Pericarpium citri reticulatae comprehensive quality evaluation method and device based on near-infrared spectrometer and analytic hierarchy process-entropy weight method

Abstract

The invention discloses a method for evaluating the comprehensive quality of dried orange peel based on a near-infrared spectrometer and an analytic hierarchy process-entropy weight method, and belongs to the technical field of optical analysis, and the method comprises the following steps: preparing a dried orange peel sample, acquiring near-infrared spectrum data of the dried orange peel sample, and preprocessing; acquiring the year information of the production place of the dried orange peel sample, and measuring the hesperidin content, the nobiletin content, the tangeretin content, the anti-inflammatory activity, the delicate flavor, the sweet taste, the bitter taste, the overall preference degree and the comprehensive score value of the dried orange peel in the dried orange peel sample; establishing a pericarpium citri reticulatae producing area year discrimination model, a hesperidin quantitative model, a nobiletin quantitative model, a tangeretin quantitative model, an anti-inflammatory activity quantitative model, a delicate flavor quantitative model, a sweet taste quantitative model, a bitter taste quantitative model, an overall preference quantitative model and a comprehensive score quantitative model; and detecting a to-be-detected dried orange peel sample, and evaluating the dried orange peel quality. The method has the advantages of economy, simplicity, convenience, rapidness, greenness and the like, and provides a basis for comprehensively evaluating the quality of the pericarpium citri reticulatae.

Description

Technical Field

[0001] The invention relates to the technical field of optical analysis, and more specifically to a method and a device for evaluating the comprehensive quality of tangerine peel based on a near-infrared spectrometer and a hierarchy analysis method-entropy weight method. Background Art

[0002] Tangerine peel is the dried mature peel of the citrus plant (Citrus reticulata Blanco) and its cultivated varieties of the Rutaceae family. The medicinal materials are divided into "tangerine peel" and "Guang tangerine peel (Xinhui tangerine peel)", and Guang tangerine peel is of better quality. Tangerine peel was first recorded in "Shennong's Herbal Classic". It tastes bitter and pungent, is warm in nature, and belongs to the lung and spleen meridians. It can regulate qi and strengthen the spleen, dry dampness and resolve phlegm. It is often used for abdominal distension, anorexia, vomiting, cough and phlegm. Tangerine peel is rich in flavonoids, volatile oils, polysaccharides and trace elements, and plays an important role in anti-inflammatory, anti-oxidation, anti-tumor, anti-viral and anti-atherosclerosis.

[0003] The quality of Chinese medicine affects the safety and effectiveness of Chinese medicine, and is related to the quality of clinical efficacy and the development of scientific research. At present, Chinese medicine quality evaluation mainly adopts single Chinese medicine quality evaluation mode such as property evaluation, chemical evaluation and biological evaluation, and its evaluation system has certain one-sidedness and limitation. As the Chinese patent of announcement number CN114252520A discloses a detection and classification method of dried tangerine peel, it adopts high performance liquid chromatography, particularly cooperates with suitable high performance liquid chromatography conditions, detects the difference components existing between dried tangerine peels of different quality levels, and analyzes the peak area of ​​these difference components, particularly the ratio of the peak area of ​​these difference components to the peak area of ​​vitaenin-2, determines the type of dried tangerine peel to be measured, and finally realizes the quality evaluation of dried tangerine peel to be measured, but the evaluation mode is single, not comprehensive enough.

[0004] Therefore, the present invention proposes a comprehensive evaluation method for tangerine peel quality based on near infrared spectrometer and analytic hierarchy process-entropy weight method. Summary of the invention

[0005] The invention aims to provide a method for evaluating the comprehensive quality of tangerine peel based on a near-infrared spectrometer and an analytic hierarchy process-entropy weight method. The method collects near-infrared spectra of tangerine peels from different origins and years, performs multiple preprocessing on the spectra, establishes an origin year qualitative model, a hesperidin quantitative model, a nobiletin quantitative model, a tangerine peel quantitative model, an ANTI-NO quantitative model, an umami quantitative model, a sweetness quantitative model, a bitterness quantitative model, an overall liking quantitative model and a comprehensive scoring quantitative model, comprehensively detects the quality indicators of tangerine peel, and improves the comprehensiveness and accuracy of tangerine peel quality evaluation.

[0006] In order to achieve these objectives of the present invention, the present invention provides a method for evaluating the comprehensive quality of tangerine peel based on a near infrared spectrometer and a hierarchy analysis method-entropy weight method, comprising:

[0007] Step 1: prepare a dried tangerine peel sample and obtain near infrared spectral data of the dried tangerine peel sample; the near infrared spectral data includes: original near infrared spectral data and pre-processed near infrared spectral data, and the pre-processing includes: performing a first-order derivative (1 st derivative), second-order derivative (2 nd derivative), Savitzky-Golay smoothing (SG), multivariate scatter correction (MSC), first-order derivative + Savitzky-Golay smoothing (1 st derivative+SG), second-order derivative+Savitzky-Golay smoothing (2 nd derivative+SG), multivariate scatter correction+first-order derivative+Savitzky-Golay smoothing (MSC+1 st derivative+SG), multivariate scatter correction+second-order derivative+Savitzky-Golay smoothing (MSC+2 nd derivative+SG);

[0008] Step 2, obtaining the origin year information of the tangerine peel sample, and determining the hesperidin content, nobiletin content, tangeretin content, tangerine peel anti-inflammatory activity (ANTI-NO), umami, sweetness, bitterness, overall liking and comprehensive score value in the tangerine peel sample;

[0009] Step 3, according to the near infrared spectrum data of the tangerine peel sample, the information of the origin year, the content of hesperidin, the content of nobiletin, the content of tangerine peel, the anti-inflammatory activity (ANTI-NO) of tangerine peel, the umami, the sweetness, the bitterness, the overall liking and the comprehensive score are respectively associated to establish any one or more of the tangerine peel origin year discrimination model, the hesperidin quantitative model, the nobiletin quantitative model, the tangerine peel quantitative model, the anti-inflammatory activity quantitative model, the umami quantitative model, the sweetness quantitative model, the bitterness quantitative model, the overall liking quantitative model and the comprehensive score quantitative model;

[0010] Step 4, substituting the near-infrared spectral data of the tangerine peel sample to be measured into the tangerine peel origin year discrimination model, hesperidin quantitative model, nobiletin quantitative model, tangerine peel quantitative model, anti-inflammatory activity (ANTI-NO) quantitative model, umami quantitative model, sweetness quantitative model, bitterness quantitative model, overall liking quantitative model, and comprehensive score quantitative model to obtain tangerine peel quality indicators such as origin year information, hesperidin content, nobiletin content, tangerine peel anti-inflammatory activity (ANTI-NO), umami, sweetness, bitterness, overall liking and comprehensive score value of the tangerine peel sample to be measured, which are used to evaluate the quality of tangerine peel.

