Material quality rapid detection and sorting equipment and Chinese medicine material production system

Abstract

The embodiments of the present invention provide a material quality rapid detection and sorting equipment and a Chinese medicine material production system, which relate to the field of material sorting equipment. The material quality rapid detection and sorting equipment includes a body, a spectrometer and a rejector. The body is provided with a feed port and a discharge port. A conveyor belt for transmitting materials between the feed port and the discharge port is provided in the body. The spectrometer is installed in the body and located above the conveyor belt. The rejector is installed in the body and located at the end of the conveyor belt. The spectrometer is connected to the rejector for communication. The discharge port includes a first outlet and a second outlet. The rejector selectively opens the first outlet or the second outlet. It can be sorted based on the quality of medicinal materials according to the quality, greatly reducing the uncertainty and time-consuming nature of manual detection. It makes up for the defects of only single sorting standards such as size and weight at present, and provides a more powerful quality assurance for the sorting of Chinese medicine materials.

Description

Technical Field

[0001] The present invention relates to the field of material sorting equipment, and in particular to a material quality rapid detection and sorting equipment and a traditional Chinese medicine material production system. Background Art

[0002] At present, the authenticity and quality of Chinese medicine materials are mainly sorted manually on the production line of Chinese medicine materials, and the classification is done by naked eyes and experience. This sorting method has high labor intensity, low production efficiency, and is easily affected by subjective factors, resulting in poor quality stability of Chinese medicine materials. Therefore, it is necessary to use a sorting machine to sort materials, improve production efficiency, and accelerate the modernization of Chinese medicine.

[0003] The phenomenon of "fakes being indistinguishable from the real" in the traditional Chinese medicine market is endless. It is difficult to distinguish the authenticity with the naked eye, and the quality of traditional Chinese medicine cannot be ensured. In addition, the quality of traditional Chinese medicine is not only reflected in the appearance, but more importantly, the internal material composition. It is of great significance to sort the quality of traditional Chinese medicine by combining the effective component content index. The existing traditional Chinese medicine material sorting technology usually uses blowers, vibrating screens and other equipment, focusing on the classification of size and weight, with a single classification standard, and has not yet used spectral technology to evaluate and sort the authenticity and quality of traditional Chinese medicine materials.

[0004] In view of this, this application is hereby filed. Summary of the invention

[0005] The purpose of the present invention includes, for example, providing a material quality rapid detection and sorting equipment and a Chinese medicine material production system, which can quickly, effectively and in real time evaluate and sort the authenticity and quality of Chinese medicine materials.

[0006] The embodiments of the present invention can be implemented as follows:

[0007] In a first aspect, the present invention provides a material quality rapid detection and sorting device, which includes a body, a spectrometer and a rejector, the body is provided with a feed port and a discharge port, the body is provided with a conveyor belt located between the feed port and the discharge port for transmitting materials, the spectrometer is installed in the body and located above the conveyor belt, the rejector is installed in the body and located at the end of the conveyor belt, the spectrometer is communicatively connected with the rejector, the discharge port includes a first outlet and a second outlet, and the rejector selectively opens the first outlet or the second outlet.

[0008] In an optional embodiment, the material quality rapid detection and sorting equipment further includes a display, which is mounted on the outside of the body and is connected to both the spectrometer and the rejector.

[0009] In an optional embodiment, the spectrometer is a portable near-infrared spectrometer, a portable ultraviolet spectrometer or a hyperspectral imager.

[0010] In an optional embodiment, the spectrometer is detachably connected to the body.

[0011] In an optional embodiment, a qualitative model for adulteration screening or a quantitative model for medicinal material content is introduced into the spectrometer.

[0012] In an optional embodiment, the method for constructing the qualitative model for adulteration screening includes: respectively identifying the macroscopic and microscopic characteristics of the true sample and the false sample, and then crushing the true sample and the false sample to prepare adulterated samples with different contents of the true sample; collecting near-infrared spectra of multiple adulterated samples, recording them continuously for multiple times, and calculating the average spectrum; applying PCA to the average spectrum to visualize the sample distribution, and using the Kennard-Stone method to divide the multiple adulterated samples into a training set for model calibration and a test set for external validation to screen samples; using the PLS-DA analysis method to classify the screened samples into true samples and false samples; and using PLS or LS-SVM to quantify the adulterated samples to predict the adulteration rate.

[0013] In an optional embodiment, the method for constructing the quantitative model of medicinal material content includes: using a hyperspectral imager to collect spectra of medicinal material samples, collecting 8-12 spectra for each batch of samples, and calculating the average value as the hyperspectral data of the medicinal material samples; optimizing the hyperspectral data and combining the partial least squares method to establish the quantitative model of medicinal material content;

[0014] The optimization of the hyperspectral data includes: using RHM and MCCV methods to remove abnormal samples in the original spectrum and the known index component content data, and determining valid samples to establish a quantitative prediction model for each index component; using 1 st D, SNV, PSNV, or 1 st The D+PSNV method is used to preprocess the spectral reflectance data to improve the signal-to-noise ratio of the spectrum; the correlation coefficient between the preprocessed spectrum and the content of known index components is analyzed; the wavelength is screened using the SPA analysis method, the correlation coefficient method or the full spectrum analysis method;

[0015] Preferably, 1 st The D+PSNV method was used to preprocess the spectral reflectance data;

[0016] Preferably, the wavelengths are screened using SPA analysis.

