Rapid determination method for storage time of edible vegetable oil based on multispectral information fusion
By using a multispectral information fusion model for detecting the oxidation state of edible vegetable oils, and combining three-dimensional fluorescence spectroscopy and low-field nuclear magnetic resonance spectroscopy data with a support vector machine regression algorithm, the problems of speed and accuracy in detecting the oxidation state of edible vegetable oils have been solved, enabling rapid detection of the storage time of edible vegetable oils.
Patent Information
- Application Number
- CN202110456163.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-04-26
- Publication Date
- 2025-11-25
- Estimated Expiration
- 2041-04-26
AI Technical Summary
Existing technologies are difficult to detect the oxidation state of edible vegetable oils quickly and accurately, and have drawbacks such as long detection cycles, complex interfering components, and poor detection accuracy.
By constructing a rapid detection model for the oxidation state of edible vegetable oils based on multispectral information fusion, and utilizing three-dimensional fluorescence spectroscopy and low-field nuclear magnetic resonance spectroscopy data, combined with support vector machine regression algorithm, a rapid detection method for the storage time of edible vegetable oils is established, noise data is removed, and feature extraction and information fusion are performed.
It enables rapid and accurate detection of the oxidation state of edible vegetable oils, improves the objectivity and robustness of the detection, and meets the complex detection needs of the oxidation state of edible vegetable oils.
Smart Images
Figure CN115248198B_ABST
Abstract
Description
[Technical Field]
[0001] This invention relates to rapid detection technology for edible oils, and particularly to a method for rapidly determining the storage time of edible vegetable oils based on multispectral information fusion. [Background Technology]
[0002] Acrylamide, polycyclic aromatic hydrocarbons, and other food hazards formed during the excessive oxidation of edible vegetable oils during storage are carcinogenic and toxic. Rapid detection and intelligent assessment of the oxidation state of oils have long been a challenge for food production enterprises and regulatory authorities.
[0003] Traditional methods for detecting the oxidation state of edible vegetable oils typically use peroxide value, acid value, and total polar components—products of oil oxidation metabolism—as physicochemical indicators, employing acid-base titration for qualitative and quantitative analysis. These existing methods suffer from drawbacks such as susceptibility to color changes in edible vegetable oils, complex operation, and difficulty in determining the titration endpoint. Potentiometric titration, as a method that effectively avoids color interference in edible vegetable oil samples, offers a high degree of automation, and provides accurate results, has gradually gained acceptance from quality inspection departments and food companies. However, due to the complex composition of edible vegetable oils, changes in polar components such as small-molecule acids, esters, and aldehydes formed during oxidation metabolism can interfere with the results of potentiometric titration. Therefore, these existing detection methods all suffer from drawbacks such as complex interfering components and long detection cycles, becoming significant factors restricting the rapid detection of edible vegetable oils.
[0004] Edible vegetable oils are rich in endogenous antioxidants such as vitamin E, chlorophyll, and carotenoids. These, along with olefins such as conjugated dienes and conjugated trienes in the secondary metabolites produced by oxidation, all exhibit characteristic fluorescence spectra. Studies have found that by analyzing fluorescence spectra or examining changes in low-field NMR relaxation time and peak area caused by oil degradation, spectral information related to the oxidation state of edible vegetable oils can be directly or indirectly resolved, representing a potential rapid identification technique for oil oxidation states. However, the information obtained from fluorescence spectroscopy or low-field NMR analysis of edible vegetable oils is large and complex, exhibiting drawbacks such as poor detection accuracy, narrow response range, and poor robustness. Currently, a rapid detection method for the oxidation state of edible vegetable oils remains unsolved. [Summary of the Invention]
[0005] The purpose of this invention is to overcome the shortcomings of existing technologies by constructing a rapid detection model for the oxidation state of edible vegetable oils based on multispectral information fusion using chemometric algorithms. This model enables rapid detection of the oxidation state of edible vegetable oils, improving the objectivity, robustness, and speed of the detection process.
[0006] To achieve the above objectives, the present invention provides a rapid method for determining the storage time of edible vegetable oils based on multispectral information fusion, the method comprising the following steps:
[0007] (1) Preparation of edible vegetable oil standards
[0008] Take edible oil and prepare n standard samples;
[0009] (2) Acquisition of three-dimensional fluorescence spectroscopy data of standard samples
[0010] Three-dimensional fluorescence spectral data were collected from the n standard samples in step (1) to obtain n sets of three-dimensional fluorescence spectral data. Each set of three-dimensional fluorescence spectral data consisted of 18 sets of two-dimensional fluorescence spectral data.