[0011] Preferably, in the step 1, tangerine peel samples from different origins and years are crushed and sieved to obtain tangerine peel powder; the tangerine peel powder is loaded into a quartz dish with a filling height of 1 to 2 cm, and near-infrared spectrum data of the tangerine peel sample is measured using a spectrometer, with a measured wavelength range of 900 to 1700 nm (nanometers).

[0012] Preferably, in step three:

[0013] 1) The establishment process of the tangerine peel origin year discrimination model is as follows:

[0014] The partial least squares discriminant analysis model (PLS-DA) or orthogonal partial least squares discriminant analysis model (OPLS-DA) was established based on the origin year data and near-infrared spectrum data of tangerine peel samples.

[0015] 2) The process of establishing the hesperidin quantitative model is as follows:

[0016] Obtaining the reference value of hesperidin content: The hesperidin content in dried tangerine peel was determined by high performance liquid chromatography;

[0017] According to the hesperidin content and near-infrared spectral data of tangerine peel samples, the full-spectrum partial least squares regression (full-PLS) model, the joint interval partial least squares regression (si-PLS), the backward interval partial least squares regression (bi-PLS) and the support vector regression (SVR) model were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness and stability of the hesperidin content prediction model.

[0018] 3) The process of establishing the quantitative model of nobiletin is as follows:

[0019] Obtaining the reference value of nobiletin content: The content of nobiletin in tangerine peel was determined by high performance liquid chromatography;

[0020] According to the nobiletin content and near-infrared spectral data of tangerine peel samples, the full-spectrum partial least squares regression (full-PLS) model, the joint interval partial least squares regression (si-PLS), the backward interval partial least squares regression (bi-PLS) and the support vector regression (SVR) model were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness and stability of the nobiletin content prediction model.

[0021] 4) The process of establishing the quantitative model of tangeretin is as follows:

[0022] Obtaining the reference value of tangeretin content: The content of tangeretin in dried orange peel was determined by high performance liquid chromatography;

[0023] According to the tangeretin content and near-infrared spectral data of tangerine peel samples, the full-spectrum partial least squares regression (full-PLS) model, the joint interval partial least squares regression (si-PLS), the backward interval partial least squares regression (bi-PLS) and the support vector regression (SVR) model were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness and stability of the tangeretin content prediction model.

[0024] 5) The process of establishing the anti-inflammatory activity (ANTI-NO) quantitative model is as follows:

[0025] Acquisition of ANTI-NO reference value: RAW264.7 cells were used to construct an inflammatory model, and a NO detection kit was used to determine the NO content in the cell supernatant after administration of tangerine peel water extract, and the inhibition rate of nitric oxide production was calculated;

[0026] According to the anti-inflammatory activity of tangerine peel and near-infrared spectral data of tangerine peel samples, the full-spectrum partial least squares regression (full-PLS) model, the joint interval partial least squares regression (si-PLS), the backward interval partial least squares regression (bi-PLS) and the support vector regression model (SVR) were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness and stability of the ANTI-NO prediction model.

[0027] 6) The process of establishing the umami quantitative model is as follows:

[0028] Umami reference value acquisition: Calculate umami value through sensory evaluation;

[0029] According to the umami and near-infrared spectral data of tangerine peel samples, the full-spectrum partial least squares regression (full-PLS) model, the joint interval partial least squares regression (si-PLS), the backward interval partial least squares regression (bi-PLS) and the support vector regression (SVR) model were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness and stability of the umami prediction model.

[0030] 7) The process of establishing the sweetness quantitative model is as follows:

[0031] Obtaining sweetness reference value: Calculate sweetness value through sensory evaluation;

[0032] According to the sweetness and near-infrared spectral data of tangerine peel samples, the full-spectrum partial least squares regression (full-PLS) model, the joint interval partial least squares regression (si-PLS), the backward interval partial least squares regression (bi-PLS) and the support vector regression (SVR) model were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness and stability of the sweetness prediction model.

[0033] 8) The establishment process of the bitterness quantitative model is as follows:

[0034] Bitterness reference value acquisition: Calculate bitterness value through sensory evaluation;

[0035] According to the bitterness and near-infrared spectral data of tangerine peel samples, the full-spectrum partial least squares regression (full-PLS) model, the joint interval partial least squares regression (si-PLS), the backward interval partial least squares regression (bi-PLS) and the support vector regression (SVR) model were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness and stability of the bitterness prediction model.

[0036] 9) The process of establishing the overall likeability quantitative model is as follows:

[0037] Obtaining the reference value of overall liking: Calculate the overall liking through sensory evaluation;

[0038] According to the overall liking degree and near-infrared spectral data of tangerine peel samples, a full-spectrum partial least squares regression (full-PLS) model, a joint interval partial least squares regression (si-PLS), a backward interval partial least squares regression (bi-PLS) and a support vector regression (SVR) model were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness and stability of the overall liking degree prediction model.

[0039] 10) The process of establishing the comprehensive scoring quantitative model is as follows:

[0040] Obtaining the reference value of the comprehensive score: The comprehensive score of all samples was calculated based on the data of hesperidin, nobiletin, tangeretin, nitric oxide production inhibition rate, umami, sweetness, bitterness and overall liking, combined with the weights determined by the hierarchical analysis method and the entropy weight method;

[0041] According to the comprehensive score values ​​and near-infrared spectral data of tangerine peel samples, the full-spectrum partial least squares regression (full-PLS) model, the joint interval partial least squares regression (si-PLS), the backward interval partial least squares regression (bi-PLS) and the support vector regression model (SVR) were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness and stability of the comprehensive score prediction model.

[0042] Furthermore, the PLS-DA model and OPLS-DA model establishment method are as follows:

[0043] All tangerine peel near-infrared spectroscopy data were divided into a training set and a prediction set at a ratio of 3:1. The training set was used to build the model, and the prediction set was used to verify the model.

[0044] The pre-processed tangerine peel near-infrared spectroscopy data were used as input variables, and the origin year variable was used as the output variable. The SIMCA14.1 software was used for normalization to obtain the corresponding model.

[0045] Furthermore, the full-PLS model, si-PLS model, bi-PLS model, and SVR model are established as follows:

[0046] Divide the data of near infrared spectra of tangerine peel samples into a training set and a prediction set; the training set is 3 / 4 of the number of all tangerine peel samples; the training set model is used to build the model, and the prediction set is used to verify the model;

[0047] The preprocessed near-infrared spectral data were used as input variables, and the hesperidin content, nobiletin content, tangeretin content, ANTI-NO, umami value, sweetness value, bitterness value, overall liking and comprehensive score values ​​were imported into matlab 2023a software as output variables to establish full-PLS, si-PLS, bi-PLS and SVR models respectively.