[0017] In an optional embodiment, the conveyor belt includes a plurality of conveyor belts;

[0018] Preferably, the transmission speed of the conveyor belt is 0.1-0.5 m / s.

[0019] In an optional embodiment, the material quality rapid detection and sorting equipment also includes a feeding mechanism, which includes a feeding hopper, a screw feeder and a motor. The feeding hopper is installed at the feeding port of the machine body, and the screw feeder is vertically installed in the feeding hopper. One end of the screw feeder is connected to the motor, and the other end extends into the bottom of the feeding hopper.

[0020] In a second aspect, the present invention provides a Chinese medicine material production system, which includes a material quality rapid detection and sorting device as described in any of the aforementioned embodiments.

[0021] The beneficial effects of the embodiments of the present invention include, for example:

[0022] The material quality rapid detection and sorting equipment provided by the present application installs a spectrometer in the body. The spectrometer can be selected according to actual needs. The rejector identifies the authenticity and quality of the materials on the conveyor belt according to the collected spectral data, and then sorts them. The quality can be sorted based on the quality of the medicinal materials, which greatly reduces the uncertainty and time-consuming nature of manual inspection. It makes up for the defects of the current single sorting standards such as size and weight, and provides a more powerful quality assurance for the sorting of traditional Chinese medicine materials. The traditional Chinese medicine material production system provided by the present application includes the above-mentioned material quality rapid detection and sorting equipment. By introducing the above-mentioned material quality rapid detection and sorting equipment into the traditional Chinese medicine material production system, the quality control of the traditional Chinese medicine materials produced by the traditional Chinese medicine material production system can be better. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 A schematic diagram of the structure of the material quality rapid detection and sorting equipment provided in this application;

[0025] Figure 2 This is a schematic diagram of collecting spectrum of Hypericum perforatum canopy in constructing the quantitative model of medicinal material content provided in this application;

[0026] Figure 3 A schematic diagram of using the RHM method to eliminate abnormal samples in the original spectrum in constructing a quantitative model for medicinal material content provided in this application;

[0027] Figure 4Schematic diagram of using the MCCV method to determine abnormal samples in the content of rutin (a), isoquercetin (b), hyperoside (c), quercetin (d) and hypericin (e) in the quantitative model of medicinal material content provided in the present application;

[0028] Figure 5 Hyperspectral images of the original canopies of 154 batches of Hypericum perforatum used in the construction of the quantitative model of medicinal material content provided in this application;

[0029] Figure 6 This is the hyperspectral image after 1st D+PSNV preprocessing in the quantitative model of medicinal material content provided in this application;

[0030] Figure 7 Correlation coefficient diagrams of the original spectrum and rutin (a) content and the pre-treated spectrum and rutin (b), isoquercetin (c), hyperoside (d), quercetin (e) and hypericin (f) content in the quantitative model of medicinal material content constructed in this application;

[0031] Figure 8 A relationship diagram between the true value and predicted value of the quantitative model of rutin (a), isoquercetin (b), hyperoside (c), quercetin (d) and hypericin (e) in the quantitative model of medicinal material content provided in this application;

[0032] Fig. 9 The average spectrum of the Pueraria lobata and Pueraria lobata samples used in the quantitative model of medicinal material content construction provided in this application;

[0033] Fig.10 In the quantitative model of medicinal material content provided in this application, 1 st D. Average spectrum after processing;

[0034] Fig.11 PCA visualization distribution diagram of Pueraria root and Pueraria lobata samples in constructing the quantitative model of medicinal material content provided in this application;

[0035] Fig.12 The predicted response values ​​of Pueraria root and Pueraria lobata in the quantitative model of medicinal material content construction provided in the present application, wherein (a) is the predicted response value of the Pueraria root group, and (b) is the predicted response value of the Pueraria lobata group;

[0036] Fig.13 The response coefficients of Pueraria root and Pueraria lobata in the quantitative model of medicinal material content provided in the present application, wherein (a) is the response coefficient of the Pueraria root group, and (b) is the response coefficient of the Pueraria lobata group;

[0037] Fig.14 A scatter plot of the predicted values ​​and measured values ​​of the PLS model in the quantitative model for building medicinal material content provided in this application;

[0038] Fig.15This is a scatter plot of the LS-SVR model predicted values ​​and measured values ​​in the quantitative model for medicinal material content constructed in this application.

[0039] Icons: 100-material quality rapid detection and sorting equipment; 110-machine body; 111-feed port; 112-discharge port; 1121-first outlet; 1122-second outlet; 113-conveyor belt; 1131-first conveyor belt; 1132-second conveyor belt; 1133-third conveyor belt; 114-speed knob; 115-high-grade material bin; 116-low-grade material bin; 120-feeding mechanism; 121-feeding hopper; 122-screw feeder; 123-motor; 130-spectrometer; 140-rejector; 150-display. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0041] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0043] In the description of the present invention, it should be noted that if the terms "upper", "lower", "inside", "outside", etc. appear to indicate an orientation or position relationship, they are based on the orientation or position relationship shown in the accompanying drawings, or are the orientation or position relationship in which the product of the invention is usually placed when used. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0044] In addition, the terms “first”, “second”, etc., if used, are merely used to distinguish between the descriptions and should not be understood as indicating or implying relative importance.

