Coptis chinensis moisture content analysis method based on heterogeneous sensor data fusion

Abstract

A Coptis chinensis moisture content analysis method based on heterogeneous sensor data fusion. The analysis method comprises the following steps: respectively collecting spectra of Coptis chinensis samples by using two types of spectrometers; gathering the collected spectra of the Coptis chinensis samples to form a sample set, and removing outliers to obtain a calibration set and a validation set; preprocessing the spectra, and selecting an optimal processing algorithm of the spectrometers to reduce spectral noise and simplify background spectral information; and performing low-level data fusion on spectra data, stitching spectral bands of near-infrared spectra and infrared spectra, so as to obtain moisture content prediction results, establishing a heterogeneous data fusion model, and on the basis of RMSE values and RPD values, evaluating the model and the prediction results, so as to obtain accurate analysis results.

Description

Method for analyzing the water content of Coptis chinensis based on heterogeneous sensor data fusion TECHNICAL FIELD

[0001] The present application relates to a method for analyzing the water content of Coptis chinensis based on heterogeneous sensor data fusion. BACKGROUND

[0002] Traditional Chinese medicines of plant origin contain a certain amount of water due to the influence of natural conditions and their own properties, and the water content varies due to the differences in their composition and internal structure. The water content of traditional Chinese medicines during processing is an important indicator for ensuring quality during storage and is a major factor affecting the quality of traditional Chinese medicines in storage, playing a leading role in quality changes. Reasonable water content can prevent insect infestation, mold growth, and decomposition of active ingredients during storage.

[0003] Generally, the water content of Coptis chinensis can be determined by drying method, which has high overall measurement accuracy and good reproducibility, but consumes a lot of time and cannot meet the needs of rapid and automated traditional Chinese medicine production. At the same time, the information obtained by a single sensor is insufficient to reveal the complex composition information of traditional Chinese medicines, and it cannot accurately and quickly analyze the water content of traditional Chinese medicine materials.

[0004] SUMMARY

[0005] The present application provides a method for analyzing the water content of Coptis chinensis based on heterogeneous sensor data fusion, which is reasonable in design and uses the combination of near-infrared spectroscopy and infrared spectroscopy for heterogeneous data fusion to compensate for the information deficiency of a single sensor. The spectral characterization technology is applied to the rapid analysis of the water content of Coptis chinensis traditional Chinese medicine, replacing the complex methods of traditional testing, and can accurately and quickly analyze the water content of traditional Chinese medicine materials, taking into account the overall analysis efficiency and accuracy, thereby meeting the needs of rapid and automated traditional Chinese medicine production and solving the problems in the prior art.

[0006] The technical solution adopted by the present application to solve the above technical problems is:

[0007] The method for analyzing the water content of Coptis chinensis based on heterogeneous sensor data fusion comprises the following steps:

[0008] S1. Two types of spectrometers are used to collect the spectra of Coptis chinensis samples;

[0009] S2. The collected spectra of Coptis chinensis samples are summarized into a sample set, and after removing outliers, a calibration set and a validation set are obtained;

[0010] S3. The spectra are preprocessed, and the best processing algorithm of the spectrometer is selected to reduce spectral noise and simplify background spectral information;

[0011] S4, low-level data fusion is performed on the spectral data, the spectral bands of the near-infrared spectrum and the infrared spectrum are spliced, a moisture content prediction result is obtained, a heterogeneous data fusion model is established, the model and the prediction result are evaluated based on RMSE (root mean square error) value and RPD (residual prediction deviation) value, and an accurate analysis result is obtained.

[0012] The spectrometer comprises an Antaris II Fourier near-infrared spectrometer and an ALPHA II infrared spectrometer to collect the near-infrared spectrum and the mid-infrared spectrum of the Coptis chinensis sample respectively.

[0013] The Antaris II Fourier near-infrared spectrometer collects one background spectrum before each sample, the scanning range is 10000cm -1 ~ 4000cm -1 , the scanning number is 32, the resolution is 8cm -1 , the gain is 2X, each sample is collected three times, and the average spectrum is taken as the analysis spectrum;

[0014] The ALPHA II infrared spectrometer collects one background spectrum before each sample, the scanning range is 400-4000nm, and the same sample is collected three times, and the average spectrum is taken as the analysis spectrum.

[0015] The collected Coptis chinensis sample spectrum is summarized as a sample set, and after removing outliers, a calibration set and a validation set are obtained, including the following steps:

[0016] S2.1, using principal component analysis combined with Mahalanobis distance, setting 95% confidence limit to judge outliers, and removing outliers of the Antaris II Fourier near-infrared spectrometer and the ALPHA II infrared spectrometer respectively;

[0017] S2.2, using batch division method to obtain the calibration set and the validation set.

