table 4 Average spectral data of one-dimensional near-infrared fiber probe of six species of Meconopsis

[0119]

[0120]

[0121] 3.2 Cluster analysis

[0122] The one-dimensional near-infrared spectra of different varieties of Meconopsis thunbergii from 48 populations were analyzed using the hierarchical clustering method in IBM SPSS Statistics 25 software. The dendrograms generated by cluster analysis were constructed using different linkage methods. Fig.14 Among them, samples 1-14 are Meconopsis integrifolia, samples 15-19 are Meconopsis quinqueneri, samples 20-23 are Meconopsis spinulosa, samples 24-31 are Meconopsis rubra, samples 32 are Meconopsis hybrida, and samples 33-48 are Meconopsis racemose. Fig.14A is the clustering result of one-dimensional near-infrared integrating sphere diffuse reflectance spectrum, and the results show that the samples of 48 populations are clustered into three categories, namely, 1, 8, 9, 16, 23, 28, 35, 38, 40, 42-47 are clustered into one category, 2, 15, 17-22, 24-27, 29-34, 36-37, 39, 41, 48 are clustered into one category, and 3-7, 10-14 are clustered into one category, indicating that cluster analysis can realize the species division of Meconopsis to a certain extent. Compared with the actual classification results, except for the classification of the entire leaf Meconopsis sample, there is a certain degree of confusion in the other varieties of Meconopsis. Fig.14 B is the clustering result of the one-dimensional near-infrared fiber probe spectrum. The results show that the 48 population samples are generally clustered into four categories, namely, 1-14 is clustered into one category, 15-19, 23, 26-31, 38, 42-48 is clustered into one category, 20-22, 25, 32, 34-35, 37, 39-41 is clustered into one category, and 24, 33, and 36 are clustered into one category. Compared with the actual classification results, except for the entire leaf Meconopsis, the other varieties of Meconopsis are partially confused, which is similar to the clustering result of the one-dimensional near-infrared integrating sphere diffuse reflectance spectrum. The classification results of the six Meconopsis samples are different from the actual classification results to a certain extent. In summary, the classification of the six Meconopsis plants cannot be accurately achieved through cluster analysis. Therefore, it is particularly important to establish a near-infrared species discrimination model for Meconopsis plants.

[0123] 3.3Discrimination model under TQ Analyst software conditions

[0124] 3.3.1 Selection of modeling conditions

[0125] The near infrared spectral data of the samples collected by near infrared integrating sphere diffuse reflectance and fiber optic probe were imported into TQ analyst software to optimize the modeling method and spectrum preprocessing method. The recognition rate and prediction rate of the model were used as indicators to screen the best combination of modeling conditions. The optimization results of modeling conditions are shown in Table 5. The best modeling condition combination of near infrared integrating sphere diffuse reflectance is DM+MSC+FD+unsmoothing. At this time, the recognition rate of the model is 87.37%, the prediction rate is 87.88%, and the external verification recognition rate is 82.71%; the best modeling condition combination of near infrared fiber optic probe is DM+SNV+FD+unsmoothing. At this time, the recognition rate of the model is 97.30%, the prediction rate is 93.29%, and the external verification recognition rate is 91.93%. The discrimination effect of the latter model is better than that of the former.

[0126] Table 5 Combination of modeling conditions in TQ Analyst software

[0127]

[0128] 3.3.2 Model establishment

[0129] After modeling based on the optimal spectral preprocessing, the 3D graph of the model is obtained ( Fig.15 ). The results showed that samples of the same species clustered in the same area of ​​space, while samples of different species were separated from each other, indicating that different species of Meconopsis can be distinguished from each other. Fig.15 A is the one-dimensional near-infrared integrating sphere diffuse reflectance spectrum Figure 3 D model, the recognition rate and prediction rate of the model for Meconopsis quinqueneri were the highest, both reaching 100%, and the recognition rate and prediction rate for Meconopsis multi-spined were the lowest, which were 33.33% and 0, respectively. Furthermore, the species discrimination model of Meconopsis plants was established using the optimized modeling conditions, and the ER value, TPR value, TNR value and F1 value of the model were calculated. The discrimination accuracy of the model for the six Meconopsis plants was from high to low as follows: Meconopsis quinqueneri > Meconopsis entire leaf > Meconopsis racemose > Meconopsis hybrid > Meconopsis roxburghii > Meconopsis multi-spined.