[0048] Furthermore, a cross-validation method was used to verify the performance of the established model. 7-fold cross-validation was used as the validation method.

[0049] Preferably, in establishing full-PLS, si-PLS, bi-PLS and SVR models, internal cross-validation is used to optimize modeling parameters and the prediction correlation coefficient (R P ), root mean square error of prediction (RMSEP), and relative analytical error (RPD) were used to evaluate the model performance.

[0050] Furthermore, among the models for discriminating the year of origin, the best one is when the spectral data is the original spectrum or the SG, MSC preprocessed spectrum, and the best model is the PLS-DA or OPLS-DA model.

[0051] The optimal hesperidin quantitative model was the bi-PLS model.

[0052] The optimal quantitative model for nobiletin was the si-PLS model.

[0053] The optimal quantitative model for tangeretin is the SVR model.

[0054] The optimal anti-inflammatory activity (ANTI-NO) quantitative model was the bi-PLS model.

[0055] The optimal umami quantitative model is the full-PLS model.

[0056] The optimal sweetness quantitative model is the si-PLS model

[0057] The optimal bitterness quantitative model was the bi-PLS model.

[0058] The optimal quantitative model of overall liking is the si-PLS model.

[0059] The optimal comprehensive scoring quantitative model is the si-PLS model.

[0060] Hesperidin content detection: The spectral data of unknown samples were imported into the optimized bi-PLS model, and the hesperidin content was obtained based on the results.

[0061] Nobiletin content detection: The spectral data of unknown tangerine peel samples were imported into the si-PLS model, and the nobiletin content was obtained based on the results.

[0062] Tangeretin content detection: The spectral data of unknown samples were imported into the SVR model, and the tangeretin content was obtained based on the results.

[0063] ANTI-NO detection: Import the unknown sample spectral data into the bi-PLS model and obtain ANTI-NO based on the results.

[0064] Umami value detection: The spectral data of unknown samples are imported into the full-PLS model, and the umami value is obtained based on the results.

[0065] Sweetness value detection: The spectral data of unknown samples are imported into the si-PLS model, and the sweetness value is obtained based on the results.

[0066] Bitterness value detection: The spectral data of unknown samples are imported into the bi-PLS model, and the bitterness value is obtained based on the results.

[0067] Overall likeability test: The unknown sample spectral data is imported into the si-PLS model, and the overall likeability is obtained based on the results.

[0068] Comprehensive score detection: Import the unknown sample spectral data into the si-PLS model and obtain the comprehensive score based on the results.

[0069] The preprocessing method of the above unknown sample spectral data is the same as that of the optimal model.

[0070] Preferably, the tangerine peel powder is sieved through an 80-mesh sieve.

[0071] Preferably, the detection range of the near infrared spectrum is 900-1700 nm.

[0072] Preferably, the scanning resolution of the near infrared spectrum is 10 nm.

[0073] Preferably, the near infrared spectrum is scanned 64 times.

[0074] Preferably, a portable near-infrared spectrometer is used for spectral detection.

[0075] A device for evaluating the comprehensive quality of dried tangerine peel based on a near-infrared spectrometer and a hierarchy analysis method-entropy weight method comprises a storage unit and a processing unit, wherein the storage unit stores a program and data of the method for evaluating the comprehensive quality of dried tangerine peel based on a near-infrared spectrometer and a hierarchy analysis method-entropy weight method, and the processing unit executes the method for evaluating the comprehensive quality of dried tangerine peel based on a near-infrared spectrometer and a hierarchy analysis method-entropy weight method.

[0076] The present invention has at least the following beneficial effects:

[0077] The method of the present invention establishes a qualitative model of the year of origin, a quantitative model of hesperidin, a quantitative model of nobiletin, a quantitative model of tangeretin, a quantitative model of ANTI-NO, a quantitative model of umami, a quantitative model of sweetness, a quantitative model of bitterness, a quantitative model of overall liking, a quantitative model of comprehensive scoring, etc., and comprehensively detects the quality indicators of tangerine peel, thereby improving the comprehensiveness and accuracy of tangerine peel quality evaluation.

[0078] The method of the present invention performs multiple preprocessing on the near infrared spectrum, and optimizes to obtain the best tangerine peel origin year qualitative model, hesperidin quantitative, nobiletin quantitative, tangeretin quantitative, ANTI-NO quantitative, umami quantitative, sweet quantitative, bitter quantitative, overall liking quantitative and comprehensive score quantitative model. The method has a qualitative accuracy of origin year of 100%; the R of the hesperidin quantitative model is 100%. p was 0.9735, and the RPD was 4.25; the R p is 0.9501, and the RPD is 3.10; the R p is 0.9198, and RPD is 2.22; R p is 0.9152, and the RPD is 2.46; the R p is 0.8273, and the RPD is 1.44; the R p is 0.8064, and the RPD is 1.53; the R p is 0.8190, and the RPD is 1.51; the R p is 0.8049, and the RPD is 1.56; the R p The value of the chromatogram was 0.9198, and the RPD was 2.02. This provides a basis for evaluating the quality of dried tangerine peel in a simple, fast, accurate, green and economical way.

[0079] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1are the original and average near-infrared spectra of all tangerine peel samples;

[0081] Figure 2 For all dried tangerine peel samples 1 st Average near infrared spectrum after derivative pretreatment;

[0082] Figure 3 For all dried tangerine peel samples 2 nd Average near infrared spectrum after derivative pretreatment;

[0083] Figure 4 is the average near infrared spectrum of all dried tangerine peel samples under SG pretreatment;

[0084] Figure 5 The average near infrared spectra of all dried tangerine peel samples pretreated with MSC;

[0085] Figure 6 For all dried tangerine peel samples 1 st Average near-infrared spectra under derivative+SG pretreatment;

[0086] Figure 7 For all dried tangerine peel samples 2 nd Average near-infrared spectra under derivative+SG pretreatment;

[0087] Figure 8 MSC+1 for all dried tangerine peel samples st Average near-infrared spectra under derivative+SG pretreatment;

[0088] Fig. 9 MSC+2 for all dried tangerine peel samples nd Average near-infrared spectra under derivative+SG pretreatment;

[0089] Fig.10 This is the three-dimensional score diagram of the PLS-DA model of the qualitative model of the year of origin of tangerine peel;

[0090] Fig.11 This is the three-dimensional score diagram of the OPLS-DA model of the qualitative model of the year of origin of dried tangerine peel;

[0091] Fig.12 It is the correlation diagram of the measured value and predicted value of the hesperidin quantitative model;

[0092] Fig.13 This is the correlation diagram of the measured value and predicted value of the quantitative model of nobiletin;

[0093] Fig.14 This is the correlation diagram between the measured value and the predicted value of the quantitative model of tangeretin;

[0094] Fig.15 It is the correlation diagram of the measured value and predicted value of the ANTI-NO quantitative model;

[0095] Fig.16 It is the correlation diagram between the measured value and the predicted value of the umami quantitative model;

[0096] Fig.17 It is the correlation diagram between the measured value and the predicted value of the sweetness quantitative model;

[0097] Fig.18 It is the correlation diagram between the measured value and the predicted value of the bitterness quantitative model;

[0098] Fig.19 This is the correlation diagram between the measured value and the predicted value of the quantitative model of overall liking;

[0099] Fig. 20 It is a correlation diagram of the measured value and predicted value of the quantitative model of the comprehensive score. DETAILED DESCRIPTION

[0100] The present invention is further described in detail below with reference to examples so that those skilled in the art can implement the invention with reference to the description.