[0011] The acquisition parameters for the three-dimensional fluorescence spectral data are: the difference between excitation wavelength and emission wavelength Δλ = 10-180 nm, the excitation wavelength is 400-1100 nm, the excitation slit is 5 nm, and the emission slit is 5 nm.
[0012] (3) Extracting characteristic variables of the three-dimensional fluorescence spectrum of the sample
[0013] The parallel factor algorithm was used to extract the feature variables of each group of two-dimensional fluorescence spectra in step (2). The excitation wavelength with the highest fluorescence signal intensity was selected as the feature variable according to the loading value of the parallel factor algorithm. 351 fluorescence spectral feature data were obtained for each group of three-dimensional fluorescence spectral data.
[0014] (4) Obtain the characteristic peak variables of the two-dimensional fluorescence spectrum of the standard.
[0015] Extract the highest intensity peak and surrounding spectral data of the excitation spectrum of each group of two-dimensional fluorescence spectral data in step (2). Each group of two-dimensional fluorescence spectral data yields 11 spectral data points. Therefore, each group of three-dimensional fluorescence spectral data yields 198 characteristic variable data points.
[0016] (5) Acquisition of low-field nuclear magnetic resonance spectroscopy data for standard samples
[0017] Low-field nuclear magnetic resonance spectral data were collected for each standard sample in step (1), and the low-field nuclear magnetic resonance spectral data of each standard sample contained 500 variables.
[0018] (6) Fusion of standard product characteristic data
[0019] (6.1) Obtaining fluorescence features - multi-feature peak fusion data
[0020] For each standard sample, the fluorescence spectral characteristic data obtained in step (3) and the characteristic variable data of the three-dimensional fluorescence spectral data obtained in step (4) are fused to obtain 549 fluorescence characteristic-multi-characteristic peak fusion data;
[0021] (6.2) Acquiring low-field NMR-fluorescence feature fusion data
[0022] For each standard sample, the fluorescence spectral characteristic data obtained in step (3) is fused with the low-field nuclear magnetic resonance spectral data obtained in step (5) to obtain 851 low-field nuclear magnetic resonance-fluorescence characteristic fusion data;
[0023] (7) Determination of physicochemical properties of standard products
[0024] The peroxide value physicochemical index of n standard samples in step (1) was determined respectively to obtain the peroxide value of each standard sample;
[0025] (8) Establish a rapid detection model for edible oil storage time
[0026] Using the low-field nuclear magnetic resonance spectral data of the standard in step (5), the fluorescence characteristic-multi-peak fusion data in step (6.1) and the low-field nuclear magnetic-fluorescence characteristic fusion data in step (6.2) as input values, and the physicochemical index in step (7) as output values, a rapid detection model for the storage time of edible vegetable oil based on multispectral information fusion is established for the standard using the support vector machine regression algorithm.
[0027] (9) Rapid determination of the sample to be tested
[0028] Take the edible oil sample to be tested and obtain its low-field nuclear magnetic resonance spectral data, 549 fluorescence feature-multi-characteristic peak fusion data and 851 low-field nuclear magnetic resonance-fluorescence feature fusion data using the same method as steps (2)-(6). Use the obtained low-field nuclear magnetic resonance spectral data, 549 fluorescence feature-multi-characteristic peak fusion data and 851 low-field nuclear magnetic resonance-fluorescence feature fusion data as input values and substitute them into the rapid detection model in step (8) to obtain the peroxide value physicochemical index of the edible oil sample to be tested. Calculate the storage time of the edible oil sample to be tested based on the obtained peroxide value physicochemical index.
[0029] In this invention, the acquisition of three-dimensional fluorescence spectral data and low-field nuclear magnetic resonance spectral data is carried out using methods existing in the art. The three-dimensional fluorescence spectral data is a three-dimensional matrix data composed of the difference between excitation wavelength and emission wavelength (Δλ), excitation wavelength (Ex), and fluorescence intensity (FI). Three-dimensional fluorescence spectral data can be acquired for edible vegetable oil standards stored for different times using a fluorescence spectrometer.