[0130] Fig.15 B is a one-dimensional near-infrared fiber optic probe 3D model. The model has high recognition and prediction rates for six species of Meconopsis. Among them, the recognition rates of Meconopsis integrinus, Meconopsis quinqueneri, Meconopsis roxburghii, Meconopsis hybridis and Meconopsis racemosa are all higher than 90%, and the prediction rates are all higher than 85%. Although the recognition rate of Meconopsis spinosus reaches 100%, the prediction rate is 0, indicating that the model is overfitting. Except for Meconopsis spinosus, the model has high discrimination accuracy for the other five species of Meconopsis. Further, the species discrimination model of Meconopsis was established using the optimized modeling conditions, and the ER value, TPR value, TNR value and F1 value of the model were calculated. The discrimination accuracy of the model for the six species of Meconopsis is from high to low: Meconopsis integrinus > Meconopsis hybridis > Meconopsis racemosa > Meconopsis quinqueneri > Meconopsis roxburghii > Meconopsis spinosus.

[0131] The 3D model discrimination results under the two collection methods both showed that the discrimination effect of Meconopsis spinosa was relatively low. This was due to the small sample size of Meconopsis spinosa. If the number of samples increased, the model accuracy and fitting would also improve.

[0132] 3.4 Discriminant model under Python software conditions

[0133] 3.4.1 Modeling Condition Selection

[0134] The performance of SVM, DT, RF, ET and Voting fusion models were evaluated by recognition rate, prediction rate, total correct rate and external verification recognition rate, and the results are shown in Table 6. The results show that the performance of the models established using Python is good. For the near-infrared integrating sphere diffuse reflectance spectrum acquisition condition, the SVM model has the best performance, with a recognition rate of 97.95% and a prediction rate of 97.82%; under the near-infrared fiber optic probe acquisition condition, the SVM model also has the best performance, with a recognition rate of 99.68.%, and a prediction rate of 98.47%. The latter is better than the former.

[0135] Table 6 Comparison of model performance under different spectral acquisition methods

[0136]

[0137]

[0138] 3.4.2 Model establishment

[0139] Fig.16 This is the effect diagram of the classification model established by using Python software for the spectrum collected under the condition of near-infrared integrating sphere diffuse reflectance. Fig.16 A is the decision tree (DT) classification model, Fig.16 B is the Extreme Tree (ET) classification model, Fig.16 C is the random forest (RF) classification model, Fig.16 is the support vector machine (SVM) classification model, Fig.16 E is the effect diagram after the above four classification models are fused based on voting. Voting is a simple and effective data fusion method that directly votes or averages the results predicted by different models without changing the model. It can be seen from the figure that the SVM classification model has the best effect. Among them, Meconopsis integrinus, Meconopsis rubra, Meconopsis quinquenervata and Meconopsis racemosa are accurately predicted (100%), while the prediction rate of Meconopsis hybrida is poor, only 62%, and most of Meconopsis spinosa is identified as Meconopsis racemosa, with a prediction rate of 0%. The fusion models of DT, ET, RF, SVM and Voting failed to correctly predict Meconopsis spinosa, which is due to the small number of Meconopsis spinosa samples.