[0101] It should be understood that the terms such as “having”, “including” and “comprising” used herein do not exclude the existence or addition of one or more other elements or combinations thereof.

[0102] The analytic hierarchy process (AHP) and the entropy weight method (EWM) are two common index weighting methods, which are often used in traditional Chinese medicine, engineering, management and other disciplines. The analytic hierarchy process is a subjective weighting method with the advantages of subjective logical judgment and analysis. It aggregates and combines factors at different levels according to the mutual influence and affiliation between factors to form a multi-level analysis structure model, so that the problem is ultimately reduced to the determination of the relative importance of the lowest level (schemes and measures for decision-making) relative to the highest level (overall goal) or the ranking of the relative order of superiority and inferiority. The entropy weight method is an objective weighting method used to solve the relationship between the weights of each index in the multi-index comprehensive evaluation problem. It determines the weight according to the discrete degree of the index. The greater the discrete degree, the greater the weight of the index.

[0103] The combined weighting of the analytic hierarchy process and the entropy weight method can not only reduce the subjectivity of the analytic hierarchy process weighting, but also reduce the problem of small weights of important indicators due to small discrete degrees caused by the entropy weight method weighting, and reduce weight fluctuations caused by data changes. The combination of the two can complement each other and make the combined weight more scientific.

[0104] Near infrared light is an electromagnetic wave between mid-infrared and visible light, and has strong penetrating power. Therefore, by collecting the near infrared spectrum of the sample, the composition and molecular structure of the organic matter in the sample can be obtained by using the double frequency and combined frequency absorption of the vibration of hydrogen-containing groups in organic matter. Near infrared spectroscopy is a fast, non-destructive and pollution-free analytical method. Combined with chemometrics, it has gradually become a new method for evaluating the quality of traditional Chinese medicine.

[0105] The present invention discloses a method for evaluating the comprehensive quality of dried tangerine peel based on a portable near-infrared spectrometer and a hierarchical analysis method-entropy weight method, comprising the following steps:

[0106] (1) Sample pretreatment: All dried tangerine peel samples from different origins and years were crushed into powder using a grinder and passed through an 80-mesh sieve to obtain dried tangerine peel powder.

[0107] (2) Near infrared spectroscopy data acquisition: Tangerine peel powder was placed in a quartz dish with a filling height of 1-2 cm, and the original near infrared spectroscopy data of the sample was measured using a spectrometer with a measurement wavelength range of 900-1700 nm; the near infrared spectroscopy detection resolution was 10 nm; and the near infrared spectroscopy detection scanning number was 64 times.

[0108] (3) Spectral preprocessing: The spectral data obtained in step (2) was imported into TQ Analyst EZ Edition software for first-order derivative analysis (1 st derivative), second-order derivative (2 nd derivative), Savitzky-Golay smoothing (SG), multivariate scatter correction (MSC), first-order derivative + Savitzky-Golay smoothing (1 st derivative+SG), second-order derivative+Savitzky-Golay smoothing (2 nd derivative+SG), multivariate scatter correction+first-order derivative+Savitzky-Golay smoothing (MSC+1 st derivative+SG), multivariate scatter correction+second-order derivative+Savitzky-Golay smoothing (MSC+2 nd derivative+SG); the number of smoothing points is 5.

[0109] (4) Establishing a discriminant model for the year of origin of dried tangerine peel: Using the data obtained in step (3) and the original near-infrared spectral data, a partial least squares discriminant analysis (PLS-DA) model and an orthogonal partial least squares discriminant analysis (OPLS-DA) model were established respectively, and the optimal data processing method and its optimal analysis model were selected.

[0110] (5) Determination of origin year: The spectral data of unknown samples were imported into the PLS-DA and OPLS-DA models to obtain the identification result of the origin year of tangerine peel.

[0111] (6) Determination of hesperidin content: According to the guidance of the Chinese Pharmacopoeia (2020 edition), the hesperidin content was determined by high performance liquid chromatography (HPLC).

[0112] (7) Determination of nobiletin content: According to the guidance of the Chinese Pharmacopoeia (2020 edition), the nobiletin content was determined by high performance liquid chromatography (HPLC).

[0113] (8) Determination of tangeretin content: According to the guidance of the Chinese Pharmacopoeia (2020 edition), the tangeretin content was determined by high performance liquid chromatography (HPLC).

[0114] (9) Determination of anti-inflammatory activity: Dried tangerine peel was extracted by water extraction and heated under reflux, filtered and freeze-dried to obtain a dried tangerine peel water extract for subsequent cell experiments. RAW264.7 cells were used to induce an inflammation model by lipopolysaccharide (LPS) stimulation to determine the anti-inflammatory activity of different dried tangerine peels. The anti-inflammatory activity was determined by using a NO detection kit (Biyuntian) to determine the NO content in the cell supernatant. The anti-inflammatory activity was expressed as the inhibition rate of carbon monoxide production (ANTI-NO), and the calculation formula was as follows:

[0115] ANTI-NO=[(A 模型组 -A 给药组 ) / (A 模型组 -A 空白组 )]*100%

[0116] (10) Umami score acquisition: The dried tangerine peel was brewed, and 10 scorers rated the umami taste of the dried tangerine peel water on a scale of 1 to 9.

[0117] (11) Sweetness rating: Dried tangerine peel was brewed, and 10 reviewers rated the sweetness of the dried tangerine peel water on a scale of 1 to 9.

[0118] (12) Bitterness score acquisition: The dried tangerine peel was brewed, and 10 scorers scored the bitterness of the dried tangerine peel water on a scale of 1 to 9.

[0119] (13) Obtaining the overall liking rating: The dried tangerine peel was brewed, and 10 raters rated the overall liking of the dried tangerine peel water on a scale of 1 to 9.

[0120] (14) Obtaining a comprehensive score: In order to comprehensively and fully evaluate the quality of dried tangerine peel, the present invention uses a hierarchical analysis method combined with an entropy weight method to determine the weights of hesperidin, nobiletin, tangeretin, anti-inflammatory activity, umami, sweetness, bitterness, and overall liking to calculate a comprehensive score, including the following steps:

[0121] 1) The weights of hesperidin, nobiletin, tangeretin, anti-inflammatory activity, umami, sweetness, bitterness and overall liking were calculated by analytic hierarchy process. Including:

[0122] a) Score the indicators in pairs according to the TLsaaty1~9 scaling method and construct a judgment matrix.

