Method for detecting amylose content in high-amylose wheat noodles based on near infrared spectrum

Abstract

The invention belongs to the technical field of food nondestructive testing, and particularly relates to a method for detecting the content of starch in high-amylose wheat noodles based on a near infrared spectrum. The detection method provided by the invention comprises the following specific steps: collecting near infrared spectrums of high-amylose wheat noodles with different amylose contents, carrying out spectrum pretreatment, establishing a prediction model, and predicting the amylose content in an unknown noodle sample through the model. The detection method provided by the invention is simple to operate and high in automation, the detection process is non-destructive, the prediction precision of the amylose content is effectively improved by combining with the deep learning BP model, and a detection model with high stability and high accuracy is provided for rapid and nondestructive detection of the amylose content in the high-amylose wheat noodles.

Description

Technical Field

[0001] The invention belongs to the technical field of nondestructive food detection, and particularly relates to a method for detecting starch content in high-amylose wheat noodles based on near infrared spectroscopy. Background Art

[0002] High-amylose wheat noodles, due to their high amylose content, show stronger resistance to cooking after cooking, lower stickiness, and have better elasticity and toughness. In addition, they are rich in resistant starch, which helps lower the glycemic index and improve intestinal health. They are especially suitable for diabetic patients and consumers with low glycemic diet needs, and are therefore deeply loved by consumers.

[0003] However, this also makes high-amylose wheat noodles a target for adulteration by unscrupulous merchants. In order to reduce costs, some unscrupulous merchants use low-amylose wheat or other starches for adulteration, which not only reduces the quality of noodles, but also may threaten the health of consumers. Therefore, the adulteration of high-amylose wheat noodles has become an important issue in the field of food safety.

[0004] The content of amylose is one of the important indicators for evaluating the quality of high-amylose wheat noodles. The content will directly affect the edible quality, texture characteristics and digestibility of noodles.

[0005] Traditional methods for detecting the content of amylose mainly include iodine colorimetry, chromatography, enzymatic hydrolysis, etc. Although these methods are highly accurate, they usually require complex sample preparation processes, expensive chemical reagents, and long detection times. In view of this, it is particularly important to establish a rapid and effective method for detecting the content of amylose in noodles to identify and prevent the adulteration of high-amylose wheat noodles.

[0006] In recent years, near-infrared spectroscopy has been widely used in the field of food quality testing due to its advantages of being non-destructive, rapid and efficient. However, in grain processing, especially in the online real-time monitoring of grain terminal product processing and the determination of product characteristic indicators, there are few reports. In addition, the chemical composition of grain terminal products (such as moisture, starch, protein, etc.) may change their internal structure after being processed by physical and chemical methods. This change may exceed the data range during model training, resulting in uncertainty and stability in model establishment, increasing the difficulty of model establishment and introducing prediction errors.

[0007] Therefore, how to combine advanced data analysis methods, optimize the spectral data processing process, and build a high-precision prediction model for amylose content has become an urgent problem to be solved. Summary of the invention

[0008] In order to solve the above technical problems, the present invention provides a method for detecting the content of amylose in high-amylose wheat noodles, which can quickly and accurately detect the content of amylose in a sample.

[0009] The method for detecting the content of amylose in high-amylose wheat noodles based on near infrared spectroscopy provided by the present invention comprises the following steps:

[0010] (1) preparing calibration set samples, wherein the calibration set samples are noodles made from mixed flour to which waxy corn starch is added in an amount of 0% to 55% by weight of high-amylose wheat flour;

[0011] (2) using a near infrared spectrometer to collect spectral information of the calibration set samples in (1), and at the same time, using an iodine colorimetric method to determine the chemical content of amylose in the calibration set samples;

[0012] (3) performing spectral preprocessing on the near infrared spectra of the calibration set samples collected in (2);

[0013] (4) Screening characteristic variables of the spectral information obtained in (3);

[0014] (5) The characteristic variables selected in (4) are used to construct a quantitative prediction model for the content of amylose in the calibration set samples.