[0140] The spectral data of six species of Meconopsis plants collected under the conditions of near-external fiber probes were modeled using Python software. The model effect diagram is shown in Fig.17 .in, Fig.17 A is the decision tree (DT) classification model, Fig.17 B is the Extreme Tree (ET) classification model, Fig.17 C is the random forest (RF) classification model, Fig.17 is the support vector machine (SVM) classification model, Fig.17 E is the effect diagram after the above four classification models are fused based on voting. The results show that the SVM model has the best classification effect and can accurately identify six species of Meconopsis plants. Among them, the prediction rates of Meconopsis quinqueneri, Meconopsis integrifolia, Meconopsis racemosa and Meconopsis rubra are all above 98%, and the prediction rates of Meconopsis spinosa and Meconopsis hybrid are also above 70%. The model after Voting fusion is second, and the ET and RF models have better effects. The DT model has the worst recognition effect. The models after ET, RF and Voting fusion did not correctly identify Meconopsis spinosa. Similar to the near-infrared integrating sphere diffuse reflectance spectroscopy model, the recognition effect of various models on Meconopsis spinosa is the lowest, followed by Meconopsis hybrid. This is because the number of samples of these two Meconopsis is small. If the number of samples is sufficient, the recognition effect of the model on the samples will be improved accordingly. In the face of the conditions of large samples in actual production, the model is applicable to the discrimination of samples. Therefore, further in-depth research is needed in subsequent research.

[0141] 3.5 Comparison of model effects

[0142] According to the data in Tables 5 and 6, and Fig.15 , Fig.16 and Fig.17 By comparison, it can be found that the performance of the models under the two modeling software is good. The performance of the SVM model established by Python software is slightly better than that of the DM model established by TQ Analyst software. In addition, the SVM model also performs better than the DM model in the identification ability of six species of Meconopsis plants. By comparing the modeling effect of near-infrared data under the two acquisition methods and the discrimination results of samples, the performance of the model established by the spectral data collected by the near-infrared fiber optic probe and the recognition ability of Meconopsis are better than those of the spectrum collected by the near-infrared integrating sphere diffuse reflectance.

[0143] 4. Discussion

[0144] Near infrared spectroscopy analysis method belongs to the category of weak signal analysis, and has obvious advantages in rapid identification of traditional Chinese medicine. Therefore, the present invention uses near infrared spectroscopy technology combined with chemometrics to establish an identification model for different varieties of Meconopsis, realizes rapid and non-destructive identification of closely related species of Meconopsis plants, and provides a simple and effective means for quality control of Meconopsis resources. We analyzed the one-dimensional near infrared spectra of six Meconopsis plants obtained by two different spectral acquisition methods, and obtained 3 common peaks and 6 characteristic peaks under the condition of near infrared integrating sphere diffuse reflection, 3 common peaks and 7 characteristic peaks under the near infrared fiber probe, the latter has slightly more characteristic absorption peaks, and the collection information is more comprehensive. The existing classification of different varieties of Zhongning wolfberry and Dong'e Ejiao from different manufacturers and storage times shows that the spectrum under the integrating sphere diffuse reflection mode is more suitable for modeling, which may be due to the difference in the properties of different samples. Due to the different optical principles of near infrared integrating sphere diffuse reflection and fiber optic probe scanning samples, the best near infrared spectrum acquisition method can be selected for modeling analysis according to the differences of different samples in subsequent studies.

[0145] Furthermore, systematic cluster analysis is an important method for comparing the similarities and differences of things, but our study shows that systematic cluster analysis cannot accurately identify the closely related species of Meconopsis. Therefore, it is necessary to establish a more accurate species classification model to achieve accurate identification of a variety of Meconopsis plants. Comparison of the best models under the two modeling software showed that the robustness, predictiveness and classification ability of the SVM model established by Python were better than those of the DM model established by TQ Analyst. In actual research, the SVM algorithm is widely used in various research fields because of its strong classification ability and generalization performance in high-dimensional and nonlinear regression or classification problems, and has also achieved good results in variety classification. Comprehensively comparing the performance of the model and the recognition ability of the sample, it is believed that the SVM classification model established by using Python software combined with the spectra collected by the near-infrared fiber optic probe is the best. It achieves rapid and accurate identification of six species of Meconopsis, and its recognition rate, prediction rate and external validation set recognition rate are 99.68%, 98.47% and 97.98% respectively. The model has good performance and strong applicability. Compared with traditional identification methods, the application of near-infrared spectroscopy technology in this study has great advantages in the rapid and non-destructive identification of different varieties of Meconopsis, and is more suitable for actual production applications.