[0045] It should be noted that, in the absence of conflict, the features in the embodiments of the present invention may be combined with each other.

[0046] Please refer to Figure 1 This embodiment provides a material quality rapid detection and sorting device 100, which includes a body 110, a feeding mechanism 120, a spectrometer 130, a rejector 140 and a display 150.

[0047] The machine body 110 is provided with a feed port 111 and a discharge port 112. A conveyor belt 113 for transmitting materials is provided between the feed port 111 and the discharge port 112. The conveyor belt 113 can control the material transmission process. Different materials have different properties, and different spectrometers 130 have different detection times. Therefore, the number of conveyor belts 113 in the present application can be 1-5, and there are multiple arrangements. Specifically, in the present embodiment, the number of conveyor belts 113 is 3, namely, the first conveyor belt 1131, the second conveyor belt 1132 and the third conveyor belt 1133. The first conveyor belt 1131, the second conveyor belt 1132 and the third conveyor belt 1133 are arranged from top to bottom in opposite directions, which can effectively extend the material transmission path and provide a buffer time interval for data transmission of the rejector 140. The transmission speed and transmission direction of the conveyor belt 113 can be adjusted or closed respectively, and the overall equipment occupies a small space, which is convenient for storage. Preferably, the transmission speed of the conveyor belt 113 is 0.1-0.5 m / s.

[0048] Furthermore, in order to more conveniently adjust the conveying speed of the conveyor belt 113 , the present application also provides a speed regulating knob 114 in the body 110 , and the speed regulating knob 114 can control the conveying speed of the three conveyor belts 113 at the same time.

[0049] The feeding mechanism 120 includes a feeding hopper 121, a screw feeder 122 and a motor 123. The feeding hopper 121 is installed at the feeding port 111 of the machine body 110. The screw feeder 122 is vertically installed in the feeding hopper 121. One end of the screw feeder 122 is connected to the motor 123, and the other end extends into the bottom of the feeding hopper 121. When the motor 123 is turned on, the screw feeder 122 rotates in the feeding hopper 121, stirring the materials to break up the sticky medicinal materials, and the screw feeder 122 is conducive to the uniform transmission of the Chinese medicine materials, which is convenient for subsequent sorting, avoids the medicinal materials from agglomerating together, and controls the uniformity of the material transmission. At the same time, the bottom pipe of the feeding hopper 121 is narrow, which can realize slow feeding and avoid too much medicinal materials from entering the conveyor belt 113.

[0050] The spectrometer 130 is installed in the body 110 and located above the conveyor belt 113. The spectrometer 130 is detachably connected to the body 110. It can be replaced according to needs. For example, the spectrometer 130 can be a portable near-infrared spectrometer 130, a portable ultraviolet spectrometer 130 or a hyperspectral imager.

[0051] The rejector 140 is installed in the body 110 and is located at the end of the conveyor belt 113. The spectrometer 130 is in communication connection with the rejector 140. The discharge port 112 includes a first outlet 1121 and a second outlet 1122. The rejector 140 selectively opens the first outlet 1121 or the second outlet 1122. In the present application, the rejector 140 is detachably connected to the body 110. There are many options for the rejector 140, which can be replaced according to the properties of the material, for example, it can be a blowing rod, a stripping rack, etc.

[0052] The first outlet 1121 and the second outlet 1122 can be distinguished according to the quality or authenticity of the materials, for example, they can be divided into the high-quality first outlet 1121 and the low-quality second outlet 1122. The rejector 140 is equipped with a signal receiving device, which can quickly and effectively receive the identification results and commands of the spectrometer 130, and sort the corresponding medicinal materials. If the quality is good, the first outlet 1121 is opened to enter the high-quality material warehouse 115. If the quality is poor, the second outlet 1122 is opened to enter the low-quality material warehouse 116, so as to realize the sorting of the authenticity of the Chinese medicine materials.

[0053] In the present application, the inclined guide position of the first outlet 1121 corresponds to the preset position of the high-grade material bin 115 outside the body 110 , and the inclined guide position of the internal second outlet 1122 corresponds to the preset position of the low-grade material bin 116 outside the body 110 .

[0054] The display 150 is installed outside the body 110. The display 150 is connected to the spectrometer 130 and the rejector 140 at the same time. The display 150 can reflect the material quality information in real time, realize rapid detection and sorting of material quality, and effectively distinguish and separate authentic and superior medicinal materials.

[0055] It should be noted that the spectrometer 130 of the present application is introduced with a qualitative model for adulteration screening or a quantitative model for medicinal material content, so that qualitative or quantitative analysis can be performed on different materials as needed.

