A tea oil variety identification method based on multi-source data

Abstract

The present invention discloses a method for identifying tea oil varieties based on multi-source data, which belongs to the technical field of edible oils. The present invention obtains first target data, second target data, and third target data by continuously collecting a first target image, a second target image, and a third target image of the tea oil to be analyzed, calculates fusion similarity based on a standard fingerprint library of tea oil, and determines whether it is a single variety of tea oil. Furthermore, the present invention collects first feature data, second feature data, and third feature data, screens feature Raman shifts and feature wavelengths to establish a pure oil spectrum database, establishes a variety recognition model based on the pure oil spectrum database, improves the accuracy and comprehensiveness of tea oil recognition, establishes a variety ratio model based on the second feature data and the third feature data, and establishes a verification model based on the first feature data to improve recognition speed and recognition accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of edible oils, and in particular to a tea oil variety identification method based on multi-source data. Background Art

[0002] Camellia oil identification method has near infrared spectroscopy, capillary gas chromatography, headspace-gas chromatography-mass spectrometry etc., wherein near infrared spectroscopy has the characteristics of fast and simple, is suitable for industrial large-scale detection, but the precision of identification by near infrared spectroscopy is not high, can only detect the oil variety of part. Publication number is that Chinese patent application CN108169169A discloses a kind of rapid detection method of camellia oil adulteration based on near infrared spectroscopy technology, described method is combined with Mahalanobis distance cluster analysis method by near infrared spectroscopy technology. Combine different band selection, different spectral pretreatment methods to establish the model of five kinds of edible vegetable oil type discrimination, near infrared spectroscopy technology combines self-organizing competitive neural network to carry out pattern recognition to five kinds of edible vegetable oil. Under the binary adulteration system of camellia oil, establish the qualitative discriminant analysis model of three kinds of adulterated edible oils, adopt statistical method to carry out singularity analysis to adulterated sample, combine partial least squares method (PLS) to establish the quantitative analysis model of three kinds of adulterated edible oils. Due to the limited wavelength range of near-infrared light, this method uses near-infrared spectroscopy technology to detect the quality of camellia oil. The types of oils that can be detected are limited, and the model is not comprehensive.

[0003] Publication number CN106950241A Chinese patent application discloses a method for predicting other adulterated oil types and contents in tea oil. The method first collects tea oil, corn oil, sunflower oil, and rapeseed oil of different types and brands, mixes corn oil, sunflower oil, and rapeseed oil in different proportions into tea oil, performs 1HNMR detection, obtains corresponding spectrum, determines the best pre-treatment method by PCA score graph, further adopts orthogonal partial least squares method-discriminant analysis (OPLS-DA) and partial least squares method (PLS), establishes qualitative and quantitative models, and verifies the model, and finally utilizes prediction set to identify the model. The qualitative model and quantitative model established by the above method are only for the identification of some oils, and cannot achieve comprehensive and reliable detection. The prior art has the need for further improvement. Summary of the Invention

[0004] To address the shortcomings of the aforementioned prior art, the present invention proposes a tea oil variety identification method based on multi-source data. This method continuously acquires first, second, and third target images of the tea oil to be analyzed to obtain first, second, and third target data. Fusion similarity is calculated based on a standard tea oil fingerprint library to determine whether the tea oil is a single variety. Furthermore, the present invention establishes a variety identification model, a variety ratio model, and a verification model based on a pure oil spectral database, improving the comprehensiveness and accuracy of tea oil identification.

[0005] The technical solution of the present invention is achieved as follows:

[0006] A tea oil variety identification method based on multi-source data comprises the following steps:

[0007] Step 1: Prepare the tea oil to be analyzed, place it at the starting position of the workbench, and install image acquisition devices on the first image acquisition channel, second image acquisition channel, and third image acquisition channel of the workbench respectively;

[0008] Step 2: Acquire the background image and the reference image. The tea oil to be analyzed enters the first image acquisition channel. The image acquisition device acquires the first target image, enters the second image acquisition channel, acquires the second target image, enters the third image acquisition channel, and acquires the third target image.