[0146] The above specific implementation methods are detailed descriptions of the present invention. It cannot be determined that the specific implementation methods of the present invention are limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions and substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for rapid identification of different species of Meconopsis plants by near infrared spectroscopy, characterized in that: The method comprises the following steps: S1. Collect samples of different species of Meconopsis; S2, collecting near-infrared integrating sphere diffuse reflectance spectra of different species of Meconopsis plant samples, and collecting near-infrared fiber optic probe spectra of different species of Meconopsis plant samples; S3. Analyze the near-infrared integrating sphere diffuse reflectance spectral characteristics and near-infrared fiber optic probe spectral characteristics of different species of Meconopsis plant samples; S4. Establish an integrating sphere diffuse reflectance spectrum discrimination model for different species of Meconopsis plant samples based on near-infrared integrating sphere diffuse reflectance spectrum; establish an optical fiber probe spectrum discrimination model for different species of Meconopsis plant samples based on near-infrared optical fiber probe spectrum; S5. Compare the prediction effects of the integrating sphere diffuse reflectance spectral discrimination model of different species of Meconopsis plant samples; compare the prediction effects of the optical fiber probe spectral discrimination model of different species of Meconopsis plant samples; S6. Compare the prediction effects of the integrating sphere diffuse reflectance spectrum discrimination model and the optical fiber probe spectrum discrimination model, and select the discrimination model with the best prediction effect.

2. The method for rapid identification of different species of Meconopsis plants by near infrared spectroscopy according to claim 1, characterized in that: The different species of Meconopsis plant samples include Meconopsis spinulosa, Meconopsis roxburghii, Meconopsis racemosa, Meconopsis hybrid, Meconopsis integrifolia and Meconopsis quinqueneri. The hybrid Meconopsis is a hybrid of Meconopsis quinqueneri and Meconopsis roxburghii.

3. The method for rapid identification of different species of Meconopsis plants by near infrared spectroscopy according to claim 1, characterized in that: The method of collecting near-infrared integrating sphere diffuse reflectance spectra of different species of Meconopsis plant samples comprises: Take a proper amount of the whole herb sample of Meconopsis thunbergii after passing through a 200-mesh sieve and place it in a sample cup, flatten it, and compact it. The sample thickness is 0.1 cm. During the collection process, the background interference of CO2 and water is deducted in real time. The scanning resolution is 8 cm -1 , scan times 64 times.

4. The method for rapid identification of different species of Meconopsis plants by near infrared spectroscopy according to claim 1, characterized in that: The near-infrared optical fiber probe spectrum of different species of Meconopsis plant samples is collected, including: The near-infrared fiber probe spectrum of the whole herb sample of Meconopsis thunbergii after passing through a 200-mesh sieve was collected using the fiber probe accessory of the Fourier transform infrared spectrometer. The background interference of CO2 and water was deducted in real time during the collection process. The scanning resolution was 4 cm -1 , scan times 64 times.