[0056] The method for constructing a quantitative model for medicinal material content includes: using a hyperspectral imager to collect spectra of medicinal material samples, collecting 8-12 spectra for each batch of samples, and calculating the average value as the hyperspectral data of the medicinal material samples; optimizing the hyperspectral data and combining the partial least squares method to establish a quantitative model for medicinal material content; wherein, the optimization of the hyperspectral data includes: using the RHM and MCCV methods to eliminate abnormal samples existing in the original spectra and the known index component content data, and determining the effective samples to establish a quantitative prediction model for each index component; using 1 st D, SNV, PSNV, or 1 stThe D+PSNV method was used to preprocess the spectral reflectance data to improve the signal-to-noise ratio of the spectrum; the correlation coefficient between the preprocessed spectrum and the content of known index components was analyzed; and the wavelength was screened using SPA analysis method, correlation coefficient method or full spectrum analysis method.

[0057] Specifically, taking Hypericum perforatum as an example, a quantitative model of medicinal material content is constructed, which includes the following steps:

[0058] S101, Materials

[0059] 154 batches of field samples of Hypericum perforatum came from the Lixian Demonstration Base in Gansu, China. They are perennial herbaceous plant Hypericum perforatum L. of the Garciniaaceae family.

[0060] S102, Hypericum perforatum hyperspectral collection

[0061] The portable hyperspectral instrument collected the spectrum of the canopy of Hypericum perforatum with sampling intervals of 1.4 nm (350 nm-1000 nm) and 1.1 nm (1001 nm-2500 nm). The instrument field of view was 25°, the field of view was a circle with a radius of 15 cm, and the fiber optic probe was 32 cm above the canopy (e.g. Figure 2 ). Before measurement, the standard reference plate needs to be calibrated. When the reflectivity value is a flat line of 1.00, the probe is measured vertically downward. 10 spectra are collected for each batch of samples, and the average value is taken as the spectral reflectivity of the sample.

[0062] S103. Results and Analysis

[0063] (1) Eliminate abnormal samples

[0064] RHM and MCCV methods were used to remove abnormal samples in the original spectrum and the content data of the five index components. The results are as follows: Figure 3 , Figure 4 Nine batches of abnormal samples were screened out from the original spectral data, and the number of abnormal samples in the contents of rutin, isoquercetin, hyperoside, quercetin and hypericin were 13, 13, 12, 6 and 13 batches, respectively. Finally, the number of valid samples was determined to be 132, 132, 133, 139 and 132 batches, respectively, to establish the quantitative prediction model of each index component.

[0065] (2) Spectral preprocessing

[0066] The hyperspectral curves of 154 batches of original canopy of Hypericum perforatum are as follows Figure 5 As shown in (a), it can be seen from the figure that the signal-to-noise ratio of the start and end bands is low, so the wavelength variables of 350nm-409nm and 2441nm-2500nm are removed before modeling, and the spectrum obtained is as follows Figure 5 As shown in (b).

[0067] Use 1 st D, SNV, PSNV and 1 st The D+PSNV method was used to preprocess the spectral data, and the comparison results are shown in Table 1. As can be seen from the table, for the prediction of the content of the five index components, all models achieved relatively ideal results, and the Rc and Rp values ​​were both higher than 0.80. st The model with D+PSNV preprocessed spectra had the best effect. The content prediction models of rutin (RMSEP=0.0922, Rp=0.9564), isoquercetin (RMSEP=0.0597, Rp=0.9914), hyperoside (RMSEP=0.0588, Rp=0.9710), quercetin (RMSEP=0.0191, Rp=0.8619) and hypericin (RMSEP=0.0092, Rp=0.9929) all obtained the smallest RMSEP and the largest Rp. Figure 6 As shown in the figure, it is found that the pre-processed spectrum is significantly different from the original spectrum, the spectrum overlap is higher, and the influence of baseline drift is significantly eliminated. st The effectiveness of the D+PSNV method in processing hyperspectral data was analyzed. The correlation coefficients between the pre-processed spectra and the contents of each index component were analyzed. Figure 7 , Figure 7 The maximum absolute value of the correlation coefficient between the original spectrum of a and rutin content is less than 0.2, while the correlation coefficient after pretreatment increases significantly, with the maximum absolute value exceeding 0.5 ( Figure 7 b), indicating that the signal-to-noise ratio of the spectrum can be improved after spectral pretreatment. The correlation coefficients of isoquercetin, hyperoside, quercetin and hypericin with the pretreated spectrum are shown in Figure 7 In cf, both showed a large correlation.

[0068] Table 1. Effects of different preprocessing methods on the PLS quantitative model of five index components

[0069]

[0070]

[0071] (3) Wavelength variable selection

[0072] In order to reduce the redundancy and collinearity of hyperspectral data, simplify the computational complexity and improve the model operation efficiency, two variable selection methods are summarized in Table 2 to select appropriate wavelength variables for the content prediction models of rutin, isoquercetin, hyperoside, quercetin and hypericin. Among them, the PLS prediction model established by SPA for the five index components has better performance, all of which obtain the smallest RMSEP and the largest Rp. Compared with the number of wavelengths in the full spectrum, the number of wavelengths of rutin, isoquercetin, hyperoside, quercetin and hypericin is reduced by 96.45%, 96.11%, 96.75%, 96.36% and 96.36%, respectively. The wavelengths screened by SPA show good potential, making the established quantitative analysis model extremely accurate and greatly simplifying the complexity of the model.