[0123] b) Normalize the judgment matrix:

[0124] c) Add the processed judgment matrix row by row:

[0125] d) Vector Perform normalization: i=1,2,…,m; then N=[N1,N2,…,N m ] T is the desired eigenvector, N i is the weight of each indicator.

[0126] e) Perform a one-time test on the judgment matrix: the largest eigenvalue

[0127] CI=(λ max -m) / (m-1)(m is the index number)

[0128] CR = CI / RI (where CR is the random consistency ratio, CI is the consistency index, and RI is the average random consistency index of the matrix)

[0129] f) The weight values ​​obtained by the hierarchical analysis method are: hesperidin (0.2756), nobiletin (0.0940), tangeretin (0.0940), anti-inflammatory activity (0.3927), umami (0.0170), sweetness (0.0400), bitterness (0.0204), and overall liking (0.0664)

[0130] 2) The weights of hesperidin, nobiletin, tangeretin, anti-inflammatory activity, umami, sweetness, bitterness and overall liking were calculated by the entropy weight method.

[0131] a) Construct the original data matrix B = (b ij ) m*n , n is the evaluation object, m is the evaluation index, b ij Represents the i-th indicator under the j-th evaluation object.

[0132] b) Normalization of the original data matrix

[0133] For positive indicators

[0134] For negative indicators

[0135] In addition, the weight of the jth evaluation object under the i-th indicator needs to be calculated based on the standardized value.

[0136] c) Entropy weight method is used to calculate the entropy value e under the i-th indicator i :

[0137] d) Calculate the coefficient of variation g i :g i =1-e i

[0138] e) Calculate the weight of each indicator:

[0139] f) The weight values ​​obtained by entropy weight method are: hesperidin (0.1681), nobiletin (0.1808), tangeretin (0.1312), anti-inflammatory activity (0.1383), umami (0.1114), sweetness (0.1094), bitterness (0.0700), and overall liking (0.0907).

[0140] 3) Determination of the combined weight of AHP-EWM: (The weight values ​​of each index are: hesperidin (0.3225), nobiletin (0.1183), tangeretin (0.0858), anti-inflammatory activity (0.3780), umami (0.0132), sweetness (0.0304), bitterness (0.0099), overall liking (0.0419))

[0141] 4) Calculate the comprehensive score of each sample

[0142]

[0143] (15) Establishing a quantitative model for hesperidin: Using the data obtained in steps (3) and (6), a full-spectrum partial least squares regression (full-PLS) model, a joint interval partial least squares regression (si-PLS) model, a backward interval partial least squares regression (bi-PLS) model, and a support vector regression model (SVR) were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness, and stability of the hesperidin content prediction model.

[0144] (16) Establishing a quantitative model for citrus nobiletin: Using the data obtained in steps (3) and (7), a full-spectrum partial least squares regression (full-PLS) model, a joint interval partial least squares regression (si-PLS), a backward interval partial least squares regression (bi-PLS), and a support vector regression model (SVR) were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness, and stability of the citrus nobiletin content prediction model.

[0145] (17) Establishment of a quantitative model for tangeretin: The data obtained in steps (3) and (8) were used to establish a full-spectrum partial least squares regression (full-PLS) model, a joint interval partial least squares regression (si-PLS), a backward interval partial least squares regression (bi-PLS), and a support vector regression model (SVR), and the optimal modeling parameters were selected to ensure the accuracy, robustness, and stability of the tangeretin content prediction model.

[0146] (18) Establishing ANTI-NO quantitative model: Using the data obtained in steps (3) and (9), a full-spectrum partial least squares regression (full-PLS) model, a joint interval partial least squares regression (si-PLS), a backward interval partial least squares regression (bi-PLS) and a support vector regression model (SVR) were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness and stability of the ANTI-NO prediction model.

[0147] (19) Establishing a quantitative umami model: Using the data obtained in steps (3) and (10), a full-spectrum partial least squares regression (full-PLS) model, a joint interval partial least squares regression (si-PLS), a backward interval partial least squares regression (bi-PLS), and a support vector regression model (SVR) were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness, and stability of the umami prediction model.

[0148] (20) Establishing a sweetness quantitative model: Using the data obtained in steps (3) and (11), a full-spectrum partial least squares regression (full-PLS) model, a joint interval partial least squares regression (si-PLS), a backward interval partial least squares regression (bi-PLS), and a support vector regression model (SVR) were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness, and stability of the sweetness prediction model.

[0149] (21) Establishing a bitterness quantitative model: Using the data obtained in steps (3) and (12), a full-spectrum partial least squares regression (full-PLS) model, a joint interval partial least squares regression (si-PLS), a backward interval partial least squares regression (bi-PLS), and a support vector regression model (SVR) were established, and the optimal modeling parameters were selected to ensure the accuracy, robustness, and stability of the bitterness prediction model.

[0150] (22) Establishing a quantitative model of overall likeability: Using the data obtained in steps (3) and (13), a full-spectrum partial least squares regression (full-PLS) model, a joint interval partial least squares regression (si-PLS), a backward interval partial least squares regression (bi-PLS) and a support vector regression model (SVR) are established, and the optimal modeling parameters are selected to ensure the accuracy, robustness and stability of the overall likeability prediction model.

[0151] (23) Establishing a quantitative model for the comprehensive score value: The data obtained in steps (3) and (14) are used to establish a full-spectrum partial least squares regression (full-PLS) model, a joint interval partial least squares regression (si-PLS), a backward interval partial least squares regression (bi-PLS) and a support vector regression model (SVR), and the optimal modeling parameters are selected to ensure the accuracy, robustness and stability of the comprehensive score value prediction model.

[0152] (24) Hesperidin content detection: The spectral data of unknown samples were imported into the optimized bi-PLS model, and the hesperidin content was obtained based on the results.

[0153] (25) Nobiletin content detection: The spectral data of unknown samples were imported into the optimized si-PLS model, and the Nobiletin content was obtained based on the results.

[0154] (26) Tangeretin content detection: The spectral data of unknown samples were imported into the optimized SVR model, and the tangeretin content was obtained based on the results.

[0155] (27) ANTI-NO detection: The spectral data of unknown samples were imported into the optimized bi-PLS model, and ANTI-NO was obtained based on the results.

[0156] (28) Umami value detection: The spectral data of unknown samples were imported into the optimized full-PLS model, and the umami value was obtained based on the results.

[0157] (29) Sweetness value detection: The spectral data of unknown samples were imported into the optimized si-PLS model, and the sweetness value was obtained based on the results.