5. The method for rapid identification of different species of Meconopsis plants by near infrared spectroscopy according to claim 1, characterized in that: The method for analyzing the near-infrared integrating sphere diffuse reflectance spectral characteristics and near-infrared optical fiber probe spectral characteristics between different species of Meconopsis plant samples includes: The near-infrared one-dimensional spectra of the samples were characterized by analysis, and the common peaks were determined with the missing rate less than 5% as an indicator. The characteristic peaks were attributed, and the one-dimensional spectra were processed by first-order derivative and second-order derivative, and the amplification of the differences between the peaks of the first-order derivative spectrum and the second-order derivative spectrum were analyzed. At the same time, the three types of spectra were analyzed for similarity using the correlation coefficient as an indicator. Finally, the absorbance values ​​at the common peaks were subjected to one-way analysis of variance using SPSS software to determine whether there were differences between different species of Meconopsis.

6. The method for rapid identification of different species of Meconopsis plants by near infrared spectroscopy according to claim 1, characterized in that: The method further comprises: Cluster analysis was performed using near-infrared one-dimensional spectra of different species of Meconopsis plant samples to obtain different cluster groups.

7. The method for rapid identification of different species of Meconopsis plants by near infrared spectroscopy according to claim 1, characterized in that: The method of establishing an integrating sphere diffuse reflectance spectrum discrimination model for different species of Meconopsis plant samples based on near-infrared integrating sphere diffuse reflectance spectrum comprises: The best modeling condition is DM+MSC+D1+unsmoothed, and the modeling band is 9881.46-4119.21cm -1 .

8. The method for rapid identification of different species of Meconopsis plants by near infrared spectroscopy according to claim 1, characterized in that: The optical fiber probe spectrum discrimination model for different species of Meconopsis plant samples is established based on the near-infrared optical fiber probe spectrum, comprising: The best modeling conditions are: DM+SNV+D1+unsmoothed, and the modeling band is 9885.32-4364.12cm -1 .

9. The method for rapid identification of different species of Meconopsis plants by near infrared spectroscopy according to claim 2, characterized in that: Compare the prediction effects of the integrating sphere diffuse reflectance spectroscopy discrimination model for different species of Meconopsis, including: The discrimination accuracy of the integrating sphere diffuse reflectance spectroscopy discrimination model of the six species of Meconopsis plants is from high to low: Meconopsis quinquenervii > Meconopsis integrinus > Meconopsis racemosa > Meconopsis hybridus > Meconopsis roxburghii > Meconopsis spicata; The method of comparing the prediction effects of the optical fiber probe spectral discrimination model of different species of Meconopsis plant samples includes: The discrimination accuracy of the fiber optic probe spectral discrimination model for the six species of Meconopsis plants is as follows from high to low: Meconopsis integrifolia > Meconopsis hybrida > Meconopsis racemosa > Meconopsis quinqueneri > Meconopsis roxburghii > Meconopsis spicata.

10. A near infrared spectrum rapid identification system for different species of Meconopsis, characterized in that: The system comprises: A sample collection module, used to collect samples of different species of Meconopsis; A spectrum acquisition module, used for acquiring near-infrared integrating sphere diffuse reflectance spectra of different species of Meconopsis plant samples, and acquiring near-infrared optical fiber probe spectra of different species of Meconopsis plant samples; Spectral feature analysis module, used to analyze near-infrared integrating sphere diffuse reflectance spectral features and near-infrared fiber optic probe spectral features between different species of Meconopsis plant samples; A discrimination model building module is used to build an integrating sphere diffuse reflectance spectrum discrimination model for different species of Meconopsis plant samples based on near-infrared integrating sphere diffuse reflectance spectrum; and to build a fiber optic probe spectrum discrimination model for different species of Meconopsis plant samples based on near-infrared fiber optic probe spectrum; The discrimination model comparison module is used to compare the prediction effects of the integrating sphere diffuse reflectance spectrum discrimination model of different species of Meconopsis plant samples; and to compare the prediction effects of the optical fiber probe spectrum discrimination model of different species of Meconopsis plant samples; The discrimination model selection module is used to compare the prediction effects of the integrating sphere diffuse reflectance spectrum discrimination model and the optical fiber probe spectrum discrimination model, and select the discrimination model with the best prediction effect.

Images