[0073] Table 2. Comparison of different wavelength variable selection methods

[0074]

[0075] (4) Construction of quantitative models for five index components

[0076] Based on the above optimization process, the quantitative detection models of rutin, isoquercetin, hyperoside, quercetin and hypericin were established. The regression results of each calibration model are shown in Table 3 and Figure 8 ,It can be seen that the predicted values ​​of quality indicators such as rutin, isoquercetin, hyperoside, quercetin and hypericin content are consistent with the reference values, the RMSEC and RMSEP of each model are low, and the Rc and Rp are high, indicating that all five models have good prediction and generalization performance.

[0077] Table 3. Results of PLS ​​quantitative model for five index components

[0078] Serial number Indicator ingredients RMSEC Rc RMSEP R 1 Rutin 0.0812 0.9466 0.0922 0.9564 2 Isoquercetin 0.1160 0.9433 0.0597 0.9914 3 Hyperoside 0.0561 0.9288 0.0588 0.9710 4 Quercetin 0.0191 0.8625 0.0191 0.8619 5 Hypericin 0.0280 0.9054 0.0092 0.9929

[0079] The quantitative detection model of rutin, isoquercetin, hyperoside, quercetin and hypericin was established by using the hyperspectral data of Hypericum perforatum combined with the partial least squares method. st After D+PSNV optimal preprocessing and SPA spectral variable selection, the five quantitative models all have good prediction accuracy. Compared with traditional quality detection methods and modern near-infrared spectroscopy technology, hyperspectral technology not only shows strong advantages in non-destructive, rapid and real-time detection, but also can effectively combine the intrinsic quality attributes of samples with external spatial attributes, fully and accurately evaluate the quality of samples, better guarantee the quality of raw medicinal materials, and has important significance for the intelligent development of the traditional Chinese medicine industry.

[0080] Among them, the method for constructing a qualitative model for adulteration screening includes: identifying the macroscopic and microscopic characteristics of authentic samples and fake samples respectively, and then crushing the authentic samples and fake samples to prepare adulterated samples with different contents of authentic samples; collecting near-infrared spectra of multiple adulterated samples, recording them continuously for multiple times, and calculating the average spectrum; applying PCA to the average spectrum to visualize the sample distribution, and using the Kennard-Stone method to divide multiple adulterated samples into a training set for model calibration and a test set for external validation to screen samples; using the PLS-DA analysis method to classify the screened samples into authentic samples and fake samples; and using PLS or LS-SVM to quantify the adulterated samples to predict the adulteration rate.

[0081] Specifically, a qualitative adulteration screening model was constructed by taking the rapid screening of adulteration of Pueraria root and Pueraria lobata as an example, which includes the following steps:

[0082] The specific construction methods include:

[0083] S201. Sample preparation

[0084] Pueraria root (n = 20 samples) and Pueraria lobata (n = 20 samples) were obtained from Sichuan Province, China. All plant materials were identified and authenticated by comparing macroscopic and microscopic characteristics. All fresh materials were cut into 1 cm 3 Cubes were prepared and dried in a drying oven at 60°C for 12 hours to constant weight. The dried samples were crushed and passed through a 50-mesh sieve to make the sample powder evenly distributed and stored in a desiccator at room temperature. For the purpose of adulteration level prediction analysis, 6 different batches of kudzu root and kudzu vine were randomly selected to prepare adulterated samples. The mixing ratio of kudzu root and kudzu vine is shown in Table 4. A total of 126 samples were obtained, including 114 mixed samples, 6 pure kudzu root and 6 pure kudzu vine test samples.

[0085] Table 4. Information of two adulterated Pueraria root datasets

[0086]

[0087] S202, NIR spectrum acquisition

[0088] Under normal temperature conditions, turn on the machine and preheat for 1 hour; take about 2g of powder sample and put it into the rotating quartz cup, shake it evenly, and the sample thickness must not be less than 10mm. The main working conditions of the instrument are: ambient temperature is 20-25℃, relative humidity is 40%-50%; sample measurement method: integrating sphere diffuse reflection, air as reference, resolution 8cm -1 , scanning wavelength range 4000~10000cm -1 To improve the measurement accuracy, the spectrum of each measurement was recorded continuously by scanning 64 times, repeated 3 times for each sample, and the average spectrum was calculated for further analysis.

[0089] S203, Spectral Analysis

[0090] The average spectra obtained from the kudzu vine and kudzu root samples are shown in Fig. 9 As shown. At 8370cm -1 The absorbance at 6310 cm is the second overtone of CH stretching. -1 and 6850cm -1 Related to the first overtone of OH stretching. 5400-6000cm -1 It is the first overtone of CH stretching of CH and CH2. -1 The peak at 4760cm comes from the stretching and deformation vibration of OH. -1 The peak at comes from the combination of OH deformation and CH stretching.