[0158] (30) Bitterness value detection: The spectral data of unknown samples were imported into the optimized bi-PLS model, and the bitterness value was obtained based on the results.

[0159] (31) Overall likeability test: The spectral data of unknown samples were imported into the optimized si-PLS model, and the overall likeability was obtained based on the results.

[0160] (32) Comprehensive score detection: The spectral data of unknown samples are imported into the optimized si-PLS model, and the comprehensive score is obtained based on the results.

[0161] Wherein, the PLS-DA and OPLS-DA model establishment methods are as follows:

[0162] a) Sample pretreatment. Same as step (1);

[0163] b) Near infrared spectroscopy data collection. Same as step (2);

[0164] c) Spectral preprocessing. Same as step (3);

[0165] d) Model establishment. All the tangerine peel near infrared spectroscopy data were divided into a training set and a prediction set according to a ratio of 3:1. The training set was used to build the model, and the prediction set was used to verify the model. The pre-processed data obtained in step c) was used as an input variable, and the origin year variable was used as an output variable to import SIMCA 14.1 software for normalization processing to obtain a model.

[0166] Among them, for the above steps (15), (16), (17), (18), (19), (20), (21), (22) and (23), the full-PLS model is modeled using the original spectrum and different preprocessed spectra, and the models established after different preprocessing are respectively established and compared to obtain the optimal full-PLS model.

[0167] The si-PLS model divides the full spectrum into 10, 15, 20, 25, and 30 equal-width intervals, and the best si-PLS models of 2, 3, and 4 sub-interval combinations are established and compared, respectively.

[0168] The bi-PLS model divides the full spectrum into 10, 15, 20, 25, and 30 intervals of equal width, and the best bi-PLS model is obtained by comparison.

[0169] The SVR model is modeled using the original spectrum and different pre-processed spectra, and the models established after different pre-processing are respectively established and compared to obtain the best SVR model.

[0170] Furthermore, the full-PLS, si-PLS, bi-PLS, and SVR models are established as follows:

[0171] a) Sample pretreatment. Same as step (1);

[0172] b) Near infrared spectroscopy data collection. Same as step (2);

[0173] c) Spectral preprocessing. Same as step (3);

[0174] d) Model establishment. The data of the near infrared spectrum of the tangerine peel samples are divided into a training set and a prediction set; the training set is 3 / 4 of the number of all tangerine peel samples; the training set model is used to build the model, and the prediction set is used to verify the model; the pre-processed data obtained in step c) are used as input variables, and the hesperidin content, nobiletin content, tangeretin content, ANTI-NO, umami value, sweetness value, bitterness value, overall liking and comprehensive score values ​​are used as output variables. Import the matlab 2023a software to establish full-PLS, si-PLS, bi-PLS and SVR models respectively.

[0175] Example 1

[0176] 1. Materials and instruments

[0177] 1.1 Sample

[0178] In the present embodiment, 9 kinds of dried orange peels of different origin years are collected, with a total of 150 samples. Specifically, they are Guangdong 2018 (n=16), Guangdong 2020 (n=16), Guangdong 2022 (n=16), Guangxi 2018 (n=18), Guangxi 2020 (n=18), Guangxi 2022 (n=18), Sichuan 2018 (n=16), Sichuan 2020 (n=16), and Sichuan 2022 (n=16).

[0179] The dried tangerine peel was crushed and passed through an 80-mesh sieve to obtain dried tangerine peel powder. Each sample was individually packed in a plastic sealed bag and finally stored in a desiccator. The reference values ​​of hesperidin, nobiletin and tangeretin contents were obtained by HPLC. The ANTI-NO reference value was determined by a NO detection kit in the lipopolysaccharide-induced RAW264.7 cell inflammation model to determine the NO content in the cell supernatant after administering dried tangerine peel water extract, and the anti-inflammatory activity was expressed by the carbon monoxide production inhibition rate (ANTI-NO). The reference values ​​of umami, sweetness, bitterness and overall liking were obtained by sensory scoring. The reference value of the comprehensive score was calculated by determining the indicator weights using the hierarchical analysis method combined with the entropy weight method.

[0180] 1.2 Instrumentation

[0181] A portable near-infrared spectrometer (NIR-M-F1-C, Shenzhen Spectronix Interconnect Technology Co., Ltd.); an ELISA reader (Infinite M 200PRO, TECAN, Switzerland); a high performance liquid chromatograph (Agilent 1260InfinityⅡ, USA); and TQAnalyst EZ Edition software were used to preprocess the near-infrared spectra.

[0182] 2. Methods and Results

[0183] 2.1 Near infrared spectrum acquisition

[0184] The sample was scanned at room temperature, with a scanning frequency of 64 times, a resolution of 10 nm, and a scanning range of 900-1700 nm. Before the measurement, the background was scanned and the influence of the background spectrum was subtracted. The tangerine peel powder sample was placed in a quartz dish with a filling height of about 1-2 cm. Each sample was measured 6 times and the average value was taken. The original and average near-infrared spectra of all tangerine peel samples collected are shown in Figure 1. Figure 1 As shown, the left side is the original near-infrared spectrum data, and the right side is the average near-infrared spectrum data.

[0185] 2.2 Spectral preprocessing method

[0186] The original near-infrared spectra contain a lot of redundant information and are affected by baseline drift and light scattering effects, which may reduce the model performance. In order to eliminate baseline drift, background noise and light scattering effects and improve model performance, the original near-infrared spectra are preprocessed using TQAnalyst EZ Edition software. The preprocessing method used in the present invention includes 1 st derivative, 2 nd derivative, SG, MSC, 1 st derivative+SG, 2 nd derivative+SG、MSC+1 st derivative+SG、MSC+2 nd derivative+SG. The processing result is as follows Figures 2 to 10 Shown are the average near-infrared spectra of all tangerine peel samples under different pretreatments.

[0187] The qualitative models of origin year can be compared by the discrimination accuracy, the larger the accuracy, the better the performance.

[0188] The quantitative model was performed using R P The performance of the system is evaluated by RPD. RPD is judged as follows: RPD ≥ 2.5, excellent performance; 2.5>RPD ≥ 2.0, very good performance; 2.0>RPD ≥ 1.8, good performance; 1.8>RPD ≥ 1.4, average performance; 1.4>RPD ≥ 1.0, poor performance; 1.0>RPD ≥ 0, extremely poor performance.

[0189] 2.3 Determination of origin and year

[0190] The prediction set data was imported into the PLS-DA and OPLS-DA models to identify the year of origin of dried tangerine peel. The classification results of the models under different pretreatments are shown in Table 1. The number of misjudgments was 0 and the classification accuracy was 100%, indicating that the near-infrared spectroscopy combined with the PLS-DA and OPLS-DA models can accurately identify the year of origin of dried tangerine peel without pretreatment. The three-dimensional score graphs of the PLS-DA and OPLS-DA models are shown in Table 1. Fig.10 and Fig.11 shown.

