A rice origin tracing method

Abstract

The invention belongs to the field of rice origin tracing, and specifically relates to a rice origin tracing method; the method can realize accurate tracing of rice; the method obtains rice sample powder spectrum data through an X-ray fluorescence spectrometer, processes the rice sample powder spectrum data to obtain a processed spectrum characteristic curve; according to the processed spectrum characteristic curve, a rice origin tracing model is trained through an orthogonal partial least squares discriminant analysis method, a principal component analysis method, a random forest algorithm or a support vector machine model to obtain a trained rice origin tracing model; rice of unknown origin is traced through the trained rice origin tracing model, and the rice origin is output; the method does not need to be carried out by professionals in specific experimental operations, and does not need to establish a standard curve, and only needs to collect rice samples to obtain the processed spectrum characteristic curve of the rice, thereby improving the convenience of the rice origin tracing method.

Description

Technical Field

[0001] The invention belongs to the field of rice origin tracing, and in particular relates to a rice origin tracing method. Background Art

[0002] Due to the large differences in climate and geographical conditions for rice cultivation, rice is grown from the northeast to Hainan, from the plains of the middle and lower reaches of the Yangtze River to the plateaus of the northwest, thus forming rice products with different characteristics. Various rice origin traceability technologies have also emerged, among which multi-element analysis technology is the most widely used. In addition, rice traceability technologies such as gas chromatography, liquid chromatography, and isotope analysis technology are also included. However, these traditional technologies often require complex pretreatment, extremely professional laboratory equipment and personnel to operate, and are extremely unfriendly to the environment. With the development of science and technology, green non-destructive analysis technologies such as hyperspectral imaging, infrared spectroscopy, Raman spectroscopy, and sensor arrays have been applied to the field of rice origin traceability and identification. However, these technologies also have certain defects. Infrared spectroscopy and Raman spectroscopy technologies may be affected by sample storage, processing and other factors that may affect the ingredients in food; when hyperspectral imaging collects images, the signal may be affected by the surrounding environment (such as scattering, lighting, etc.); the sensor array shows poor sensitivity and selectivity for similar substances or taste attributes, and the temperature level, humidity and other background odors will destroy the function of the sensor array; in summary, the rice traceability technology in the existing technology cannot achieve accurate traceability of rice. Summary of the invention

[0003] In order to solve the problem that the rice traceability technology in the prior art cannot accurately trace the rice, a rice origin traceability method is proposed.

[0004] A rice origin tracing method, comprising:

[0005] Step 1: Pre-treating rice samples from different known origins to obtain rice sample powder;

[0006] Step 2: Use an X-ray fluorescence spectrometer to scan the rice sample powder at different voltages and different filter conditions for different time periods to obtain a plurality of rice sample powder spectrum data; each spectrum data includes a plurality of energy spectrum intensity data, and each energy spectrum intensity data corresponds to an energy characteristic data;

[0007] Step 3: using the energy characteristic data as the horizontal coordinate and the energy spectrum intensity data as the vertical coordinate, constructing a plurality of spectrum characteristic curves corresponding to the plurality of spectrum data respectively according to the plurality of energy spectrum intensity data in each spectrum data and the energy characteristic data corresponding to each energy spectrum intensity data;

[0008] Step 4: intercepting multiple spectral characteristic curves according to different energy characteristic data segments to obtain multiple spectral characteristic curve segments, and concatenating and smoothing the obtained multiple spectral characteristic curve segments to obtain a processed spectral characteristic curve;

[0009] Step 5: Establish a rice origin traceability model, and train the rice origin traceability model according to the processed spectral characteristic curve to obtain a trained rice origin traceability model;

[0010] Step 6: Obtain the processed spectral characteristic curve of the rice from unknown origin, input the processed spectral characteristic curve of the rice from unknown origin into the trained rice origin tracing model, and output the rice origin.