[0091] At 4000cm -1 The strongest absorption band observed near CH is caused by the stretching and deformation combination CH (4020 cm -1 ) and CH2(4319cm -1 However, it is challenging to observe clear differences between Pueraria lobata and Pueraria root. st The spectrum after D processing is as follows Fig.10 As shown, differences between the samples are identified, especially in terms of peak intensity. These plots are at 7500 cm -1 The above range shows a very similar shape. For 6000cm -1 Below this wave number, differences become apparent, which can serve as a basis for source identification. At about 5100 cm -1 and 5260cm -1 The difference at is believed to be a combination of trace amounts of water OH stretching, as there may be hardly any difference in moisture content between the samples.

[0092] The main components of the dry material are carbohydrates, including starch, water-soluble carbohydrates and cellulose. It is reported that the starch accumulation of roots is higher than that of vines. In addition, at 5700-6000cm -1 and 4200-4400cm -1 The wide absorption differences among regions may be caused by the CH groups in starch.

[0093] S204, Rapid Identification of Pueraria lobata and Pueraria lobata

[0094] PCA

[0095] As mentioned above, the first part of the study was about developing a fast method suitable for distinguishing Pueraria root from Pueraria lobata. In order to get a clearer overview of the dataset, PCA was applied to visualize the sample distribution. Fig.11As can be seen, the two categories are almost separate, but there is some overlap. The first and second principal components account for 90% of the total variance, and by including the third principal component, the explained variance reaches 98.7%.

[0096] PLS-DA

[0097] The Kennard-Stone algorithm was used to divide the 40 samples into a training set and a test set. The training set was used for model calibration, and the test set was only used for external validation. For PLS-DA, the optimal number of latent variables determined by the leave-one-out cross-validation (LOOCV) method for the calibration samples was only 3. Parameters such as error rate, classification specificity, and sensitivity were used to evaluate the classification effect. As shown in Table 5, the PLS-DA model provided 100% prediction accuracy for Pueraria and Pueraria lobata. The results are also shown graphically in Fig.12 In the figure, the predicted response of each class and all samples are plotted. The red dashed line represents the discrimination threshold. As can be seen from the figure, the Pueraria root and Pueraria lobata samples can be clearly separated without sample separation errors.

[0098] Table 5 Classification and evaluation results of Pueraria root and Pueraria lobata

[0099]

[0100] In order to understand which spectral variables contribute most to the classification, the regression coefficients of the variables for each class were calculated and studied. Coefficients with high absolute values ​​are associated with important variables for class discrimination, and variables with positive coefficients contribute to an increase in the class calculation response. Fig.13 As shown, the following conclusions are drawn: The spectral range of interest (with high absolute values):

[0101] ①4250cm -1 and 4400cm -1 Nearby area (C–H combination)

[0102] ②5150cm -1 Large range centered on (OH combination)

[0103] C–H and O–H groups belong to starch, flavonoids, cellulose, and other water-soluble carbohydrates. These two ranges showed negative and positive contributions, respectively, to the kudzu-like behavior. The latter range associated with kudzu-like gave the most important weight to the reaction as expected, which is compatible with kudzu chemistry, as kudzu is rich in starch and flavonoids. Fig.13 Comparing (a) and (b), we can assume that the variables associated with positive coefficients are higher in samples of category 1 (Kuard), while the predicted values ​​corresponding to negative coefficients are higher in individuals of category 2 (Kuard).

[0104] S205, Identification of adulteration rates of kudzu root and kudzu vine

[0105] PLS

[0106] Four different signal preprocessings were applied, and the model performance is shown in Table 6. The number of latent variables for each PLS model was determined by 10-fold cross validation, and the minimum root mean square error of cross validation (RMSECV) was selected to optimize the prediction model to avoid overfitting and underfitting. Therefore, as shown in Table 6, 1 combined with SG smoothing st D obtained the lowest RMSECV value and outperformed other pretreatments while providing good prediction performance, obtaining a high Rp value of 0.9933 and a low RMSEP value of 3.3754.

[0107] Table 6 Prediction performance of kudzu root and kudzu vine models under four different pretreatment methods

[0108]

[0109] LS-SVM

[0110] Gaussian radial basis function (RBF) is selected as the kernel function of LS-SVM. Therefore, the performance of the LS-SVM model depends on two parameters: the regularization parameter γ and the kernel parameter σ. The regularization parameter γ determines the trade-off between minimizing the training error and minimizing the model complexity. The kernel parameter σ is the bandwidth of the ε-insensitive region used to fit the training data. To select the optimal parameters of the LS-SVM model, the LOOCV grid search technique is used to avoid blindness and randomness. γ and σ 2 The best combination is 1.0796×10 6 and 7.4206×10 4 The prediction results are shown in Table 7.

[0111] Table 7 Comparison of prediction results of different quantitative models

[0112] <![CDATA[R c ]]> RMSEC <![CDATA[R p ]]> RMSEP Bias RPD PLS 0.9914 4.0308 0.9933 3.3754 -0.1026 8.7130 LS-SVM 0.9991 1.3065 0.9973 2.1339 0.2611 13.6818

[0113] The R values ​​of the LS-SVM model for the calibration and validation sets were 0.9990 and 0.9972, both higher than those of the PLS model. In addition, the calibration and validation set root mean square (RMSE) values ​​of the LS-SVM were lower than those of the PLS dataset. The high R values ​​and low RMSE values ​​indicate that the LS-SVM model has good prediction and generalization performance.