[0009] Step 3: Obtain the first target data, the second target data, and the third target data respectively according to the first target image, the second target image, and the third target image, establish a standard fingerprint library of tea oil, and calculate the fusion similarity;

[0010] Step 4: If the fusion similarity exceeds the similarity threshold, the tea oil to be analyzed is a single variety of tea oil, the recognition result is output, and the program ends; otherwise, go to step 5;

[0011] Step 5: Establish a pure oil spectrum database, and establish a variety identification model, a variety ratio model, and a verification model based on the pure oil spectrum database;

[0012] Step 6: Substitute the first target data, the second target data, and the third target data into the variety recognition model to obtain the oil variety set, and obtain the variety ratio set according to the variety ratio model. Verify according to the verification model. If the verification is successful, proceed to step 7; otherwise, output recognition failure and end the program.

[0013] Step 7: Determine the target oil type based on the variety ratio set, and output the target oil type and variety ratio set.

[0014] In the present invention, in step 2, the image acquisition device includes a light generator, a light receiver and a photoelectric sensor. The photoelectric sensor acquires a background image and a reference image. The reference image is an optical signal image of a blank sample. The light generator of the first image acquisition channel generates Raman light, and the light receiver acquires a first target image to generate Raman spectrum data. In the second image acquisition channel, the light generator generates ultraviolet-visible light, and the light receiver acquires a second target image to generate ultraviolet-visible spectrum data. In the third image acquisition channel, the light generator generates infrared light, and the light receiver acquires a third target image to generate infrared spectrum data.

[0015] In the present invention, in step 3, the Raman spectrum data is baseline corrected and intensity normalized to generate the first target data Y1(α), where α is the Raman shift, and the UV-visible absorbance A is calculated based on the UV-visible spectrum data. λ , and then generate the second target data Y2(λ), λ is the wavelength of ultraviolet visible light, and the infrared absorbance A is calculated according to the infrared spectrum data β , and then generate the third target data Y3(β), where β is the wavelength of infrared light. The steps of establishing a standard fingerprint library for tea oil are as follows: collect pure tea oil samples from multiple different origins, obtain three spectral data of the pure tea oil samples and obtain the first standard data, second standard data, and third standard data after preprocessing, and establish a standard fingerprint library for tea oil according to the origin, first standard data, second standard data, and third standard data of each pure tea oil sample.

[0016] In the present invention, in step 3, the first target data, the second target data, and the third target data are respectively compared with the first standard data, the second standard data, and the third standard data of different origins in the tea oil standard fingerprint library, and the fusion similarity is calculated. If the fusion similarity exceeds the similarity threshold, the tea oil to be analyzed is pure tea oil. The fusion similarity S=(S1+S2+S3) / 3, S1 is the maximum similarity between the first target data and the first standard data, S2 is the maximum similarity between the second target data and the second standard data, and S3 is the maximum similarity between the third target data and the third standard data. m different Raman shifts {α1,..., α m}, , d is the origin serial number of the pure tea oil sample, D is the total number of origins of the pure tea oil samples, 1≤d≤D, A d1α is the first standard data of the d-th origin.

[0017] In the present invention, in step 5, the steps of establishing a pure oil spectrum database are as follows: preparing a plurality of pure oil samples, obtaining three types of spectrum data of the pure oil samples and performing preprocessing to obtain first characteristic data, second characteristic data, and third characteristic data, screening characteristic Raman shift and characteristic wavelength, and establishing a pure oil spectrum database U1, U1={u1,u2,...,ui ,...,u n},u i ={C 1i , C 2i , C 3i},u i is the characteristic data set of the i-th pure oil, C 1i is the characteristic Raman shift set of the i-th pure oil, C 2i is the first characteristic wavelength set of the i-th pure oil, C 3i is the second characteristic wavelength set of the i-th pure oil, and n is the number of pure oil types.

[0018] In the present invention, in step 6, an identification database U2 is established based on the pure oil spectrum database, a variety recognition model is established based on the identification database, a first set X1 and a second set X2 are determined based on the variety recognition model, a variety ratio model is established based on the second feature data, the third feature data, the first set X1 and the second set X2, and a verification model is established based on the first feature data, the first set X1 and the second set X2, U2={u 21 ,u 22 ,...,u 2i ,...,u 2n},u 2i ={C 21i , C 22i , C 23i}, ,u 2i is the identification data set of the i-th pure oil, C 21i is the first identification data set of the i-th pure oil, C 22i is the second identification data set of the i-th pure oil, C 23i is the third identification data set of the i-th pure oil.