[0191] 2.4 Quantitative model establishment

[0192] 2.4.1 Full-PLS model establishment

[0193] The full spectrum band (900-1700nm) was used to establish the PLS model.

[0194] 2.4.2 si-PLS model establishment

[0195] The spectrum is divided into 10, 15, 20, 25 and 30 intervals, which form a combination of 2, 3, and 4 sub-intervals to establish the PLS model. The latent variable (LV) has a great influence on the prediction performance of the model. When the LV is too large, overfitting often occurs. When the LV is too small, the prediction results of the model will be very poor due to insufficient fitting. The LV range of this embodiment is set to 0-15.

[0196] 2.4.3bi-PLS model establishment

[0197] The entire spectrum was divided into n equal intervals. The intervals were modeled according to the seven-fold cross-validation method to obtain the RMSECV values ​​of the corresponding intervals. The worst interval was removed and the remaining n-1 intervals were iteratively modeled using the seven-fold cross-validation method until the minimum value of RMSECV was obtained. The bi-PLS model was established by dividing the spectrum into 10, 15, 20, 25, and 30 equal intervals.

[0198] 2.4.4 SVR model establishment

[0199] The SVR model was established using full spectrum data with different preprocessing.

[0200] 2.5 Prediction of Hesperidin Content

[0201] From Table 2, we can see that using 1 st The bi-PLS model after derivative pretreatment has the best prediction performance for hesperidin in tangerine peel, so the detection of unknown samples also undergoes 1 st After derivative preprocessing, it is input into the bi-PLS model to achieve accurate prediction. PThe value was 0.9735, the RMSEP value was 4.9756, and the RPD value was 4.25, indicating that the model has excellent performance. The bi-PLS method effectively selected the spectral interval closely related to the hesperidin content, thereby improving the prediction performance of the hesperidin quantitative model. Fig.12 The correlation between the measured and predicted values ​​of hesperidin content is shown.

[0202] 2.6 Prediction of Nobiletin Content

[0203] As shown in Table 2, after MSC preprocessing, the optimal model is the si-PLS model. Using si-PLS for wavelength selection can reduce irrelevant variables and improve the prediction accuracy of the model. P The value is 0.9501, the RMSEP value is 0.4960, and the RPD value is 3.10. Fig.13 This is the correlation diagram between the measured and predicted values ​​of nobiletin content.

[0204] 2.7 Prediction of tangeretin content

[0205] As can be seen from Table 2, compared with the PLS model, the MSC+1 st The SVR model after derivative+SG preprocessing obtains the best prediction results. P The value was 0.9198, the RMSEP value was 0.2139, and the RPD value was 2.22. The correlation between the measured and predicted values ​​of tangeretin is shown in Fig.14 The RPD value of the optimal SVR model was greater than 2, showing very good performance. Compared with the PLS model, the SVR model provided better prediction results, indicating that there was a nonlinear relationship between the near-infrared spectral data and tangeretin.

[0206] 2.8 ANTI-NO Prediction

[0207] From Table 2, we can see that the optimal model is nd The bi-PLS model after derivative preprocessing can improve the effect of the model by selecting the bands related to anti-inflammatory activity for modeling. The best model has an R P The value is 0.9152, the RMSEP value is 0.0481, and the RPD value is 2.46. Fig.15 This is a correlation diagram between the measured value and the predicted value of the anti-inflammatory activity of dried tangerine peel.

[0208] 2.9 Umami Value Prediction

[0209] As shown in Table 2, the best model compared with other models is the full-PLS model without preprocessing, with an R PThe value is 0.8273, the RMSEP value is 0.3144, and the RPD value is 1.44. The RPD value of the optimal model is greater than 1.4, and it is considered that the model can be used for prediction. Fig.16 The correlation between the measured and predicted values ​​of umami taste is shown.

[0210] 2.10 Sweetness value prediction

[0211] As can be seen from Table 2, the optimal model is the si-PLS model without preprocessing. Through variable selection, the bands related to sweetness can be selected. The R P The value is 0.8064, the RMSEP value is 0.3731, and the RPD value is 1.53. Fig.17 Correlation plot between measured and predicted values ​​of sweetness.

[0212] 2.11 Bitterness value prediction

[0213] As shown in Table 2, after MSC+2 nd The bi-PLS model after derivative+SG preprocessing has the best performance. P The value was 0.8190, the RMSEP value was 0.8661, and the RPD value was 1.51. Among them, bi-PLS effectively selected the spectral interval closely related to bitterness, which improved the prediction performance of the model. Fig.18 The correlation between the measured and predicted values ​​of bitterness values ​​is shown.

[0214] 2.12 Overall Likeability Prediction

[0215] As shown in Table 2, by comparing the performance of different models under different pretreatments, the optimal model is the si-PLS model under the original spectrum. Wavelength selection can reduce irrelevant variables, and the R of the optimal model is P The value is 0.8049, the RMSEP value is 0.6880, and the RPD value is 1.56. Fig.19 This is a correlation diagram of the overall liking measurement and predicted value.

[0216] 2.13 Comprehensive score prediction

[0217] As can be seen from Table 2, the performance of all models of the comprehensive score value is good. The best model is the model established after si-PLS preprocessing with SG. Its RPD value is 2.02, which is greater than 2.0. The model performance is very good and has great potential for practical application. The correlation diagram of the measured and predicted values ​​of the comprehensive score value is shown in Figure 2. Fig. 20 .

[0218] Table 1. PLS-DA and OPLS-DA model discrimination results under different preprocessing

[0219]

[0220] Table 2. Results of the quantitative model of the effects of different pretreatment methods on hesperidin content, nobiletin content, tangeretin content, tangerine peel anti-inflammatory activity (ANTI-NO), umami, sweetness, bitterness, overall liking and comprehensive score value.

[0221]

[0222]

[0223]

[0224] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and the embodiments. They can be applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily realized.