[0011] Beneficial Effects

[0012] The invention proposes a rice origin tracing method, which can realize accurate tracing of rice. The method obtains rice sample powder spectral data through an X-ray fluorescence spectrometer, processes the rice sample powder spectral data to obtain a processed spectral characteristic curve; trains a rice origin tracing model through an orthogonal partial least squares discriminant analysis method, a principal component analysis method, a random forest algorithm or a support vector machine model according to the processed spectral characteristic curve to obtain a trained rice origin tracing model; traces rice of unknown origin through the trained rice origin tracing model, and outputs the rice origin; the method does not need to be carried out by professionals in specific experimental operations, and does not need to establish a standard curve. It only needs to collect rice samples to obtain the processed spectral characteristic curve of the rice, thereby improving the convenience of the rice origin tracing method. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a flow chart of the rice origin tracing method according to the specific implementation method of this application;

[0014] Figure 2 This is a score graph of orthogonal partial least squares discriminant analysis of rice samples of known origin according to a specific implementation method of the present application;

[0015] Figure 3 A linear discriminant function score graph of rice samples of known origin according to a specific embodiment of the present application;

[0016] Figure 4 This is a graph showing the relationship between the error rate and the number of decision trees in the support vector machine algorithm of the specific implementation mode of this application. DETAILED DESCRIPTION

[0017] Specific implementation method 1: The following will be combined with the attached embodiment of the present invention Figure 2 To Attachment Figure 4 , illustrate this implementation mode, and clearly and completely describe the technical solutions in the embodiments of the present invention:

[0018] A rice origin tracing method, comprising:

[0019] Step 1: Pre-treating rice samples from different known origins to obtain rice sample powder;

[0020] Step 2: Use an X-ray fluorescence spectrometer to scan the rice sample powder at different voltages and different filter conditions for different time periods to obtain a plurality of rice sample powder spectrum data; each spectrum data includes a plurality of energy spectrum intensity data, and each energy spectrum intensity data corresponds to an energy characteristic data;

[0021] Step 3: using the energy characteristic data as the horizontal coordinate and the energy spectrum intensity data as the vertical coordinate, constructing a plurality of spectrum characteristic curves corresponding to the plurality of spectrum data respectively according to the plurality of energy spectrum intensity data in each spectrum data and the energy characteristic data corresponding to each energy spectrum intensity data;

[0022] Step 4: intercepting multiple spectral characteristic curves according to different energy characteristic data segments to obtain multiple spectral characteristic curve segments, and concatenating and smoothing the obtained multiple spectral characteristic curve segments to obtain a processed spectral characteristic curve;

[0023] Step 5: Establish a rice origin traceability model, and train the rice origin traceability model according to the processed spectral characteristic curve through orthogonal partial least squares discriminant analysis, principal component analysis, random forest algorithm or support vector machine algorithm to obtain a trained rice origin traceability model;

[0024] Step 6: Obtain the processed spectral characteristic curve of the rice from unknown origin, input the processed spectral characteristic curve of the rice from unknown origin into the trained rice origin tracing model, and output the rice origin.

[0025] Specifically, representative rice samples were collected on-site at the rice producing areas. The rice used was collected from four regions, namely Yilan County and Ning'an City in Heilongjiang Province, Dazhou City in Sichuan Province and Qiandongnan in Guizhou Province, including 30 samples from Yilan County (YLX), 20 samples from Ning'an City (NAS), 31 samples from Dazhou City (DZS), and 30 samples from Qiandongnan (QDN), totaling 111 samples.

[0026] Furthermore, the rice samples with different known origins were pretreated, including: drying the rice samples at 60° C. for 20 hours, and crushing them for 2 minutes after cooling.

[0027] Specifically, the rice sample was dried at 60°C for 20 hours to constant weight, and then directly crushed in a grinder for 2 minutes after cooling. 10 g of powder was accurately weighed using a 1 / 100,000 balance, and then directly loaded into a US Chemplex 40 mm XRF sample cup, and compacted using a US Chemplex 250 Mylar film for testing.