[0019] In the present invention, in step 6, the variety identification model: if , and and , then the tea oil to be analyzed contains the i-th pure oil, and the i-th pure oil is added to the first set X1. , and and , then the tea oil to be analyzed contains the i-th pure oil, and the i-th pure oil is added to the second set X2, C 10 is the characteristic Raman shift of the first target data, C 20 is the characteristic wavelength of the second target data, C 30 is the characteristic wavelength of the third target data.

[0020] In the present invention, in step 6, the variety ratio model: , δ1 is the target threshold, a 1j is the composition ratio of the jth pure oil in the first set, a 2k is the composition ratio of the kth pure oil in the second set, y 21j (λ) is the second characteristic data of the jth pure oil in the first set, y 22k (λ) is the second characteristic data of the kth pure oil in the second set, y 31j (β) is the third characteristic data of the jth pure oil in the first set, y 32k (β) is the third characteristic data of the kth pure oil in the second set.

[0021] In the present invention, in step 6, the model is verified: ,y 11j is the first characteristic data of the jth pure oil in the first set, y 12k is the first characteristic data of the kth pure oil in the second set, b 1j is the calibration ratio of the jth pure oil in the first set, b 2k is the calibration ratio of the kth pure oil in the second set.

[0022] In the present invention, in step 7, the oil variety set X3 is determined according to the variety identification model, X3=X1+X2, the composition ratio of each pure oil is obtained according to the variety ratio model, and the variety ratio set is generated according to the composition ratio of each pure oil. If any j-th pure oil in the first set satisfies |a 1j -b 1j |<δ2 and any k-th pure oil in the second set satisfies |a 2k -b 2k |<δ2, the verification succeeds, otherwise the verification fails, and δ2 is the verification threshold.

[0023] The implementation of the tea oil variety identification method based on multi-source data of the present invention has the following beneficial effects: the present invention obtains first target data, second target data, and third target data by continuously collecting the first target image, second target image, and third target image of the tea oil to be analyzed, calculates the fusion similarity based on the tea oil standard fingerprint library, and determines whether it is a single variety of tea oil. Furthermore, the present invention collects the first feature data, second feature data, and third feature data of multiple pure oil samples, screens the characteristic Raman shift and characteristic wavelength to establish a pure oil spectrum database, establishes a variety identification model based on the pure oil spectrum database, improves the accuracy and comprehensiveness of tea oil identification, establishes a variety ratio model based on the second target data, third target data, second feature data, and third feature data, and establishes a verification model based on the first target data and the first feature data, thereby improving the recognition speed and recognition accuracy, and is suitable for industrial large-scale detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a flow chart of the tea oil variety identification method based on multi-source data of the present invention;

[0025] Figure 2 A schematic diagram of multiple image acquisition channels of a workbench of the present invention;

[0026] Figure 3 This is a light path diagram of the image acquisition device of the present invention;

[0027] Figure 4 is a schematic diagram of the second target data and the third target data of the present invention;

[0028] Figure 5 Schematic diagram of identifying data sets for the present invention. DETAILED DESCRIPTION

[0029] In order to more clearly understand the purpose, technical solutions and advantages of the present application, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0030] Because detecting only one type of spectral data of tea oil can easily lead to low recognition accuracy and limited types of oils, the present invention identifies oils based on the ultraviolet-visible spectral data and infrared spectral data of the tea oil to be analyzed, and verifies them through Raman spectral data.

[0031] Infrared spectral data is based on the selective absorption of infrared light by molecules in the tea oil being analyzed. Different chemical bonds vibrate at specific frequencies, forming absorption peaks. The wavelength of the absorption peaks allows for identification of the tea oil being analyzed, and the absorbance of the absorption peaks provides a basis for quantitative analysis. Ultraviolet-visible spectral data, similar to infrared spectral data, is based on the absorption of ultraviolet or visible light by conjugated systems or chromophores in the molecules, forming absorption peaks. Tea oil variety identification based on both UV-visible and infrared spectral data expands the wavelength range and provides complementary information.