[0030] The low-field nuclear magnetic resonance (NMR) data is constructed with relaxation time as the x-axis and signal intensity as the y-axis. Two-dimensional low-field NMR data can be acquired from edible vegetable oil standards stored for different times using a low-field NMR spectrometer.
[0031] In this invention, the selection of the penalty factor c, kernel function g, and parameter p in the support vector machine regression algorithm is crucial to the validity and accuracy of the final result. In this invention, the penalty factor c is set to 1.6303, the kernel function g is set to 83.5389, and the parameter p is set to 0.1.
[0032] According to a preferred embodiment, step (8) further includes optimizing the support vector machine regression algorithm using a genetic algorithm (GA). Those skilled in the art can optimize the support vector machine regression algorithm based on existing technology. For example, a solution vector of the data to be optimized can be encoded using a genetic algorithm to form a string, ultimately creating a string group. The fitness of each data point is solved using a fitness function, and the optimal data is selected to form new offspring data. After several genetic mutations, the optimal data is obtained, which is then used for the next step of supporting vector machine model construction.
[0033] This invention, on the one hand, removes interfering noise data such as stray peaks and Rayleigh dispersion from the original three-dimensional fluorescence spectral data using algorithms such as parallel factor. On the other hand, it extracts features from the fluorescence spectrum and low-field NMR characteristic peak data, and then fuses the feature data through a data layer, a feature layer, and a decision layer. The rapid detection model for the storage time of edible vegetable oils established using the support vector machine algorithm enables rapid sample detection, meeting the complex detection requirements of the oxidation state of edible vegetable oils, which is unique to temperature, concentration, and matrix effects. [Attached Image Description]
[0034] Figure 1 Three-dimensional fluorescence spectra of rapeseed oil in different oxidation states;
[0035] Among them, (a) three-dimensional fluorescence spectrum of fresh rapeseed oil, (b) three-dimensional fluorescence spectrum of semi-fresh rapeseed oil, (c) three-dimensional fluorescence spectrum of semi-rotten rapeseed oil, and (d) three-dimensional fluorescence spectrum of rotten rapeseed oil.
[0036] Figure 2 Low-field nuclear magnetic resonance spectra of rapeseed oil under different oxidation states;
[0037] Among them, Figure (a) shows the T2 distribution spectrum of rapeseed oil during different storage periods, and (b) shows the T2 distribution inversion spectrum of rapeseed oil obtained using T-Invfit software;
[0038] Figure 3 For the extraction of three-dimensional fluorescence spectral features of rapeseed oil based on the parallel factor algorithm;
[0039] Figure (a) shows the results of principal components 2-12 under the parallel factor algorithm, and Figure (b) shows the loading score obtained by the parallel factor algorithm.
[0040] Figure 4 Extraction diagram of characteristic peaks in the three-dimensional fluorescence spectrum of rapeseed oil
[0041] Figure 5 This is a schematic diagram illustrating the fusion of three-dimensional fluorescence spectroscopy and low-field nuclear magnetic resonance multispectral data of rapeseed oil.
Detailed Implementation Methods
[0042] The following examples are used to explain the technical solutions of the present invention in a non-limiting manner.
[0043] Example 1
[0044] I. Preparation of Standards
[0045] In this embodiment, rapeseed oil is used as an example to prepare standard products. Rapeseed oil is stored at room temperature for different durations to obtain standard products of fresh, semi-fresh, semi-rancid, and rancid rapeseed oil.
[0046] like Figure 1 As shown, fresh rapeseed oil had the highest content of vitamin E (an antioxidant), the least degradation of conjugated olefin fatty acids, and the best oil quality. Its characteristic peak was observed at 490–600 nm (a). With prolonged storage, the content of antioxidants in rapeseed oil gradually decreased, while the content of conjugated olefin fatty acid oxidative metabolites increased. When the physicochemical indicators such as acid value and peroxide value exceeded national standards, indicating the rancidity stage (d), the fluorescence characteristics showed a decreasing trend in the characteristic peak of vitamin E at 460–600 nm. When the vitamin E content of rapeseed oil samples decreased, the content of conjugated olefins increased, and the samples were between the fresh and rancidity stages, they were artificially distinguished into the near-fresh stage (b) and the near-rancidity stage (c) based on different fluorescence spectral characteristics. In this embodiment, the fresh standard has a room temperature storage period of less than 1 day, the near-fresh standard has a room temperature storage period of 80 days, the near-rotten standard has a room temperature storage period of 480 days, and the rotten standard has a room temperature storage period of 540 days.