[0028] Furthermore, an X-ray fluorescence spectrometer is used to scan the rice sample powder at different voltages and different filter conditions for different time periods to obtain a plurality of different rice sample powder spectrum data, including:

[0029] The light tube voltage of the X-ray fluorescence spectrometer is set to 5 kV, without a filter, and the rice sample powder is scanned for 30 seconds to obtain the first spectrum data;

[0030] The light tube voltage of the X-ray fluorescence spectrometer is set to 10 kV, the primary filter material is metal aluminum, and the thickness is 40 microns, and the rice sample powder is scanned for 60 seconds to obtain the second spectrum data;

[0031] The light tube voltage of the X-ray fluorescence spectrometer was set to 20 kV, the primary filter material was metal aluminum, and the thickness was 260 microns. The rice sample powder was scanned for 120 seconds to obtain the third spectrum data.

[0032] Specifically, the NDA-300 energy dispersive X-ray fluorescence spectrometer (EDXRF) produced by Beijing Nayou Technology was turned on for preheating for 5 minutes, and the energy calibration sheet provided by the EDXRF instrument manufacturer was used for 3 minutes of energy calibration. Then the sample cup containing the rice sample was placed in the instrument. The target material of the X-ray tube was Rh (rhodium), the incident angle was 70°, the exit angle was 20°, the window material was Be (beryllium), and the thickness was 65μm; the detector material was Si (silicon), the window material was Be, and the thickness was 12.7μm. The rice sample powder was scanned for different times under different voltages and different filter conditions using an X-ray fluorescence spectrometer.

[0033] Further, a plurality of spectral characteristic curves respectively corresponding to the plurality of spectral data are constructed according to the plurality of energy spectral intensity data in each spectral data and the energy characteristic data corresponding to each energy spectral intensity data, including:

[0034] A first spectral characteristic curve is constructed based on multiple energy spectral intensity data in the first spectral data and energy characteristic data corresponding to each energy spectral intensity data; a second spectral characteristic curve is constructed based on multiple energy spectral intensity data in the second spectral data and energy characteristic data corresponding to each energy spectral intensity data; a third spectral characteristic curve is constructed based on multiple energy spectral intensity data in the third spectral data and energy characteristic data corresponding to each energy spectral intensity data.

[0035] Further, the energy characteristic data is an integer in the range of 0 to 2048, and the energy characteristic data corresponding to any two energy spectrum intensity data in the same spectrum data are different;

[0036] According to different energy characteristic data segments, multiple spectral characteristic curves are intercepted respectively to obtain multiple spectral characteristic curve segments, including: according to the energy characteristic data segment [13-251], the first spectral characteristic curve is intercepted to obtain the first spectral characteristic curve segment; according to the energy characteristic data segment [13-521], the second spectral characteristic curve is intercepted to obtain the second spectral characteristic curve segment; according to the energy characteristic data segment [13-757], the third spectral characteristic curve is intercepted to obtain the third spectral characteristic curve segment.

[0037] Specifically, there are 2048 channels corresponding to the XRF spectrum data. The first spectrum characteristic curve takes the [13-251] segment, the second spectrum characteristic curve takes the [13-521] segment, and the third spectrum characteristic curve takes the [13-757] segment, for a total of 1500 data points.

[0038] Furthermore, the obtained multiple spectral characteristic curve segments are spliced ​​and smoothed to obtain a processed spectral characteristic curve, including:

[0039] The first spectral characteristic curve segment, the second spectral characteristic curve segment and the third spectral characteristic curve segment are sequentially arranged and spliced ​​to obtain a spliced ​​spectral characteristic curve, and a polynomial smoothing process with a set window size is performed on the spliced ​​spectral characteristic curve to obtain a processed spectral characteristic curve.

[0040] Specifically, SG smoothing (polynomial smoothing) is performed on the spliced ​​spectral characteristic curve with a window size of 7.

[0041] Furthermore, the rice origin tracing model is trained according to the processed spectral characteristic curve by using the orthogonal partial least squares discriminant analysis method to obtain the trained rice origin tracing model, including:

[0042] The spectral characteristic curves of rice with known origin after processing are subjected to multivariate statistical analysis to obtain the scoring results of the rice samples with known origin; the rice origin tracing model is trained according to the spectral characteristic curves of rice with known origin after processing and the scoring results of the rice samples with known origin to obtain the trained rice origin tracing model.