[0032] Due to the inelastic scattering effect, when Raman light irradiates the tea oil being analyzed, the photons interact with the vibrational or rotational energy levels of the molecules, producing energy changes known as Raman shifts, which form Raman peaks. The main components of tea oil (such as unsaturated fatty acids and glycerides) have specific Raman peaks, while other oils (such as palm oil and soybean oil) have different fatty acid compositions, resulting in variations in the Raman shifts and intensities of these peaks. Therefore, establishing a validation model based on Raman spectral data improves identification accuracy. Example 1

[0033] like Figures 1 to 5 As shown, the tea oil variety identification method based on multi-source data of the present invention includes the following steps.

[0034] Step 1: Prepare the tea oil to be analyzed, place it on the starting position of the workbench, and install image acquisition devices on the first image acquisition channel, the second image acquisition channel, and the third image acquisition channel of the workbench. Figure 2 As shown, the tea oil to be analyzed is placed in a cuvette, which is placed at the starting position of the workbench. The starting position is driven by a conveyor belt and sequentially enters the first image acquisition channel, the second image acquisition channel, and the third image acquisition channel. Figure 3 As shown, the image acquisition device includes a light generator, a light receiver and a photoelectric sensor. The angle between the light generator and the light receiver is 30°. The light generator is a surface light source, including multiple evenly arranged point light sources.

[0035] Step 2: Collect the background image and reference image, the tea oil to be analyzed enters the first image acquisition channel, the image acquisition device acquires the first target image, enters the second image acquisition channel, acquires the second target image, enters the third image acquisition channel, and acquires the third target image. Figure 3 As shown, the photoelectric sensor collects a background image and a reference image via optical fiber. The reference image is an optical signal image of a blank sample. The light generator in the first image acquisition channel generates Raman light, and the light receiver collects the first target image to generate Raman spectral data. The light generator in the second image acquisition channel generates ultraviolet-visible light, and the light receiver collects the second target image to generate ultraviolet-visible spectral data. The light generator in the third image acquisition channel generates infrared light, and the light receiver collects the third target image to generate infrared spectral data. The first, second, and third target images are spectral images of the tea oil to be analyzed, containing multiple spectral data.

[0036] Step 3: Obtain the first target data, second target data, and third target data based on the first target image, second target image, and third target image, respectively, establish a standard fingerprint library for tea oil, and calculate the fusion similarity. After baseline correction and intensity normalization of the Raman spectrum data, the first target data Y1(α) is generated, where α is the Raman shift, and the UV-visible absorbance A is calculated based on the UV-visible spectrum data. λ , and then generate the second target data Y2(λ), λ is the wavelength of ultraviolet visible light, and the infrared absorbance A is calculated according to the infrared spectrum data β , and then generate the third target data Y3(β), where β is the wavelength of infrared light. , , B 1λ is the UV-visible spectrum data at wavelength λ, B 2β is the infrared spectrum data at wavelength β, D1 is the background image, and D2 is the reference image. The establishment of the tea oil standard fingerprint library and the calculation of the fusion similarity are described in detail in Example 2.

[0037] The wavelength range of ultraviolet and visible light is [10nm, 700nm], and the wavelength range of infrared light is [700nm, 2500nm]. Figure 4 As shown, the horizontal axis is wavelength and the vertical axis is absorbance. Wavelengths within [10nm, 700nm] are the second target data, and wavelengths within [700nm, 1200nm] are the third target data. Five absorption peaks appear in the range of 200nm to 1200nm. In the ultraviolet-visible light range, absorption peaks appear at 250nm, 430nm, and 660nm. In the infrared light range, absorption peaks appear at 930nm and 1050nm. Absorbance is a dimensionless quantity that measures the degree of absorption of light when it passes through the tea oil being analyzed.

[0038] Step 4: If the fusion similarity exceeds the similarity threshold, the tea oil to be analyzed is a single variety of tea oil. The identification result is output and the program ends. Otherwise, proceed to Step 5. The range of fusion similarity is [0, 1], and the similarity threshold is generally 0.98. Due to different origins, the three spectral data of pure tea oil have slight fluctuations. The fusion similarity is used to search for the pure tea oil sample in the tea oil standard fingerprint library that is closest to the tea oil to be analyzed based on the three different spectral data.