[0047] II. Acquisition of Fluorescence Spectrum Data for Standards
[0048] Data were collected from each standard sample using a fluorescence spectrometer. In this invention, the relevant parameters for three-dimensional synchronous fluorescence spectroscopy were set as follows: the difference between excitation and emission wavelengths Δλ = 10–180 nm, the excitation slit width was 5 nm, the excitation wavelength was 400–1100 nm, the excitation slit width was 5 nm, the emission slit width was 5 nm, the scanning speed was 1200 nm / min, and the scanning wavelength interval was 1 nm.
[0049] When adding samples, use a pipette to take 3.0 mL of each standard, filling each standard to 2 / 3 full in the cuvette. When handling the cuvette, try to avoid having oils on your hands that could affect the experimental results.
[0050] Fluorescence spectral data were scanned for each standard, resulting in 7 sets of raw fluorescence spectral data for the standards, such as... Figure 1 As shown.
[0051] III. Extracting Characteristic Variables from Three-Dimensional Fluorescence Spectroscopy
[0052] The raw fluorescence spectrum data for each standard sample includes three-dimensional fluorescence spectra acquired at 18 Δλ excitation wavelengths. Each λ excitation wavelength yields 351 variable data points, resulting in an 18×351 matrix for each standard sample. Feature variables were extracted using the Parallel Factor Analysis (PARAFAC) algorithm, with the loading value used to select the excitation wavelength with the highest fluorescence signal intensity as the feature variable. Therefore, 351 feature data points were extracted for each set of fluorescence spectra.
[0053] IV. Obtaining Characteristic Peak Variables
[0054] The three-dimensional fluorescence spectral data obtained for each standard sample consisted of 18 two-dimensional fluorescence spectral data with a variable of 351, centered on the maximum peak of each two-dimensional fluorescence spectrum (e.g., Figure 4 As shown), five spectral values are taken to the left and right to form characteristic peak data consisting of 11 spectral data, that is, 18×11 characteristic peak spectral data are obtained from the three-dimensional fluorescence spectrum data of each standard.
[0055] V. Acquiring low-field nuclear magnetic resonance spectral data of standards
[0056] To calibrate the low-field NMR instrument: Use a pipette to transfer 2 mL of low-field NMR calibration material into a 10 mL transparent detection bottle. Seal the bottle opening with polytetrafluoroethylene (PTFE). Place the calibration bottle into a 15 mm diameter NMR detection tube. Then, set the signal acquisition parameters through the low-field NMR instrument's operating interface, select the Q-Flame IonizationDetector (FID) queue name, and set the Regulate Analog Gain 1 (RG1) and Regulate Digital Gain 1 (DRG1) parameters to 10 and 3 respectively. Search for the correct center frequency parameter.
[0057] The Time Wait (TW) parameter was set to 100ms. Then, the 90-degree pulse width (P1) and 180-degree pulse width (P2) of low-field nuclear magnetic resonance were found separately using software. The relaxation time values corresponding to the 90-degree pulse width (P1) and 180-degree pulse width (P2) were recorded. The rapeseed oil samples were then sampled using the optimized 90-degree pulse width (P1) and 180-degree pulse width (P2) parameters. The low-field nuclear magnetic resonance spectral data of each sample contained 500 variables.
[0058] By integrating low-field NMR spectral information of edible vegetable oils and setting parameters using Carr-Purcell-Meiboom-Gill (CPMG), the changes in low-field NMR spectra of edible vegetable oils during different storage periods were analyzed. The final data acquisition was performed according to the parameters in Table 1, setting instrument data such as sampling time, number of echoes, and signal amplification. The detection results were optimized by increasing the signal-to-noise ratio and reducing the signal acquisition time, resulting in the following... Figure 2 The low-field nuclear magnetic resonance spectra of rapeseed oil under different oxidation states are shown.
[0059] Table 1. Low-field NMR detection parameter settings
[0060]
[0061] VI. Feature Data Fusion
[0062] Low-field NMR data containing 500 variables and fluorescence spectral feature extraction data containing 351 variables were fused together in a head-to-tail manner to form low-field NMR-fluorescence feature extraction fusion data containing 851 variables.
[0063] The spectral information of 198 characteristic peak variable data and 351 characteristic variable data were fused to obtain 549 multi-characteristic variable fused data.