[0043] Specifically, open the SIMCA software, click "File", create "New Spreadsheet", and import the spectral characteristic curves of rice from known origins into the SIMCA software. Set the horizontal row, vertical row and first row to "Primary", then click "Edit", click "Transpose" to change the data format to horizontal rows for main factors and vertical rows for sample numbers, save the data, click "Home", then click "New as", select "use advancedmodel" "Observations", group the experimental data, select OPLS-DA on "model type", fit the data after confirmation, click "Home", then click "Autofit", and then click "Scores", you can get the OPLS-DA score chart, as shown Figure 2 As shown, from Figure 2 It can be seen that rice samples from four production areas, namely Yilan County, Ning'an City, Southeast Guizhou and Dazhou City, are completely separated; this indicates that the processed spectral characteristic curve combined with the pattern recognition method has a certain potential for distinguishing the production areas.

[0044] Furthermore, the rice origin tracing model is trained by principal component analysis according to the processed spectral characteristic curve to obtain the trained rice origin tracing model, including:

[0045] The processed spectral characteristic curve is subjected to principal component analysis to obtain 11 principal components; the obtained 11 principal components are input into SPSS analysis software to obtain discriminant function coefficients; a discriminant function corresponding to the rice origin is obtained according to the 11 principal components and the discriminant function coefficients; the 11 principal components are respectively the first principal component PC1, the second principal component PC2, the third principal component PC3, the fourth principal component PC4, the fifth principal component PC5, the sixth principal component PC6, the seventh principal component PC7, the eighth principal component PC8, the ninth principal component PC9, the tenth principal component PC10 and the eleventh principal component PC11;

[0046] The discriminant function corresponding to the rice origin includes:

[0047] NAS=-0.302PC1+1.132PC2-0.206PC3-0.013PC4-1.218PC5-0.938PC6+0.417PC7-0.174PC 8-0.259PC9+0.324PC10-0.019PC11-17.109;

[0048] YLX=-0.373PC1+0.570PC2+0.227PC3+0.398PC4-630PC5-0.307PC6-0.162PC7-0.152PC8-0.034PC9+0.247PC10-0.312PC11-8.431;

[0049] QDN=0.157PC1-0.677PC2+0.097PC3+0.437PC4-0.561PC5+0.007PC6-0.267PC7-0.025PC 8+0.112PC9-0.231PC10+0.359PC11-8.227;

[0050] DZS=0.404PC1-0.627PC2-0.181PC3-0.800PC4+0.852PC5+0.895PC6+0.147PC7-0.010PC 8+0.092PC9-0.225PC10-0.033PC11-11.465;

[0051] Among them, NAS is produced in Ning'an City, YLX is produced in Yilan County, QDN is produced in Qiandongnan, and DZS is produced in Dazhou City;

[0052] The rice origin tracing model is trained according to the spectral characteristic curve of rice from known origin after processing and the discriminant function corresponding to the rice origin, so as to obtain the trained rice origin tracing model.

[0053] Specifically, the spectral characteristic curves of rice from known origins were processed using Python for PCA (principal component analysis). The implementation of PCA on Python is mainly based on toolkits such as "pandas", "numpy", "sklearn", and "matplotlib". The first 11 principal components of the preprocessed spectral data after PCA were imported into SPSS, and "Analysis", "Classification", and "Discriminant" were clicked. Then "Category" was selected as the grouping variable, and the definition range (0-4) was set. The remaining indicators were selected as independent variables, and "Input independent variables together" was selected. Click the "Statistics" button on the right, check "Average" and "Fisher" and "Unstandardized" under the function coefficients, click the "Classification" button on the right, check "Summary Table", "Leave One Classification", "Merge Group", and "Domain Map", click the "Save" button on the right, check "Predict Group Members" and "Group Member Probability", and finally click OK to calculate, and the discriminant function coefficients of rice from four origins can be obtained. The discriminant function can be obtained based on this coefficient.