[0114] The scatter plots of the predicted values ​​and actual values ​​of the calibration set and validation set of the PLS and LS-SVM models are shown in Figure 2. Fig.14 and Fig.15 Compared with the PLS model, the samples predicted by LS-SVM are closer to the regression line, indicating that the LS-SVM regression model has better ability than the PLS model in predicting the adulteration rate between Pueraria root and Pueraria lobata.

[0115] In this application, near infrared spectroscopy combined with machine learning algorithms provides a fast and effective tool to distinguish and quantify adulterated samples of Pueraria root. The results show that the PLS-DA method can be used to classify the cheap analog of Pueraria root, Pueraria lobata, with a classification accuracy of 100%. In addition, adulterated Pueraria root samples can be quantified by PLS and LS-SVM, and LS-SVM shows better performance than PLS and shows a higher Rp value of 0.9972 and a low RMSEP value of 2.1724. This is the first study to distinguish and quantify adulterated Pueraria root by machine learning methods, which can provide new options for the comprehensive quality assessment of Pueraria root.

[0116] The following describes it in detail with reference to specific embodiments.

[0117] Example 1

[0118] Rapid sorting of kudzu root and its counterfeits using portable near-infrared spectroscopy

[0119] (I) Preparation stage: Construction of the authenticity discrimination model of Pueraria root medicinal materials: 1. Sample preparation: Six different batches of Pueraria root and Pueraria lobata were randomly selected to prepare adulterated samples. Pueraria root and Pueraria lobata were mixed according to mass percentage (10, 20, 30, 40, 50, 60, 70, 80, 90%; w / w), and a total of 102 mixed samples were obtained; 2. Spectral acquisition: The average spectrum of Pueraria root and its counterfeit (Pueraria lobata) was collected using a portable near-infrared spectrometer for further analysis; 3. Model construction: The first-order derivative (1st D) was used for spectral preprocessing; the partial least squares discriminant analysis (PLS-DA) was used to construct the discrimination model; the LS-SVM algorithm was used to construct the quantitative adulteration rate prediction model of Pueraria root and Pueraria lobata.

[0120] (ii) Install a portable near-infrared spectrometer and import the constructed kudzu root and vine discrimination model into the computer terminal of the sorting equipment.

[0121] (III) Feeding hopper 121, setting the transmission speed of conveyor belt 113 to 0.2-0.5m / s, setting the kudzu root purity value range (90%-100%), if the acquired value is lower than 90%, it will turn red, indicating low-grade material, if the acquired value is higher than 90%, it will turn green, indicating high-grade material. When the material conveyor belt 113 moves under the near-infrared spectrometer, the near-infrared spectrometer collects spectral information of the material passing through, and then the display 150 displays the material quality information (whether it is adulterated, adulteration rate, etc.), and judges whether the kudzu root medicinal material meets the standard. If it does not meet the standard, the information is transmitted to the rejector 140, and the material transmission speed is calculated. For example, after 10s, when a sample with an unqualified adulteration ratio passes through, the second outlet 1122 is opened and enters the low-grade material bin 116. If it meets the standard, the second outlet 1122 is not opened, and the material enters the high-grade material bin 115. The authenticity of the material is sorted.

[0122] Example 2

[0123] Rapid quality sorting of Hypericum perforatum using hyperspectral imaging technology

[0124] (I) Preparation stage: 1. Sample preparation: 100 batches of Hypericum perforatum samples were randomly selected. 2. Spectral acquisition: Hyperspectral images of Hypericum perforatum were collected using a hyperspectral imaging spectrometer, the collected hyperspectral images were calibrated, and the average spectrum of all pixels was taken for further analysis. 3. Model construction: The first-order derivative combined with the piecewise standard normal variable transformation (1st D + SG) was used for spectral preprocessing; the SPXY method was used to divide the sample set into a calibration set and a validation set; the continuous projection algorithm (SPA) was used to screen the wavelength variables; finally, the partial least squares regression method (PLS) was used to establish the quantitative model of the effective components of Hypericum perforatum, rutin, isoquercitrin, hyperoside, quercetin and hypericin.

[0125] (ii) Install a hyperspectral imaging spectrometer and import the constructed quantitative models of the effective components of Hypericum perforatum, such as rutin, hyperoside and hypericin, into the computer terminal of the sorting equipment.

[0126] (III) Feeding material from the feed hopper 121, setting the transmission speed of the conveyor belt 113 to 0.1-0.3 m / s, setting the content range of each effective component (rutin ≥ 0.45%, hyperoside ≥ 0.30%, hypericin ≥ 0.03%), if the content value is less than the set range, it will turn red, indicating that it is a low-grade material, if the content value is greater than the set range, it will turn green, indicating that it is a high-grade material. When the material conveyor belt 113 moves under the hyperspectral imaging instrument, the hyperspectral imaging instrument collects image information of the material passing through, and then the display 150 displays the content of the effective components of rutin, hyperoside and hypericin in the material to determine whether the Hypericum perforatum medicinal material meets the standard. If it does not meet the standard, the information is transmitted to the rejector 140 to calculate the material transmission speed. For example, after 5 seconds, when a sample with unqualified medicinal material content (rutin ≤ 0.45%, hyperoside ≤ 0.30%, hypericin ≤ 0.03%) passes through, the second outlet 1122 is opened and enters the low-grade material bin 116. If it meets the standard (rutin ≥ 0.45%, hyperoside ≥ 0.30%, hypericin ≥ 0.03%), the second outlet 1122 is not opened, and the material enters the high-grade material bin 115. The quality of the material is sorted.