[0018] The optimal processing algorithm is the first derivative combined with the convolution balance algorithm.

[0019] Low-level data fusion is performed on the spectral data, the spectral bands of the near-infrared spectrum and the infrared spectrum are spliced, a moisture content prediction result is obtained, a heterogeneous data fusion model is established, the model and the prediction result are evaluated based on RMSE value and RPD value, and an accurate analysis result is obtained, including the following steps:

[0020] S4.1, using a partial least squares model to predict the moisture content of different Coptis chinensis samples, projecting the prediction variables and the observed variables into a new space;

[0021] S4.2, capturing multiple latent variables representing most of the original information, and establishing a linear regression model;

[0022] S4.3, cross-validation of the fusion model according to the residual square selection factor of the optimal process.

[0023] The RMSE value evaluates the prediction ability of the model by measuring the deviation between the predicted value and the actual value, and the lower the RMSE value, the smaller the prediction error and the better the prediction ability.

[0024] The correlation coefficients of the calculated correction set and prediction set are the corresponding RMSEP (root mean square error of prediction) values and RPD values, which are used to estimate and verify the accuracy of the established heterogeneous data fusion model; the higher the RPD value, the better the prediction ability, and RPD>3 indicates good prediction.

[0025] The application adopts the above structure, and the determination of the moisture content of Coptis chinensis is affected by the data fusion technology of near-infrared spectroscopy and infrared spectroscopy, the analysis speed is fast, and a complex pretreatment method is not needed; the near-infrared spectrum and the infrared spectrum of the Coptis chinensis sample are collected by an Antaris II Fourier near-infrared spectrometer and an ALPHA II infrared spectrometer respectively; the prediction ability of the model is evaluated by measuring the deviation between the predicted value and the actual value, and the lower the RMSE value, the smaller the prediction error and the better the prediction ability; the accuracy of the established heterogeneous data fusion model is estimated and verified by calculating the corresponding RMSEP value and RPD value, and the application has the advantages of precision, practicality, rapidness and efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0026] FIG. 1 is an abnormal spectrum elimination result diagram of the application.

[0027] FIG. 2 is a spectrum pretreatment result diagram of the application. DETAILED DESCRIPTION

[0028] In order to clearly illustrate the technical features of the present application, the application will be described in detail below with specific embodiments and in combination with the accompanying drawings.

[0029] As shown in FIGS. 1-2, the Coptis chinensis moisture content analysis method based on heterogeneous sensor data fusion includes the following steps:

[0030] S1, two types of spectrometers are used to collect the spectrum of the Coptis chinensis sample;

[0031] S2, the collected Coptis chinensis sample spectrum is summarized as a sample set, and the correction set and the verification set are obtained after removing the abnormal values;

[0032] S3, the spectrum is pretreated, the best processing algorithm of the spectrometer is selected to reduce the spectrum noise and simplify the background spectrum information;

[0033] S4, low-level data fusion is performed on the spectral data, the spectral bands of the near-infrared spectrum and the infrared spectrum are spliced, a moisture content prediction result is obtained, a heterogeneous data fusion model is established, the model and the prediction result are evaluated based on the RMSE value and the RPD value, and an accurate analysis result is obtained.

[0034] The spectrometer comprises an Antaris II Fourier near-infrared spectrometer and an ALPHA II infrared spectrometer to collect the near-infrared spectrum and the mid-infrared spectrum of the Coptis chinensis sample respectively.

[0035] The Antaris II Fourier near-infrared spectrometer has a scanning mode of collecting 1 background spectrum before each sample, a scanning range of 10000cm -1 ~ 4000cm -1 , a scanning number of 32 times, a resolution of 8cm -1 , a gain of 2X, and each sample is collected three times, and the average spectrum is taken as the analysis spectrum.

[0036] The ALPHA II infrared spectrometer has a scanning mode of collecting 1 background spectrum before each sample, a scanning range of 400-4000nm, and the same sample is collected three times, and the average spectrum is taken as the analysis spectrum.

[0037] The collected Coptis chinensis sample spectrum is summarized as a sample set, and after removing outliers, a calibration set and a validation set are obtained, including the following steps:

[0038] S2.1, using principal component analysis combined with Mahalanobis distance, setting a 95% confidence limit to judge outliers, and removing outliers for the Antaris II Fourier near-infrared spectrometer and the ALPHA II infrared spectrometer;

[0039] S2.2, using batch division method to obtain the calibration set and the validation set.

[0040] The optimal processing algorithm is the first derivative combined with the convolution balance algorithm.