Claims

1. A method for evaluating the comprehensive quality of dried tangerine peel based on near infrared spectroscopy and analytic hierarchy process-entropy weight method, characterized in that: include: Step 1, preparing a dried tangerine peel sample and obtaining near infrared spectroscopy data of the dried tangerine peel sample; The near infrared spectral data include: original near infrared spectral data and preprocessed near infrared spectral data, and the preprocessing includes: performing any one of the first-order derivative, second-order derivative, Savitzky-Golay smoothing, multivariate scattering correction, first-order derivative + Savitzky-Golay smoothing, second-order derivative + Savitzky-Golay smoothing, multivariate scattering correction + first-order derivative + Savitzky-Golay smoothing, and multivariate scattering correction + second-order derivative + Savitzky-Golay smoothing; Step 2, obtaining the origin year information of the tangerine peel sample, and determining the hesperidin content, nobiletin content, tangeretin content, tangerine peel anti-inflammatory activity, umami taste, sweetness, bitterness, overall liking and comprehensive score value in the tangerine peel sample; Step 3, according to the near infrared spectrum data of the tangerine peel sample, the information of the origin year, the content of hesperidin, the content of nobiletin, the content of tangerine peel, the anti-inflammatory activity of tangerine peel, the umami, the sweetness, the bitterness, the overall liking and the comprehensive score are respectively associated to establish a tangerine peel origin year discrimination model, a hesperidin quantitative model, a nobiletin quantitative model, a tangerine peel quantitative model, an anti-inflammatory activity quantitative model, an umami quantitative model, a sweetness quantitative model, a bitterness quantitative model, an overall liking quantitative model, and a comprehensive score quantitative model; Step 4. Substitute the near-infrared spectral data of the tangerine peel sample to be measured into the tangerine peel origin year discrimination model, hesperidin quantitative model, nobiletin quantitative model, tangeretin quantitative model, anti-inflammatory activity quantitative model, umami quantitative model, sweetness quantitative model, bitterness quantitative model, overall liking quantitative model, and comprehensive score quantitative model to obtain the origin year information, hesperidin content, nobiletin content, tangeretin content, tangerine peel anti-inflammatory activity, umami, sweetness, bitterness, overall liking and comprehensive score value of the tangerine peel sample to be measured, which is used to evaluate the quality of tangerine peel.

2. The tangerine peel comprehensive quality evaluation method based on near infrared spectroscopy and analytic hierarchy process-entropy weight method as claimed in claim 1, characterized in that, The near infrared spectrum data of the tangerine peel sample was measured by a spectrometer, and the measured wavelength range was 900 to 1700 nm.

3. The tangerine peel comprehensive quality evaluation method based on near infrared spectroscopy and analytic hierarchy process-entropy weight method as claimed in claim 2, characterized in that, The comprehensive score is calculated by combining the analytic hierarchy process with the entropy weight method to determine the weights of hesperidin, nobiletin, tangeretin, anti-inflammatory activity, umami, sweetness, bitterness and overall liking.

4. The tangerine peel comprehensive quality evaluation method based on near infrared spectroscopy and analytic hierarchy process-entropy weight method as claimed in claim 3, characterized in that, 1) The establishment process of the tangerine peel origin year discrimination model is as follows: A partial least squares discriminant analysis model or an orthogonal partial least squares discriminant analysis model is established according to the origin year data of the tangerine peel sample and the near infrared spectrum data of the tangerine peel sample, and any one of the analysis models is selected as the tangerine peel origin year discriminant model; 2) The establishment process of the hesperidin quantitative model, the nobiletin quantitative model, the tangeretin quantitative model, the anti-inflammatory activity quantitative model, the umami quantitative model, the sweetness quantitative model, the bitterness quantitative model, the overall liking quantitative model, and the comprehensive score quantitative model is as follows: According to the hesperidin content, nobiletin content, tangeretin content, anti-inflammatory activity, umami taste, sweetness, bitterness, overall liking and comprehensive score of the dried tangerine peel samples, and the near-infrared spectroscopy data of the dried tangerine peel samples; The full-spectrum partial least squares regression model, joint interval partial least squares regression, backward interval partial least squares regression and support vector regression models were established, and the optimal modeling parameters were selected to obtain the corresponding quantitative models.

5. The tangerine peel comprehensive quality evaluation method based on near infrared spectroscopy and analytic hierarchy process-entropy weight method as claimed in claim 4, characterized in that, The method for establishing the partial least squares discriminant analysis model or the orthogonal partial least squares discriminant analysis model is: All the near-infrared spectral data of tangerine peel were divided into training set and prediction set in a ratio of 3:

1. The training set was used to build the model, and the prediction set was used to verify the model. The preprocessed near-infrared spectral data of tangerine peel were used as input variables, and the year of origin variable was used as output variable. They were imported into SIMCA software for normalization to obtain the corresponding model.

6. The tangerine peel comprehensive quality evaluation method based on near infrared spectroscopy and analytic hierarchy process-entropy weight method as claimed in claim 4, characterized in that, The establishment methods of the full spectrum partial least squares regression model, the joint interval partial least squares regression model, the backward interval partial least squares regression model and the support vector regression model are as follows: The data of near infrared spectra of tangerine peel samples were divided into training set and prediction set; the training set was 3 / 4 of all tangerine peel samples; the training set model was used to build the model, and the prediction set was used to verify the model; The preprocessed near-infrared spectral data were used as input variables, and the hesperidin content, nobiletin content, tangeretin content, anti-inflammatory activity, umami value, sweetness value, bitterness value, overall liking and comprehensive score values ​​were imported into MATLAB software as output variables to establish the above regression models.

7. The method for evaluating the comprehensive quality of dried tangerine peel based on near infrared spectroscopy and analytic hierarchy process-entropy weight method as claimed in claim 6, characterized in that: In establishing the regression model, cross-validation was used to optimize the modeling parameters and the prediction correlation coefficient R was used. P , prediction root mean square error RMSEP, and relative analysis error RPD are used to evaluate the model performance.

8. The method for evaluating the comprehensive quality of dried tangerine peel based on near infrared spectroscopy and analytic hierarchy process-entropy weight method as claimed in claim 6, characterized in that: The hesperidin quantitative model is a backward interval partial least squares regression model; the nobiletin quantitative model is a joint interval partial least squares regression model; the tangeretin quantitative model is a support vector regression model; the anti-inflammatory activity quantitative model is a backward interval partial least squares regression model; the umami quantitative model is a full spectrum partial least squares regression model; the sweetness quantitative model is a joint interval partial least squares regression model; the bitterness quantitative model is a backward interval partial least squares regression model; the overall liking quantitative model is a joint interval partial least squares regression model; and the comprehensive score quantitative model is a joint interval partial least squares regression model.

9. the tangerine peel comprehensive quality evaluation method based on near infrared spectroscopy and analytic hierarchy process-entropy weight method as claimed in claim 6, the spectral data of unknown tangerine peel sample is imported in the corresponding quantitative model, obtains corresponding tangerine peel quality index.

10. A device for evaluating the comprehensive quality of tangerine peel based on a near-infrared spectrometer and a hierarchy analysis method-entropy weight method, comprising a storage unit and a processing unit, the storage unit storing the program and data of the tangerine peel comprehensive quality evaluation method based on a near-infrared spectrometer and a hierarchy analysis method-entropy weight method as described in any one of claims 1 to 9, and the processing unit executing the tangerine peel comprehensive quality evaluation method based on a near-infrared spectrometer and a hierarchy analysis method-entropy weight method as described in any one of claims 1 to 9.

Images