[0039] Step 5: Establish a pure oil spectrum database, and establish a variety identification model, a variety ratio model, and a verification model based on the pure oil spectrum database. Steps for establishing a pure oil spectrum database: Prepare multiple pure oil samples, obtain three types of spectral data of the pure oil samples and obtain the first characteristic data, second characteristic data, and third characteristic data after preprocessing, screen the characteristic Raman shift and characteristic wavelength, and establish a pure oil spectrum database U1, U1={u1,u2,...,u i ,...,u n},u i ={C 1i , C 2i , C 3i},u i is the characteristic data set of the i-th pure oil, C 1i is the characteristic Raman shift set of the i-th pure oil, C 2i is the first characteristic wavelength set of the i-th pure oil, C 3i is the second characteristic wavelength set of the i-th pure oil, and n is the number of pure oil types. The preprocessing can include moving average smoothing, first-order derivative, standard normal variate transformation, second-order derivative smoothing, multivariate scatter correction, and continuous wavelet transform. This preprocessing eliminates noise from the three spectral data of the pure oil sample and improves data accuracy. The characteristic Raman shift is the Raman shift corresponding to the Raman peak, and the characteristic wavelength is the wavelength corresponding to the absorption peak.

[0040] Step 6: Substitute the first target data, the second target data, and the third target data into the variety recognition model to obtain the oil variety set, obtain the variety ratio set based on the variety ratio model, and perform verification based on the verification model. If the verification is successful, proceed to step 7, otherwise the output is that the recognition fails and the program ends. Establish an identification database U2 based on the pure oil spectrum database, establish a variety recognition model based on the identification database, determine the first set X1 and the second set X2 based on the variety recognition model, establish a variety ratio model based on the second feature data, the third feature data, the first set X1 and the second set X2, and establish a verification model based on the first feature data, the first set X1 and the second set X2. U2={u 21 ,u 22 ,...,u 2i ,...,u 2n},u 2i ={C 21i , C 22i , C 23i}, ,u 2i is the identification data set of the i-th pure oil, C 21i is the first identification data set of the i-th pure oil, C 22i is the second identification data set of the i-th pure oil, C 23i is the third identification data set of the i-th pure oil.

[0041] Variety identification model: If , and and , then the tea oil to be analyzed contains the i-th pure oil, and the i-th pure oil is added to the first set X1. , and and , then the tea oil to be analyzed contains the i-th pure oil, and the i-th pure oil is added to the second set X2, C 10 is the characteristic Raman shift of the first target data, C 20 is the characteristic wavelength of the second target data, C 30 is the characteristic wavelength of the third target data.

[0042] When the number of pure oil types n=3, the identification data set is as follows Figure 5 As shown, the circle represents the feature data set, the shaded part is the intersection of the feature data set, and all blank parts are the identification data set. Each pure oil corresponds to an identification data set, and the identification data sets of all pure oils are combined to generate an identification database.

[0043] Variety scale models: , δ1 is the target threshold, a 1jis the composition ratio of the jth pure oil in the first set, a 2k is the composition ratio of the kth pure oil in the second set, y 21j (λ) is the second characteristic data of the jth pure oil in the first set, y 22k (λ) is the second characteristic data of the kth pure oil in the second set, y 31j (β) is the third characteristic data of the jth pure oil in the first set, y 32k (β) is the third characteristic data of the kth pure oil in the second set.

[0044] Verify the model: ,y 11j is the first characteristic data of the jth pure oil in the first set, y 12k is the first characteristic data of the kth pure oil in the second set, b 1j is the calibration ratio of the jth pure oil in the first set, b 2k is the calibration ratio of the kth pure oil in the second set.

[0045] Step 7: Determine the target oil type based on the variety ratio set, and output the target oil type and variety ratio set. Since the characteristic data sets of each pure oil have an intersection, in the process of solving the variety ratio set and verification, select multiple different α in the characteristic Raman shift set, select multiple different λ in the first characteristic wavelength set, and select multiple different β in the second characteristic wavelength set, and substitute them into the variety ratio model and verification model respectively. Determine the oil variety set X3 according to the variety identification model, X3=X1+X2, obtain the composition ratio of each pure oil according to the variety ratio model, and generate the variety ratio set according to the composition ratio of each pure oil. If any j-th pure oil in the first set satisfies |a 1j -b 1j |<δ2 and any k-th pure oil in the second set satisfies |a 2k -b 2k If |<δ2, the verification succeeds; otherwise, the verification fails. δ2 is the verification threshold, which is set based on the actual recognition accuracy and can be set to 0.1. This verification model improves the accuracy of tea oil recognition. The pure oil corresponding to the non-zero composition ratio in the variety ratio set is the target oil. By outputting the target oil and variety ratio set, the identification of the tea oil to be analyzed is completed. Example 2

[0046] This embodiment further discloses a method for establishing a standard fingerprint library of tea oil and calculating fusion similarity.