[0054] At the same time, the canonical discriminant function is obtained, which is as follows:

[0055] Function1=0.088PC1-0.205PC2-0.011PC3-0.070PC4+0.228PC5+0.161PC6-0.020PC7+0.034PC9-0.070PC10+0.038PC11

[0056] Function2=-0.038PC1-0.066PC2+0.077PC3+0.219PC4+0.017PC5-0.075PC6-0.112PC7+0.024PC8+0.023PC9-0.011PC10+0.039PC11

[0057] Function1 and Function2 are the first canonical discriminant function and the second canonical discriminant function respectively;

[0058] According to the canonical discriminant function, the score graph of PCA-LDA (linear discriminant function) can be obtained. The discriminant graphs of the four regions after PCA-LDA processing are as follows: Figure 3 As shown, it can be seen that the rice in the four regions can be well distinguished and can be clearly identified as four different origin categories. The triangle represents the rice produced in Ning'an City, Heilongjiang Province, which is mainly concentrated in the lower right corner. The square represents the rice samples in Yilan County, Heilongjiang Province, which are concentrated in the upper left area. The regular pentagon represents the samples in Dazhou City, Sichuan Province, which are concentrated in the lower right area. The rest are rice samples in Qiandongnan, Guizhou Province, which are mainly concentrated in the upper right area. The specific classification results are shown in Table 1. Table 1 shows the classification and identification results of rice origin using PCA-LDA. In general, the accuracy of the established PCA-LDA model for classifying samples of known origin is 98.2%. Only one sample in Ning'an City and Dazhou City produced a classification error. In order to verify the prediction ability of the PCA-LDA method for unknown samples, the leave-one-out cross-validation was selected, and the overall accuracy of the leave-one-out cross-validation was 96.4%, which fully demonstrated the stability and effectiveness of the method.

[0059] Table 1 Classification and identification results of rice origin using PCA-LDA

[0060]

[0061] Furthermore, the rice origin tracing model is trained by a random forest algorithm according to the processed spectral characteristic curve to obtain a trained rice origin tracing model, including:

[0062] A sample set of spectral characteristic curves of rice samples is constructed according to the processed spectral characteristic curves, and 70% of the sample set is used as a training set. The training set is used to train the rice origin traceability model through the random forest algorithm to obtain the trained rice origin traceability model.

[0063] Specifically, the random forest algorithm (RF) was used to identify the origin of rice, with 70% of the sample data used as the training set and 30% as the test set. Learning and recognition were performed using Python and the code packages used in Python were "pandas", "numpy" and "sklearn", and the most important code package used to implement pattern recognition was "sklearn". The OOB (out-of-bag sample) method classification implementation and drawing of the random forest were mainly based on the MetaboAnalyst 6.0 platform. The recognition results of the identification of the origin of rice using the random forest algorithm (RF) are shown in Tables 2 and Figure 4 ,Table 2 shows the classification and identification results of rice origin using RF model, Figure 4 It reflects the relationship between the error rate and the number of decision trees. As can be seen from the figure, RF has the lowest error rate when the number of decision trees is close to 500. In this example, the number of decision trees is set to 500 and the K value (number of feature variables) is 1402. The overall accuracy of RF's OOB is 97.3%, and the overall accuracy of the prediction set is 97%. There is only one sample prediction error in the prediction set.

[0064] Table 2 Classification and identification results of rice origin using RF model

[0065]

[0066] Furthermore, a rice origin traceability model is trained according to the processed spectral characteristic curve using a support vector machine algorithm to obtain a trained rice origin traceability model, including: constructing a sample set of spectral characteristic curves of rice samples according to the processed spectral characteristic curve, and using 70% of the sample set as a training set; and using the training set to train the rice origin traceability model using a support vector machine algorithm to obtain a trained rice origin traceability model.

[0067] Specifically, the support vector machine algorithm is used to identify the origin of rice. The support vector machine uses 70% of the sample data as the training set and the remaining 30% of the sample data as the test set. Through Python and for learning and recognition, the implementation of the algorithm is mainly based on "pandas", "numpy" and "sklearn". The results are shown in Table 3, which shows the classification and identification results of the origin of rice using the SVM model. As can be seen from Table 3, the support vector machine (SVM) algorithm has a 100% accuracy rate for the test set, with no errors; the cross-validation uses the K-Fold (K-fold cross validation) method, and the overall accuracy rate is 98.2%.

[0068] Table 3 Classification results of rice origin using SVM model

[0069]

[0070] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. It should therefore be understood that many modifications may be made to the exemplary embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in a manner different from that described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be used in other described embodiments.