[0064] VII. Determination of Physicochemical Indicators of Standards
[0065] (1) Method for determining the physicochemical properties of peroxide value: Under relatively dark conditions, take 2.000–3.000 g (accurate to three decimal places) of sample into an iodine flask, add 30.00 mL of a chloroform-glacial acetic acid mixture, and shake thoroughly. Then add 1.00 mL of saturated potassium iodide solution, shake thoroughly, and place in a completely opaque place for three minutes. After three minutes, remove and add 100.00 mL of deionized water, mix well. Finally, titrate with 0.01 mol / L sodium thiosulfate, shaking continuously until the color in the iodine flask turns pale yellow.
[0066] (2) Titration endpoint determination and result analysis: Edible vegetable oils under different oxidation states were subjected to acid-base titration according to the peroxide value titration method. When the solution turned pale yellow, photographs were taken for identification and human sensory evaluation to eliminate the influence of subjective factors of the testers on the test results. Then, two drops of starch indicator solution were quickly added and continuously shaken to mix evenly. Titration was continued with sodium thiosulfate until the solution turned blue-purple, which was the titration endpoint. The test results were then recorded as physicochemical indicators.
[0067] 8. Establish a rapid detection model for the storage time of edible vegetable oils.
[0068] Using low-field nuclear magnetic resonance data (500 variables), low-field nuclear magnetic resonance-fluorescence feature extraction data (851 variables), and fluorescence feature extraction data-multi-peak fluorescence spectral data (549 variables) as input values and physicochemical indicators as output values, a rapid identification model for the storage time of edible vegetable oil based on multispectral information fusion was established using a support vector machine regression algorithm.
[0069] In this invention, a solution vector of the data to be optimized is encoded into a string using a genetic algorithm, ultimately forming a string group. The fitness function is used to calculate the fitness of each data point, selecting the optimal data to form new offspring data. After several genetic mutations, the optimal data is obtained and used for the next step of supporting vector machine model construction.
[0070] IX. Rapid Determination of Edible Vegetable Oil Samples
[0071] Take the edible oil sample to be tested and obtain its low-field nuclear magnetic resonance spectral data, 549 fluorescence characteristic-multi-characteristic peak fusion data, and 851 low-field nuclear magnetic resonance-fluorescence characteristic fusion data using the same method as the standard. Use the obtained low-field nuclear magnetic resonance spectral data, 549 fluorescence characteristic-multi-characteristic peak fusion data, and 851 low-field nuclear magnetic resonance-fluorescence characteristic fusion data as input values and substitute them into the obtained rapid detection model to obtain the peroxide value physicochemical index of the edible oil sample to be tested. The storage time of the edible oil sample to be tested is calculated based on the obtained peroxide value physicochemical index.
[0072] 10. Results Evaluation
[0073] The models were evaluated using raw low-field NMR data (M1), fluorescence spectral feature extraction data (M2), low-field NMR-fluorescence feature extraction spectral fusion data (M3), and the multispectral information fusion data of this invention (M4) as inputs, with physicochemical indicators as outputs. The training and test sets of the four models were compared. The low-field NMR-fluorescence feature extraction spectral fusion data (M3) is a new fused spectral data obtained by concatenating the M1 and M2 data.
[0074] The squared correlation coefficients of the training and test sets in the M1 model results are 0.9183 and 0.9346, respectively; for M2, they are 0.9717 and 0.9718, respectively; for M3, they are 0.9840 and 0.9832, respectively; and for the M4 model used in this invention, the squared correlation coefficients of the training and test sets are 0.9933 and 0.9949, respectively.
[0075] Therefore, the training and test set results of the model prediction results of this invention show that the information fusion algorithm can significantly improve the robustness and relevance of the model compared with using only the original low-field NMR data or fluorescence spectral feature extraction data, or the low-field NMR-fluorescence spectral fusion data.
[0076] Table 2. Mathematical modeling results for rapid identification of storage time of edible vegetable oils based on support vector machine algorithm.
[0077]
[0078] It is evident that the rapid detection method for the oxidation state of edible vegetable oil based on multispectral information fusion of the present invention effectively achieves rapid detection of the oxidation state of edible vegetable oil, and improves the objectivity, robustness and speed of the detection of the oxidation state of edible vegetable oil.