[0047] The steps for establishing a standard fingerprint library of tea oil are as follows: collecting pure tea oil samples from multiple different origins, obtaining three spectral data of the pure tea oil samples and preprocessing them to obtain the first standard data, the second standard data, and the third standard data, and establishing a standard fingerprint library of tea oil according to the origin, the first standard data, the second standard data, and the third standard data of each pure tea oil sample.

[0048] The first target data, the second target data, and the third target data are respectively compared with the first standard data, the second standard data, and the third standard data of different origins in the tea oil standard fingerprint library, and the fusion similarity is calculated. If the fusion similarity exceeds the similarity threshold, the tea oil to be analyzed is pure tea oil. The fusion similarity S=(S1+S2+S3) / 3, S1 is the maximum similarity between the first target data and the first standard data, S2 is the maximum similarity between the second target data and the second standard data, and S3 is the maximum similarity between the third target data and the third standard data. m different Raman shifts {α1,..., α m}, , d is the origin serial number of the pure tea oil sample, D is the total number of origins of the pure tea oil samples, 1≤d≤D, A d1α is the first standard data of the d-th origin. S2 and S3 are calculated according to the calculation method of S1, and then the fusion similarity S is obtained.

[0049] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A tea oil variety identification method based on multi-source data, characterized in that: The following steps are involved: Step 1: Prepare the tea oil to be analyzed, place it at the starting position of the workbench, and install image acquisition devices on the first image acquisition channel, the second image acquisition channel, and the third image acquisition channel of the workbench respectively; Step 2: Acquire the background image and the reference image. The tea oil to be analyzed enters the first image acquisition channel. The image acquisition device acquires the first target image. Then, the image acquisition device enters the second image acquisition channel to acquire the second target image. Finally, the image acquisition device enters the third image acquisition channel to acquire the third target image. Step 3: Obtain the first target data, the second target data, and the third target data respectively according to the first target image, the second target image, and the third target image, establish a standard fingerprint library of tea oil, and calculate the fusion similarity; Step 4: If the fusion similarity exceeds the similarity threshold, the tea oil to be analyzed is a single variety of tea oil, the recognition result is output, and the program ends; otherwise, go to step 5; Step 5: Establish a pure oil spectrum database, and establish a variety identification model, a variety ratio model, and a verification model based on the pure oil spectrum database; Step 6: Substitute the first target data, the second target data, and the third target data into the variety recognition model to obtain the oil variety set, and obtain the variety ratio set according to the variety ratio model. Verify according to the verification model. If the verification is successful, proceed to step 7; otherwise, output recognition failure and end the program. Step 7: Determine the target oil type based on the variety ratio set, and output the target oil type and variety ratio set. Wherein, in step 2, the image acquisition device includes a light generator, a light receiver and a photoelectric sensor, the photoelectric sensor acquires a background image and a reference image, the reference image is a light signal image of a blank sample, the light generator of the first image acquisition channel generates Raman light, the light receiver acquires a first target image, and generates Raman spectrum data, in the second image acquisition channel, the light generator generates ultraviolet-visible light, the light receiver acquires a second target image, and generates ultraviolet-visible spectrum data, in the third image acquisition channel, the light generator generates infrared light, the light receiver acquires a third target image, and generates infrared spectrum data, In step 3, the Raman spectrum data is baseline corrected and intensity normalized to generate the first target data Y1(α), where α is the Raman shift, and the UV-visible absorbance A is calculated based on the UV-visible spectrum data. λ , and then generate the second target data Y2(λ), λ is the wavelength of ultraviolet visible light, and the infrared absorbance A is calculated according to the infrared spectrum data β , and then generate the third target data Y3(β), where β is the wavelength of infrared light. The steps of establishing the tea oil