[0127] In summary, the material quality rapid detection and sorting equipment 100 provided in the present application installs a detachable spectrometer 130 on the conveyor belt 113. The spectrometer 130 can be selected according to actual needs. The rejector 140 identifies the authenticity and quality of the materials on the conveyor belt 113 according to the collected spectral data, and then sorts them. The quality can be sorted based on the quality of the medicinal materials, which greatly reduces the uncertainty and time-consuming nature of manual inspection. It makes up for the defects of the current single sorting standards such as size and weight, and provides a more powerful quality assurance for the sorting of traditional Chinese medicine materials. The traditional Chinese medicine material production system provided in the present application includes the above-mentioned material quality rapid detection and sorting equipment 100. By introducing the above-mentioned material quality rapid detection and sorting equipment 100 into the traditional Chinese medicine material production system, the quality control of the traditional Chinese medicine materials produced by the traditional Chinese medicine material production system can be better.

[0128] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A material quality rapid detection and sorting equipment, characterized in that: It includes a body, a spectrometer and a rejector, the body is provided with a feed port and a discharge port, the body is provided with a conveyor belt located between the feed port and the discharge port for transmitting materials, the spectrometer is installed in the body and located above the conveyor belt, the rejector is installed in the body and located at the end of the conveyor belt, the spectrometer is communicatively connected with the rejector, the discharge port includes a first outlet and a second outlet, and the rejector selectively opens the first outlet or the second outlet; The material quality rapid detection and sorting equipment further comprises a display, which is installed outside the body and is connected to the spectrometer and the rejector at the same time; The spectrometer is introduced with a qualitative model for adulteration screening or a quantitative model for medicinal material content; The method for constructing the qualitative model for adulteration screening includes: respectively identifying the macroscopic and microscopic features of the true sample and the false sample, and then crushing the true sample and the false sample to prepare the adulterated samples with different contents of the true sample; collecting near-infrared spectra of multiple adulterated samples, recording them continuously for multiple times, and calculating the average spectrum; applying PCA to the average spectrum to visualize the sample distribution, and using the Kennard-Stone method to divide the multiple adulterated samples into a training set for model calibration and a test set for external verification to screen samples; using the PLS-DA analysis method to classify the screened samples into true samples and false samples; and using PLS or LS-SVM to quantify the adulterated samples to predict the adulteration rate.

2. The material quality rapid detection and sorting equipment according to claim 1 is characterized in that: The spectrometer is a portable near-infrared spectrometer, a portable ultraviolet spectrometer or a hyperspectral imager.

3. The material quality rapid detection and sorting equipment according to claim 1 is characterized in that: The spectrometer is detachably connected to the body.

4. The material quality rapid detection and sorting equipment according to claim 1 is characterized in that: The method for constructing the quantitative model of medicinal material content comprises: using a hyperspectral imager to collect spectra of medicinal material samples, collecting 8-12 spectra for each batch of samples, and calculating the average value as the hyperspectral data of the medicinal material samples; optimizing the hyperspectral data and combining the partial least squares method to establish the quantitative model of medicinal material content; The optimization of the hyperspectral data includes: using RHM and MCCV methods to remove abnormal samples in the original spectrum and the known index component content data, determining valid samples to establish a quantitative prediction model for each index component; using 1 st D, SNV, PSNV, or 1 st The D+PSNV method was used to preprocess the spectral reflectance data to improve the signal-to-noise ratio of the spectrum; the correlation coefficient between the preprocessed spectrum and the content of known index components was analyzed; and the wavelength was screened using SPA analysis method, correlation coefficient method or full spectrum analysis method.

5. The material quality rapid detection and sorting equipment according to claim 4 is characterized in that: Use 1 st The D+PSNV method is used to preprocess the spectral reflectance data.

6. The material quality rapid detection and sorting equipment according to claim 4 is characterized in that: The wavelengths were screened using SPA analysis.

7. The material quality rapid detection and sorting equipment according to claim 1 is characterized in that: The conveyor belt comprises a plurality of conveyor belts.

8. The material quality rapid detection and sorting equipment according to claim 7 is characterized in that: The transmission speed of the conveyor belt is 0.1-0.5 m / s.

9. The material quality rapid detection and sorting equipment according to claim 1 is characterized in that: The material quality rapid detection and sorting equipment also includes a feeding mechanism, which includes a feeding hopper, a screw feeder and a motor. The feeding hopper is installed at the feeding port of the machine body, and the screw feeder is vertically installed in the feeding hopper. One end of the screw feeder is connected to the motor, and the other end extends into the bottom of the feeding hopper.

10. A Chinese medicine material production system, characterized in that: It includes the material quality rapid detection and sorting equipment as described in any one of claims 1-9.

Images