Claims

1. A rice origin tracing method, characterized in that: include: Step 1: Pre-treating rice samples from different known origins to obtain rice sample powder; Step 2: Use an X-ray fluorescence spectrometer to scan the rice sample powder at different voltages and different filter conditions for different time periods to obtain a plurality of rice sample powder spectrum data; each spectrum data includes a plurality of energy spectrum intensity data, and each energy spectrum intensity data corresponds to an energy characteristic data; Step 3: using the energy characteristic data as the horizontal coordinate and the energy spectrum intensity data as the vertical coordinate, constructing a plurality of spectrum characteristic curves corresponding to the plurality of spectrum data respectively according to the plurality of energy spectrum intensity data in each spectrum data and the energy characteristic data corresponding to each energy spectrum intensity data; Step 4: intercepting multiple spectral characteristic curves according to different energy characteristic data segments to obtain multiple spectral characteristic curve segments, and concatenating and smoothing the obtained multiple spectral characteristic curve segments to obtain a processed spectral characteristic curve; Step 5: Establish a rice origin traceability model, and train the rice origin traceability model according to the processed spectral characteristic curve to obtain a trained rice origin traceability model; Step 6: Obtain the processed spectral characteristic curve of the rice from unknown origin, input the processed spectral characteristic curve of the rice from unknown origin into the trained rice origin tracing model, and output the rice origin.

2. A rice origin tracing method according to claim 1, characterized in that: The rice samples with known origins were pretreated, including drying the rice samples at 60° C. for 20 hours and crushing them for 2 minutes after cooling.

3. A rice origin tracing method according to claim 1, characterized in that: The rice sample powder was scanned at different voltages and filter conditions for different time periods using an X-ray fluorescence spectrometer to obtain multiple different rice sample powder spectral data, including: The light tube voltage of the X-ray fluorescence spectrometer is set to 5 kV, without a filter, and the rice sample powder is scanned for 30 seconds to obtain the first spectrum data; The light tube voltage of the X-ray fluorescence spectrometer is set to 10 kV, the primary filter material is metal aluminum, and the thickness is 40 microns, and the rice sample powder is scanned for 60 seconds to obtain the second spectrum data; The light tube voltage of the X-ray fluorescence spectrometer was set to 20 kV, the primary filter material was metal aluminum, and the thickness was 260 microns. The rice sample powder was scanned for 120 seconds to obtain the third spectrum data.

4. A rice origin tracing method according to claim 3, characterized in that: Constructing a plurality of spectral characteristic curves respectively corresponding to the plurality of spectral data according to the plurality of energy spectral intensity data in each spectral data and the energy characteristic data corresponding to each energy spectral intensity data, including: A first spectral characteristic curve is constructed based on multiple energy spectral intensity data in the first spectral data and energy characteristic data corresponding to each energy spectral intensity data; a second spectral characteristic curve is constructed based on multiple energy spectral intensity data in the second spectral data and energy characteristic data corresponding to each energy spectral intensity data; a third spectral characteristic curve is constructed based on multiple energy spectral intensity data in the third spectral data and energy characteristic data corresponding to each energy spectral intensity data.

5. A rice origin tracing method according to claim 4, characterized in that: The energy characteristic data is an integer in the range of 0 to 2048, and the energy characteristic data corresponding to any two energy spectrum intensity data in the same spectrum data are different; According to different energy characteristic data segments, multiple spectral characteristic curves are intercepted respectively to obtain multiple spectral characteristic curve segments, including: according to the energy characteristic data segment [13-251], the first spectral characteristic curve is intercepted to obtain the first spectral characteristic curve segment; according to the energy characteristic data segment [13-521], the second spectral characteristic curve is intercepted to obtain the second spectral characteristic curve segment; according to the energy characteristic data segment [13-757], the third spectral characteristic curve is intercepted to obtain the third spectral characteristic curve segment.