[0015] In the above method, preferably, in (1), the calibration set samples are obtained by the following method: different contents of waxy corn starch are added to high-amylose wheat flour samples, and the mixture is fully mixed to obtain mixed flour, wherein the added mass of the waxy corn starch is 0%-55% based on the mass of the high-amylose wheat flour samples;

[0016] Next, weigh the mixed flour, add 40%-50% of water to the mixed flour, stir and mix, put it into a constant temperature and humidity incubator at 20-25°C for 10-30 minutes, then roll it 3-8 times, and finally press it into noodles, cool it at room temperature and set aside.

[0017] Preferably, in (2), the acquisition conditions of the near-infrared spectrometer are: the scanning spectrum wave number range is 4000 to 10000 cm -1 , resolution 6~10cm -1 , data interval is 0.5~2cm -1 Each sample was scanned 32 to 64 times and the average was taken.

[0018] Preferably, the preprocessing described in (3) is specifically: using partial least squares (PLS) in combination with at least one of the following methods: standard normal variable transformation (SNV), multiplicative scatter correction (MSC), normalization, detrend, and Savizky-Golay convolution smoothing method (SG).

[0019] As a further preferred embodiment, the preprocessing described in (3) is specifically: using PLS and Detrend for joint processing.

[0020] Preferably, in (4), at least one of PLS ​​combined with competitive adaptive reweighted sampling (CARS), random frog algorithm (RFr), and variable combination population analysis algorithm (VCPA) is used to screen the variables.

[0021] As a further preferred method, in (4), PLS combined with CARS is used to screen the variables.

[0022] Preferably, in (5), the modeling method uses at least one of back propagation neural network (BP), partial least squares regression (PLSR), and random forest (RF) for modeling.

[0023] The model performance is expressed by the coefficient of determination of the calibration and prediction sets ( and ), the root mean square error of the calibration set and the prediction set (RMSEC and RMSEP). Usually the highest coefficient of determination and the lowest root mean square error indicate that the model has good predictive ability, among which the coefficient of determination R 2 ≥0.95, root mean square error RMSE≤1, and the best model is selected by comparison. The beneficial effects of the present invention are:

[0024] (1) The near-infrared spectroscopy technique was used to rapidly detect the amylose content in high-amylose wheat noodles. The detection process is non-destructive and does not require complicated sample preparation operations or redundant chemical reagents. It is suitable for screening large quantities of samples and is easy to operate and highly automated.

[0025] (2) By screening the key wavelength range and combining multiple models to predict the results, the errors that may occur in a single model are avoided, and the accuracy of predicting the amylose content in high-amylose wheat noodles is effectively improved. This provides a detection model with high accuracy and good stability for the rapid and nondestructive detection of the amylose content in high-amylose wheat noodles. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is the original near-infrared spectrum of the calibration set samples in Example 1 of the present invention;

[0027] Figure 2 A graph showing a change trend in the number of sampled variables of the CARS algorithm in Example 1 of the present invention;

[0028] Figure 3 A distribution diagram of the variables selected by the CARS algorithm in Example 1 of the present invention;

[0029] Figure 4 This is a scatter plot of the model predicted values ​​and measured values ​​of the amylose content in the high-amylose wheat noodles in Example 1 of the present invention. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the present invention, the present invention is further described in conjunction with specific implementation methods.

[0031] Example 1

[0032] 1.1 The method for detecting the content of amylose in high-amylose wheat noodles based on near-infrared spectroscopy is as follows:

[0033] (1) Preparation of calibration set samples

[0034] Adding different amounts of glutinous corn starch into high-amylose wheat flour, mixing thoroughly, and obtaining mixed flour;

[0035] Based on the mass of high-amylose wheat flour, the added mass of waxy corn starch is: 0%, 5%, 10%, 15%, 20%, 25%, 30%, 35%, 40%, 45%, 50%, 55%;

[0036] Next, 100 g of the mixed powders of different proportions were weighed, 45 g of water was added, the mixture was stirred and mixed, and then placed in a constant temperature and humidity incubator at 25° C. for 20 min, and then rolled for 5 times, and finally pressed into noodles with a thickness of 1.5 mm, and cooled at room temperature for later use;