[0041] Low-level data fusion is performed on the spectral data, the spectral bands of the near-infrared spectrum and the infrared spectrum are spliced, a moisture content prediction result is obtained, a heterogeneous data fusion model is established, the model and the prediction result are evaluated based on the RMSE value and the RPD value, and an accurate analysis result is obtained, including the following steps:

[0042] S4.1, using a partial least squares model to predict the moisture content of different Coptis chinensis samples, projecting the prediction variables and the observed variables into a new space;

[0043] S4.2, capturing multiple latent variables representing most of the original information, and establishing a linear regression model;

[0044] S4.3, according to the residual square selection factor number of optimal process, cross-validation is carried out to the fusion model.

[0045] The RMSE value evaluates the prediction ability of the model by measuring the deviation between the predicted value and the actual value, and the lower the RMSE value, the smaller the prediction error and the better the prediction ability;

[0046] The correlation coefficient of the calculated correction set and prediction set is the corresponding RMSEP value and RPD value respectively, which is used to estimate and verify the accuracy of the established heterogeneous data fusion model; RPD is the standard deviation of the reference value and RMSEP of the prediction set, and the higher the RPD value, the better the prediction ability, and RPD>3 indicates good prediction.

[0047] The working principle of the Coptis chinensis moisture content analysis method based on heterogeneous sensor data fusion in the embodiment of the application is that the combination of near-infrared spectroscopy and infrared spectroscopy is used to compensate for the information loss of a single sensor, the spectral characterization technology is applied to the rapid analysis of the moisture content of Coptis chinensis traditional Chinese medicine, and the traditional complex testing method is replaced, so that the moisture content of traditional Chinese medicine can be accurately and rapidly analyzed, the overall analysis efficiency and accuracy are taken into account, and the demand for rapid and automatic production of traditional Chinese medicine is met.

[0048] The near-infrared spectrum is caused by the combination of C-H, O-H and N-H groups of the stretched vibration overtone, and is the most abundant chemical component feature in biological samples, and can reveal the complex components of traditional Chinese medicine by relying on near-infrared spectroscopy; however, the information obtained by a single sensor is less, and the analysis result is not accurate enough, so the near-infrared spectroscopy technology and the mid-infrared spectroscopy technology are combined to obtain more rich feature information, and a near-mid-infrared heterogeneous data fusion model is established to perform accurate analysis.

[0049] In the overall scheme, the following steps are mainly included: two types of spectrometers are used to collect the spectra of Coptis chinensis samples; the collected spectra of Coptis chinensis samples are summarized as a sample set, and after removing outliers, a correction set and a verification set are obtained; the spectra are pretreated, the best processing algorithm of the spectrometer is selected to reduce the spectral noise and simplify the background spectral information; the low-level data fusion is performed on the spectral data, the spectral bands of the near-infrared spectrum and the infrared spectrum are spliced, the moisture content prediction result is obtained, the heterogeneous data fusion model is established, and the model is evaluated based on the RMSE value and the RPD value.

[0050] The spectrometer of the application includes an Antaris II Fourier near-infrared spectrometer and an ALPHA II infrared spectrometer, which are used to collect the near-infrared spectrum and the mid-infrared spectrum of the Coptis chinensis sample respectively; before detecting the Coptis chinensis sample, it needs to be crushed and all passed through a 60-mesh sieve.

[0051] Generally, for the Antaris II Fourier near-infrared spectrometer: turn on the Antaris II Fourier near-infrared spectrometer in advance, and after confirming that the instrument is normal and stable, collect the near-infrared spectrum of the different varieties of Coptis chinensis samples by using a sample cell module, set the scanning mode as follows: collect 1 background spectrum before each sample, the scanning range is 10000cm -1 ~4000cm -1 , the scanning number is 32, the resolution is 8cm -1 , the gain is 2X, collect each sample for three times, and take the average spectrum as the analysis spectrum; for the ALPHA II infrared spectrometer: turn on the ALPHA II infrared spectrometer in advance, and after confirming that the instrument is normal and stable for half an hour, collect the sample by using an absorption module, which is consistent with the Antaris II Fourier near-infrared spectrometer, set the scanning mode as follows: collect 1 background spectrum before each sample, the scanning range is 400~4000nm, collect the same sample for three times, and take the average spectrum as the analysis spectrum.

[0052] Preferably, the collected Coptis chinensis sample spectrum is summarized as a sample set, and after removing the abnormal values, a calibration set and a verification set are obtained, including the following steps: using principal component analysis combined with Mahalanobis distance to set a 95% confidence limit to judge abnormal values, and removing abnormal values for the Antaris II Fourier near-infrared spectrometer and the ALPHA II infrared spectrometer respectively; and using batch division method to obtain the calibration set and the verification set.