Claims
1. A rapid method for determining the storage time of edible vegetable oils based on multispectral information fusion, the method comprising the following steps: (1) Preparation of edible vegetable oil standards Edible vegetable oils were subjected to room temperature oxidation treatment according to different storage times to prepare n standard samples. (2) Acquisition of three-dimensional fluorescence spectral data of standard samples Three-dimensional fluorescence spectral data were collected from the n standard samples in step (1) to obtain n sets of three-dimensional fluorescence spectral data. Each set of three-dimensional fluorescence spectral data consisted of 18 sets of two-dimensional fluorescence spectral data. The acquisition parameters for the three-dimensional fluorescence spectroscopy data are: the difference between excitation and emission wavelengths Δλ = 10–180 nm, the excitation wavelength is 400–1100 nm, the excitation slit is 5 nm, and the emission slit is 5 nm. (3) Extracting characteristic variables of the three-dimensional fluorescence spectrum of the standard sample The parallel factor algorithm was used to extract the feature variables of each group of two-dimensional fluorescence spectra in step (2). The excitation wavelength with the highest fluorescence signal intensity was selected as the feature variable according to the loading value of the parallel factor algorithm. 351 fluorescence spectral feature data were obtained for each group of three-dimensional fluorescence spectral data. (4) Obtain the characteristic peak variables of the two-dimensional fluorescence spectrum of the standard. Extract the highest intensity peak and surrounding spectral data of the excitation spectrum of each group of two-dimensional fluorescence spectral data in step (2). Each group of two-dimensional fluorescence spectral data yields 11 spectral data points. Therefore, each group of three-dimensional fluorescence spectral data yields 198 characteristic variable data points. (5) Acquisition of low-field nuclear magnetic resonance spectral data of standard samples Low-field nuclear magnetic resonance spectral data were collected for each standard sample in step (1), and the low-field nuclear magnetic resonance spectral data of each standard sample contained 500 variables. (6) Fusion of standard sample characteristic data (6.1) Obtaining fluorescence feature-multi-feature peak fusion data For each standard sample, the fluorescence spectral characteristic data obtained in step (3) and the characteristic variable data of the three-dimensional fluorescence spectral data obtained in step (4) are fused to obtain 549 fluorescence characteristic-multi-characteristic peak fusion data; (6.2) Acquiring low-field NMR-fluorescence feature fusion data For each standard sample, the fluorescence spectral characteristic data obtained in step (3) is fused with the low-field nuclear magnetic resonance spectral data obtained in step (5) to obtain 851 low-field nuclear magnetic resonance-fluorescence characteristic fusion data; (7) Determination of physicochemical properties of standard products The peroxide value physicochemical index of n standard samples in step (1) was determined respectively to obtain the peroxide value of each standard sample; (8) Establish a rapid detection model for the storage time of edible vegetable oils. Using the low-field nuclear magnetic resonance spectral data of the standard sample in step (5), the fluorescence characteristic-multi-peak fusion data in step (6.1), and the low-field nuclear magnetic resonance-fluorescence characteristic fusion data in step (6.2) as input values, and the physicochemical indicators in step (7) as output values, a rapid detection model for the storage time of edible vegetable oil based on multispectral information fusion is established for the standard sample using the support vector machine regression algorithm; the support vector machine regression algorithm is optimized by the genetic algorithm GA. (9) Rapid determination of the sample to be tested Take the edible oil sample to be tested and obtain its low-field nuclear magnetic resonance spectral data, 549 fluorescence feature-multi-characteristic peak fusion data and 851 low-field nuclear magnetic resonance-fluorescence feature fusion data in the same way as in steps (2)-(6). Use the obtained low-field nuclear magnetic resonance spectral data, 549 fluorescence feature-multi-characteristic peak fusion data and 851 low-field nuclear magnetic resonance-fluorescence feature fusion data as input values and substitute them into the rapid detection model in step (8) to obtain the peroxide value physicochemical index of the edible oil sample to be tested. Calculate the storage time of the edible oil sample to be tested based on the obtained peroxide value physicochemical index.
2. The method for rapid determination of storage time of edible vegetable oil based on multispectral information fusion according to claim 1, characterized in that... The edible oil mentioned is edible vegetable oil.
3. The method for rapid determination of storage time of edible vegetable oil based on multispectral information fusion according to claim 1, characterized in that... The standard in step (1) is obtained by storing n fresh edible oil samples at room temperature for different times.