standard fingerprint library are as follows: collecting multiple pure tea oil samples from different origins, obtaining three spectral data of the pure tea oil samples and preprocessing them to obtain the first standard data, the second standard data, and the third standard data; establishing the tea oil standard fingerprint library according to the origin, the first standard data, the second standard data, and the third standard data of each pure tea oil sample, In step 3, the first target data, the second target data, and the third target data are respectively compared with the first standard data, the second standard data, and the third standard data of different origins in the tea oil standard fingerprint library to calculate the fusion similarity. If the fusion similarity exceeds the similarity threshold, the tea oil to be analyzed is pure tea oil. The fusion similarity S=(S1+S2+S3) / 3, S1 is the maximum similarity between the first target data and the first standard data, S2 is the maximum similarity between the second target data and the second standard data, and S3 is the maximum similarity between the third target data and the third standard data. m different Raman shifts {α1,..., α m }, , d is the origin serial number of the pure tea oil sample, D is the total number of origins of the pure tea oil samples, 1≤d≤D, A d1α is the first standard data of the d-th origin, calculate S2 and S3 according to the calculation method of S1, and then calculate the fusion similarity S. In step 5, the steps of establishing a pure oil spectrum database are as follows: preparing multiple pure oil samples, obtaining three types of spectrum data of the pure oil samples and preprocessing them to obtain first characteristic data, second characteristic data, and third characteristic data, screening characteristic Raman shifts and characteristic wavelengths, and establishing a pure oil spectrum database U1, where U1={u1,u2,...,u i ,...,u n },u i ={C 1i , C 2i , C 3i },u i is the characteristic data set of the i-th pure oil, C 1i is the characteristic Raman shift set of the i-th pure oil, C 2i is the first characteristic wavelength set of the i-th pure oil, C 3i is the second characteristic wavelength set of the i-th pure oil, n is the number of pure oil types, In step 5, an identification database U2 is established based on the pure oil spectrum database, a variety identification model is established based on the identification database, the first set X1 and the second set X2 are determined based on the variety identification model, a variety ratio model is established based on the second feature data, the third feature data, the first set X1 and the second set X2, and a verification model is established based on the first feature data, the first set X1 and the second set X2, U2={u 21 ,u 22 ,...,u 2i ,...,u 2n },u 2i ={C 21i , C 22i , C 23i }, ,u 2i is the identification data set of the i-th pure oil, C 21i is the first identification data set of the i-th pure oil, C 22i is the second identification data set of the i-th pure oil, C 23i is the third identification data set of the i-th pure oil, In step 6, the variety recognition model: If , and and , then the tea oil to be analyzed contains the i-th pure oil, and the i-th pure oil is added to the first set X1. , and and , then the tea oil to be analyzed contains the i-th pure oil, and the i-th pure oil is added to the second set X2, C 10 is the characteristic Raman shift of the first target data, C 20 is the characteristic wavelength of the second target data, C 30 is the characteristic wavelength of the third target data, In step 6, the breed proportion model: , δ1 is the target threshold, a 1j is the composition ratio of the jth pure oil in the first set, a 2k is the composition ratio of the kth pure oil in the second set, y 21j (λ) is the second characteristic data of the jth pure oil in the first set, y 22k (λ) is the second characteristic data of the kth pure oil in the second set, y 31j (β) is the third characteristic data of the jth pure oil in the first set, y 32k (β) is the third characteristic data of the kth pure oil in the second set, In step 6, verify the model: ,y 11j is the first characteristic data of the jth pure oil in the first set, y 12k is the first characteristic data of the kth pure oil in the second set, b 1j is the calibration ratio of the jth pure oil in the first set, b 2k is the calibration ratio of the kth pure oil in the second set, In step 7, the oil variety set X3 is determined according to the variety identification model, X3=X1+X2, the composition ratio of each pure oil is obtained according to the variety ratio model, and the variety ratio set is generated according to the composition ratio of each pure oil. If any j-th pure oil in the first set satisfies |a 1j -b 1j |<δ2 and any k-th pure oil in the second set satisfies |a 2k -b 2k |<δ2, the verification succeeds, otherwise the verification fails, and δ2 is the verification threshold.

Images