[0053] In one specific embodiment, 65 spectra measured by the Antaris II Fourier near-infrared spectrometer and the ALPHA II infrared spectrometer are divided according to a calibration set and a verification set ratio of 3:1, 35 calibration set samples and 13 verification set samples are obtained, and there are 13 samples in the external prediction set.

[0054] During the spectrum collection process, the sample is easily affected by external conditions, including sample state, particle size, baseline measurement environment difference, etc., resulting in various noise information errors. Therefore, before performing multivariate statistical analysis, the spectrum data needs to be preprocessed to reduce system noise and enhance the spectral characteristics of the signal.

[0055] The inventors of the present application performed SNV (standard normal variable transformation), MSC (multivariate scatter correction), 1 st (first derivative), 2 nd (second derivative) and 1 st +SG (convolution balance algorithm) on the spectra measured by the Antaris II Fourier near-infrared spectrometer and the ALPHA II infrared spectrometer after removing abnormal values, and the results are shown in Table 1.

[0056] Table 1 Model evaluation parameters of five pretreatment methods

[0057] It can be seen that the use of MSC and SNV in infrared data preprocessing improves the accuracy of model prediction to a certain extent, and the RMSEP is reduced; after the first derivative and second derivative methods are processed, the RMSE is significantly reduced, and the model prediction accuracy is improved. According to the analysis results of all samples, the potential dependent variable of the first derivative combined with SG smoothing pretreatment method used in the application is the smallest, the RMSE of the model calibration set is the lowest, and the RMSEP of the validation set and the external prediction set is also the lowest, that is, the result error of model prediction is the smallest. At the same time, the RPD value is the highest, which indicates that the correlation between the predicted value and the actual test value is better, and the model prediction ability is higher. Therefore, the first derivative combined with convolution balance algorithm is selected to preprocess the infrared spectrum data.

[0058] In the near infrared spectrum pretreatment method, all the spectrum pretreatment methods improve the accuracy of model prediction, the RMSE is significantly reduced, the RPD value of the validation set and the external prediction set is improved, which indicates that the correlation between the predicted value and the actual test value is enhanced; the RMSEP value of the model validation set using the first derivative method is the lowest, the prediction ability is the highest, and the RPD value of the validation set and the external prediction set is the highest, and the correlation between the measured value and the actual test value is the best.

[0059] Preferably, the spectrum data is subjected to low-level data fusion to obtain a moisture content prediction result, an heterogeneous data fusion model is established, and the model is evaluated, including the following steps: using a partial least squares model to predict the moisture content of different Coptis samples, projecting the predicted variables and observed variables into a new space; capturing multiple potential variables representing most of the original information to establish a linear regression model; selecting the number of factors according to the residual square of the optimal process, and cross-validating the fusion model.

[0060] For the RMSE value, the prediction ability of the model is evaluated by measuring the deviation between the predicted value and the actual value, the lower the RMSE value, the smaller the prediction error, and the better the prediction ability;

[0061] The correlation coefficients of the calibration set and the prediction set calculated are the corresponding RMSEP values and RPD values, which are used to estimate and verify the accuracy of the established heterogeneous data fusion model; the RPD is the standard deviation of the reference value and the RMSEP of the prediction set, the higher the RPD value, the better the prediction ability, and RPD> 3 indicates good prediction.

[0062] The prediction results of the data fusion and the moisture content prediction results of the infrared spectrum and the near infrared spectrum established separately are shown in Table 2.

[0063] Table 2 Model evaluation parameters of data fusion

[0064] The results show that the accuracy of the model is improved, and the prediction ability of the model after data fusion is significantly improved.

[0065] In FIG. 1, (a) is a result graph of abnormal spectrum elimination of the Antaris II Fourier near-infrared spectrometer, and (b) is a result graph of abnormal spectrum elimination of the ALPHA II infrared spectrometer based on PCA score, PC1 score is the first principal component score, and PC2 score is the second principal component score.

[0066] In FIG. 2, (a) is a result graph of first derivative-SG pretreatment of the ALPHA II infrared spectrometer, and (b) is a result graph of first derivative pretreatment spectrum based on the Antaris II Fourier near-infrared spectrometer,

[0067] It is particularly pointed out that all data analysis of the application is processed and operated in MATLAB.