6. A rice origin tracing method according to claim 5, characterized in that: The obtained multiple spectral characteristic curve segments are spliced ​​and smoothed to obtain a processed spectral characteristic curve, including: The first spectral characteristic curve segment, the second spectral characteristic curve segment and the third spectral characteristic curve segment are sequentially arranged and spliced ​​to obtain a spliced ​​spectral characteristic curve, and a polynomial smoothing process with a set window size is performed on the spliced ​​spectral characteristic curve to obtain a processed spectral characteristic curve.

7. A rice origin tracing method according to claim 6, characterized in that: The rice origin tracing model is trained according to the processed spectral characteristic curve to obtain the trained rice origin tracing model, including: The rice origin tracing model is trained according to the processed spectral characteristic curve by using the orthogonal partial least squares discriminant analysis method to obtain the trained rice origin tracing model; The spectral characteristic curves of rice with known origin after processing are subjected to multivariate statistical analysis to obtain the scoring results of the rice samples with known origin; the rice origin tracing model is trained according to the spectral characteristic curves of rice with known origin after processing and the scoring results of the rice samples with known origin to obtain the trained rice origin tracing model.

8. A rice origin tracing method according to claim 6, characterized in that: The rice origin tracing model is trained according to the processed spectral characteristic curve to obtain the trained rice origin tracing model, including: The rice origin tracing model is trained by principal component analysis according to the processed spectral characteristic curve to obtain a trained rice origin tracing model; The processed spectral characteristic curve is subjected to principal component analysis to obtain 11 principal components; the obtained 11 principal components are input into SPSS analysis software to obtain discriminant function coefficients; a discriminant function corresponding to the rice origin is obtained according to the 11 principal components and the discriminant function coefficients; the 11 principal components are respectively the first principal component PC1, the second principal component PC2, the third principal component PC3, the fourth principal component PC4, the fifth principal component PC5, the sixth principal component PC6, the seventh principal component PC7, the eighth principal component PC8, the ninth principal component PC9, the tenth principal component PC10 and the eleventh principal component PC11; The discriminant function corresponding to the rice origin includes: NAS=-0.302PC1+1.132PC2-0.206PC3-0.013PC4-1.218PC5-0.938PC6+0.417PC7-0.174PC 8-0.259PC9+0.324PC10-0.019PC11-17.109; YLX=-0.373PC1+0.570PC2+0.227PC3+0.398PC4-630PC5-0.307PC6-0.162PC7-0.152PC8-0.034PC9+0.247PC10-0.312PC11-8.431; QDN=0.157PC1-0.677PC2+0.097PC3+0.437PC4-0.561PC5+0.007PC6-0.267PC7-0.025PC 8+0.112PC9-0.231PC10+0.359PC11-8.227; DZS=0.404PC1-0.627PC2-0.181PC3-0.800PC4+0.852PC5+0.895PC6+0.147PC7-0.010PC 8+0.092PC9-0.225PC10-0.033PC11-11.465; Among them, NAS is produced in Ning'an City, YLX is produced in Yilan County, QDN is produced in Qiandongnan, and DZS is produced in Dazhou City; The rice origin tracing model is trained according to the spectral characteristic curve of rice from known origin after processing and the discriminant function corresponding to the rice origin, so as to obtain the trained rice origin tracing model.

9. A rice origin tracing method according to claim 6, characterized in that: The rice origin tracing model is trained according to the processed spectral characteristic curve to obtain the trained rice origin tracing model, including: The rice origin tracing model is trained by using a random forest algorithm according to the processed spectral characteristic curve to obtain a trained rice origin tracing model; A sample set of spectral characteristic curves of rice samples is constructed according to the processed spectral characteristic curves, and 70% of the sample set is used as a training set. The training set is used to train the rice origin traceability model through the random forest algorithm to obtain the trained rice origin traceability model.

10. A rice origin tracing method according to claim 6, characterized in that: The rice origin tracing model is trained according to the processed spectral characteristic curve to obtain the trained rice origin tracing model, including: The rice origin tracing model is trained by using a support vector machine algorithm according to the processed spectral characteristic curve to obtain a trained rice origin tracing model; A sample set of spectral characteristic curves of rice samples is constructed according to the processed spectral characteristic curves, and 70% of the sample set is used as a training set. The training set is used to train a rice origin traceability model through a support vector machine algorithm to obtain a trained rice origin traceability model.

Images