[0037] (2) using a near infrared spectrometer to collect spectral information of the calibration set samples prepared in (1) to obtain near infrared spectra of the calibration set samples;

[0038] The acquisition conditions of the near-infrared spectrometer are as follows: the scanning spectrum wave number range is 4000~10000cm -1 , resolution 6~10cm -1 , data interval is 0.5~2cm -1 , each sample was scanned 32 to 64 times and averaged;

[0039] At the same time, the chemical content of amylose in the calibration set samples was determined by iodine colorimetry;

[0040] (3) Preprocessing the near infrared spectra of the calibration set samples collected in (2), specifically using PLS combined with SNV, MSC, Normalization, Detrend, SG and other methods for preprocessing;

[0041] (4) PLS combined with CARS, RFr, VCPA and other methods were used to screen variables and obtain characteristic variables;

[0042] (5) The characteristic variables selected in (4) were used to construct different quantitative prediction models for the content of amylose in high-amylose wheat noodles. The models were screened using BP, PLSR, RF and other methods to obtain the best model.

[0043] 1.2 The near infrared spectra of the calibration set samples obtained in this example are shown in the attached Figure 1 shown.

[0044] Attached Figure 1 Visible, 4700cm -1 The absorption peak near 5500 cm comes from the interaction between OH bending and CO stretching, which is related to the starch content. -1 Up to 6000cm -1 The peak mainly reflects the fundamental frequency stretching of -CH and the overtone vibrations of -CH2 and -CH3, which indicates that both starch and gluten proteins in flour contribute.

[0045] 1.3 Effect of spectral preprocessing methods on PLS modeling results

[0046] In step (3) of the present invention, five preprocessing methods are used for comparison, and PLS is combined with SNV, MSC, Normalization, Detrend, SG and other methods to eliminate the effects of temperature fluctuation, wavelength drift and baseline drift.

[0047] Table 1 shows the prediction performance of the PLS correction model for original spectra and different spectral preprocessing methods.

[0048] Table 1 Effect of different spectral processing methods on PLS modeling

[0049]

[0050]

[0051] The data in Table 1 show that when the PLS correction model is established using the spectrum processed by Detrend, the results on the test set change significantly, reaching R 2 PThe root mean square error (RMSEP) of the test set is 0.9859, and the determination coefficient R 2 ≥0.95, root mean square error (RMSE) ≤1.

[0052] 1.4 Screening of characteristic variables

[0053] Generally speaking, full spectrum information can be used for qualitative and quantitative analysis, but too much redundant information will significantly affect the speed of model operation. The selection of characteristic wavelengths is crucial for spectral analysis because it can effectively improve the robustness and prediction accuracy of the model. Feature selection not only helps to reduce the number of input variables, but also reduces the complexity of the model.

[0054] The present invention adopts CARS, VCPA, RFr and other algorithms in combination with PLS to screen the characteristic wavelengths of variables, aiming to improve the analysis and prediction capabilities of the model and select the best screening method for modeling.

[0055] Based on the preprocessing of Detrend spectral data, the CARS algorithm was used to further predict the characteristic spectral variables of amylose content in high-amylose content wheat noodles.

[0056] Attached Figure 2 It reflects the changing trend of the number of sampling variables of the CARS algorithm, such as Figure 2 As shown in (a), the number of selected variables decreases rapidly at the beginning and then decreases slowly in the later stages. This selection process from rapid to gradual reflects the nature of CARS from rough selection to fine selection, which improves the efficiency of variable selection.

[0057] exist Figure 2 In (b), as the number of sampling rounds increases, the corrected root mean square error (RMSECV) value decreases initially and then increases. Figure 2 It can be noted in (c) that CARS-PLS is run 50 times, and the one with the lowest RMSECV is selected. The information screened out this time is used as the established feature variable. The pink vertical line shows that when the sampling number is 31 times, its RMSECV reaches the minimum. At the same time, the 31st sampling round corresponds to 34 feature variables. Finally, 34 feature variables are screened out.