[0068] In summary, the method for analyzing the moisture content of Coptis based on the data fusion of heterogeneous sensors in the embodiment of the application compensates for the information loss of a single sensor based on the data fusion of heterogeneous data combining near-infrared spectroscopy and infrared spectroscopy, applies the spectral characterization technology to the rapid analysis of the moisture of Coptis traditional Chinese medicinal materials, replaces the complex method of traditional inspection, can accurately and rapidly analyze the moisture content of traditional Chinese medicinal materials, and takes into account the overall analysis efficiency and accuracy, so as to meet the needs of rapid and automatic production of traditional Chinese medicines.

[0069] The above specific embodiments cannot be regarded as a limitation on the protection scope of the application, and any alternative improvement or change made by those skilled in the art to the embodiments of the application falls within the protection scope of the application.

[0070] The details not described in the application are well-known technology of those skilled in the art.

Claims

1. A method for analyzing the moisture content of Coptis based on the fusion of heterogeneous sensor data, characterized in that, The analysis method comprises the following steps: S1, two types of spectrometers are used to collect spectra of the Coptis chinensis samples respectively; S2, the collected spectra of the Coptis chinensis samples are summarized as a sample set, and after removing outliers, a calibration set and a verification set are obtained; S3, the spectra are preprocessed, the best processing algorithm of the spectrometer is selected to reduce the spectral noise and simplify the background spectral information; S4, low-level data fusion is performed on the spectral data, the spectral bands of the near-infrared spectrum and the infrared spectrum are spliced, a moisture content prediction result is obtained, a heterogeneous data fusion model is established, the model and the prediction result are evaluated based on the RMSE value and the RPD value, and an accurate analysis result is obtained.

2. The method for analyzing the water content of Coptis chinensis based on the data fusion of heterogeneous sensors according to claim 1, characterized in that: The spectrometer comprises an Antaris II Fourier near-infrared spectrometer and an ALPHA II infrared spectrometer, which are used to collect near-infrared spectra and mid-infrared spectra of the Coptis chinensis samples respectively.

3. The method of analyzing the water content of Coptis based on the data fusion of heterogeneous sensors according to claim 2, characterized in that: The scanning mode of the Antaris II Fourier near-infrared spectrometer is to collect one background spectrum before each sample, the scanning range is 10000cm -1 ~4000cm -1 , the scanning number is selected as 32, the resolution is selected as 8cm -1 , the gain is selected as 2X, each sample is collected three times repeatedly, and the average spectrum is taken as the analysis spectrum; The scanning mode of the ALPHA II infrared spectrometer is to collect background spectra once before each sample, the scanning range is 400-4000 nm, and the same sample is collected repeatedly for three times, and the average spectrum is taken as the analysis spectrum.

4. The method of claim 2, wherein the method is based on a fusion of heterogeneous sensor data. The collected spectra of the Coptis chinensis samples are summarized as a sample set, and after removing outliers, a calibration set and a verification set are obtained, which comprises the following steps: S2.1, principal component analysis combined with Mahalanobis distance is used to judge outliers, and the confidence limit is set to 95%, and the outliers are removed from the Antaris II Fourier near-infrared spectrometer and the ALPHA II infrared spectrometer respectively; S2.2, batch division method is used to obtain the calibration set and the verification set.

5. The method of analyzing the water content of Coptis chinensis based on the fusion of heterogeneous sensor data according to claim 1, characterized in that: The best processing algorithm is the first derivative combined with the convolution balance algorithm.

6. The method of analyzing the content of the water in the coptis based on the data fusion of the heterogeneous sensors according to claim 1, characterized in that, Low-level data fusion is performed on the spectral data, the spectral bands of the near-infrared spectrum and the infrared spectrum are spliced, a moisture content prediction result is obtained, a heterogeneous data fusion model is established, the model and the prediction result are evaluated based on the RMSE value and the RPD value, and an accurate analysis result is obtained, which comprises the following steps: S4.1, the partial least squares model is used to predict the moisture content of different Coptis chinensis samples, and the prediction variable and the observation variable are projected into a new space; S4.2, a plurality of latent variables representing most of the original information are captured, and a linear regression model is established; S4.3, the number of selection factors is selected according to the residual square of the optimal process, and the fusion model is cross-validated.

7. The method of analyzing the water content of Coptis chinensis based on the data fusion of heterogeneous sensors according to claim 1, characterized in that: The RMSE value is used to evaluate the prediction ability of the model by measuring the deviation between the predicted value and the actual value, the lower the RMSE value, the smaller the prediction error, and the better the prediction ability; The correlation coefficients of the calculated calibration set and the prediction set are the corresponding RMSEP values and RPD values respectively, which are used to estimate and verify the accuracy of the established heterogeneous data fusion model; the higher the RPD value, the better the prediction ability, and RPD>3 indicates good prediction.

Images