[0058] Attached Figure 3 The distribution location plot of the variables selected by the CARS algorithm is shown.

[0059] By the attached Figure 3 It can be seen that the selected wavelength points for CARS are clustered around the absorption peaks in the near-infrared spectrum, in particular, around 4700 cm -1The wavenumbers nearby mainly correspond to the absorption peaks from the interaction of OH bending and CO stretching, which is a feature related to the starch content of high-amylose wheat noodles.

[0060] 1.5 Establishment of quantitative prediction model

[0061] Figure 4 The regression fit plot of high amylose wheat noodles predicted by the CARS-BP model is shown. The solid line is the regression line corresponding to the ideal correlation between the measured and predicted values, given by Figure 4 It can be seen that for the calibration set and the prediction set, the predicted values ​​are basically close to the corresponding actual values. Among all the models, CARS-BP has the best prediction results. is 0.9921, RMSEP is 0.7819, and the prediction set determination coefficient is reached The requirement of root mean square error RMSEP ≤ 1 of the prediction set indicates that the regression effect is good and shows good robustness.

[0062] Based on the above experiments, it can be seen that the method for detecting the content of amylose in high-amylose wheat noodles of the present invention is feasible, the determination process is fast and the results are accurate and reliable.

Claims

1. A method for detecting the content of amylose in high-amylose wheat noodles based on near infrared spectroscopy, characterized in that: The following steps are involved: (1) preparing a calibration set of samples, wherein the calibration set of samples is noodles made from a mixture of glutinous corn starch in an amount of 0% to 55% by weight of high-amylose wheat flour; (2) using a near infrared spectrometer to collect spectral information of the calibration set samples in (1), and using an iodine colorimetric method to determine the chemical content of amylose in the calibration set samples; (3) performing spectral preprocessing on the near infrared spectra of the calibration set samples collected in (2); (4) Screening characteristic variables of the spectral information obtained in (3); (5) The characteristic variables selected in (4) are used to construct a quantitative prediction model for the amylose content in the calibration set samples.

2. The method for detecting the content of amylose in high-amylose wheat noodles based on near infrared spectroscopy according to claim 1, characterized in that: In (1), the calibration set samples are obtained by the following method: Adding different amounts of waxy corn starch to high-amylose wheat flour samples, respectively, and mixing them thoroughly to obtain mixed flour, wherein the added mass of the waxy corn starch is 0%-55% based on the mass of the high-amylose wheat flour samples; Next, weigh the mixed flour, add 40%-50% of water to the mixed flour, stir and mix, put it into a constant temperature and humidity incubator at 20-25°C for 10-30 min, then roll it 3-8 times, and finally press it into noodles, cool it at room temperature and set aside.

3. The method for detecting the content of amylose in high-amylose wheat noodles based on near infrared spectroscopy according to claim 1, characterized in that: (2) In the acquisition conditions of the near-infrared spectrometer, the scanning spectrum wave number range is 4000~10000 cm -1 , resolution 6~10 cm -1 , data interval is 0.5~2 cm -1 Each sample was scanned 32 to 64 times and the average was taken.

4. The method for detecting the content of amylose in high-amylose wheat noodles based on near infrared spectroscopy according to claim 1, characterized in that: The preprocessing described in (3) is specifically: using partial least squares method in combination with at least one of the following methods: standard normal variable transformation, multiplicative scatter correction, normalization, detrending, and Savizky-Golay convolution smoothing method.

5. The method for detecting the content of amylose in high-amylose wheat noodles based on near infrared spectroscopy according to claim 1, characterized in that: In (4), the variables are screened by using the partial least squares method combined with at least one of the competitive adaptive reweighted sampling method, the random frog algorithm, and the variable combination population analysis algorithm.

6. The method for detecting the content of amylose in high-amylose wheat noodles based on near infrared spectroscopy according to claim 1, characterized in that: In (5), the modeling method uses at least one of back propagation neural network, partial least squares regression, and random forest to perform modeling.

Images