Construction method and application of typical robust meat dish intelligent characteristic spectrum recognition system

Abstract

The invention discloses a construction method and application of an intelligent characteristic spectrum recognition system for typical robust meat dishes, and belongs to the technical field of food detection. In order to overcome the defects of research on the quality and characteristics of Lu vegetable in the prior art, near infrared spectrum and Raman spectrum technologies are utilized for the first time, and an intelligent characteristic spectrum recognition system for typical robust meat dishes is researched and established; the digital fingerprint spectrum of typical robust meat dishes established based on the system can be used for scientifically, objectively and digitally evaluating, classifying and judging the quality of the Luercai, and can be used for intelligently displaying the characteristics of the Luercai, so that an effective tool is provided for guaranteeing and improving the quality of the robust meat dishes; and a new thought and technical support are provided for the development and propagation of the Lu vegetable.

Description

Technical Field

[0001] The present invention relates to the field of food detection technology, and in particular to a construction method and application of an intelligent characteristic spectrum recognition system for typical Shandong-style meat dishes. Background Art

[0002] Shandong cuisine, one of China's four major culinary styles, boasts a long history and unique flavors, making it an integral part of Chinese culinary culture. Currently, domestic research on Shandong cuisine focuses primarily on the following aspects: first, a review and summary of its history, culture, regional characteristics, and flavor profile; second, optimization and improvement of its processing techniques, recipes, and seasonings; and third, sensory evaluation, physical and chemical analysis, and instrumental analysis of its edible qualities. However, these studies still have some shortcomings, such as a lack of comprehensive evaluation methods for Shandong cuisine's quality, neglect of the correlation between edible quality and flavor compounds, and a lack of digitally representative methods for showcasing its unique characteristics.

[0003] In foreign markets, Sichuan and Cantonese cuisines dominate Chinese cuisine, while Shandong cuisine, the foremost of China's four traditional culinary styles, remains relatively unknown. This is due to both the unique characteristics of Shandong cuisine and a lack of effective promotion and display in international markets. Therefore, research on the quality evaluation and unique presentation of Shandong cuisine holds significant theoretical and practical value.

[0004] Shandong cuisine is an important part of traditional Chinese food culture. Scientific evaluation and display of its quality and characteristics are conducive to the protection and inheritance of its historical culture. Currently, there is no digital fingerprint of the quality of typical Shandong dishes in existing technologies, which limits the research on the quality and characteristics of Shandong cuisine. Summary of the Invention

[0005] In response to the problems existing in the prior art, the purpose of the present invention is to provide a method for constructing and applying an intelligent characteristic spectrum recognition system for typical Shandong-style meat dishes. By studying three typical Shandong-style meat dishes (Shandong fried chicken, Zhucheng roast pork and Four-happiness meatballs), the present invention establishes a digital fingerprint of the quality of typical Shandong-style dishes, providing an effective tool for quality assurance and improvement of Shandong-style meat dishes.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for constructing an intelligent characteristic spectrum recognition system for typical Shandong-style meat dishes, the method comprising the following steps:

[0008] (1) Process typical Luwei meat dishes and determine the optimal processing parameters;

[0009] (2) detecting the sample processed in step (1) using near-infrared spectroscopy, Raman spectroscopy, or a combination of the two, preprocessing the collected near-infrared spectroscopy data and Raman spectroscopy data, extracting data analysis features, and establishing a classification and discrimination model based on near-infrared spectroscopy, Raman spectroscopy, or a combination of the two;

[0010] (3) Evaluate the accuracy and stability of the model obtained in step (2), establish a digital fingerprint of typical Shandong-style meat dishes, and complete the construction of an intelligent characteristic spectrum recognition system for typical Shandong-style meat dishes.

[0011] In step (1), the typical Shandong-style meat dishes tested are Shandong fried chicken, Zhucheng roast pork and four-happiness meatballs.

[0012] In step (1), the process includes: selecting, cutting, marinating and cooking typical Shandong-flavored meat dishes test samples.

[0013] In step (2), the preprocessing includes: cleaning, smoothing, and normalizing the data.

[0014] In a second aspect, the present invention provides a digital fingerprint of typical Luwei meat dishes established based on the above construction method.

[0015] The third aspect of the present invention provides the application of the digital fingerprint of the typical Luwei meat dishes in the following (1) or (2):

[0016] (1) Achieve a comprehensive evaluation of the quality of Luwei meat dishes;

[0017] (2) To classify and identify the quality of Shandong-style meat dishes;

[0018] (3) Realize the digital display of Shandong-style meat dishes.

[0019] Beneficial effects of the present invention:

[0020] In response to the deficiencies in the existing technology in the research on the quality and characteristics of Shandong cuisine, the present invention uses near-infrared spectroscopy and Raman spectroscopy technology for the first time to study and establish an intelligent characteristic spectral recognition system for typical Shandong-style meat dishes. The digital fingerprint of typical Shandong-style meat dishes established based on this system can scientifically, objectively and digitally evaluate, classify and identify the quality of Shandong cuisine, and can realize the intelligent display of the characteristics of Shandong cuisine, providing an effective tool for the quality assurance and improvement of Shandong-style meat dishes, and offering new ideas and technical support for the development and dissemination of Shandong cuisine. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The effect of different smoking times on the color of Zhucheng roast meat.

[0022] Figure 2 The effect of different amounts of lotus root starch on the color of four-happiness meatballs.

[0023] Figure 3 The effect of different ginger addition amounts on the color of Shandong fried chicken.

[0024] Figure 4 Near-infrared spectra of Taishan fried chicken PCA (A), LDA (B), PLS-DA (C) and PLSR (D).

[0025] Figure 5 Near-infrared spectra of Zhucheng roast pork PCA (A), LDA (B), PLS-DA (C) and PLSR (D).

[0026] Figure 6 Raman spectra of Sixi meatballs PCA (A), LDA (B), PLS-DA (C), and PLSR (D).

[0027] Figure 7 Near-infrared spectra of Sixi meatballs PCA (A), LDA (B), PLS-DA (C) and PLSR (D).

[0028] Figure 8 Raman spectra of Zhucheng roast pork PCA (A), LDA (B), PLS-DA (C) and PLSR (D).

[0029] Figure 9 Raman spectra of Taishan fried chicken PCA (A), LDA (B), PLS-DA (C) and PLSR (D). DETAILED DESCRIPTION

[0030] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present application belongs.

[0031] The following detailed description is for illustrative purposes only and is intended to provide further explanation of the present invention, rather than to limit the scope of the present invention.

[0032] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present application belongs.

[0033] The present invention can utilize near-infrared spectroscopy technology, Raman spectroscopy technology or a combination of near-infrared spectroscopy technology and Raman spectroscopy technology to establish a classification and discrimination model, thereby establishing a digital fingerprint of the quality of typical Shandong-style dishes.

[0034] The following detailed description is for illustrative purposes only and is intended to provide further explanation of the present invention, rather than to limit the scope of the present invention.

[0035] Example 1: Quality Analysis of Typical Luwei Meat Dishes

[0036] Shandong Taishan Fried Chicken, Zhucheng Roast Pork, and Four-Happiness Meatballs were prepared by adjusting the seasoning mass ratio and cooking method. Table 1 shows the effects of different smoking times on the TPA of Zhucheng Roast Pork. Textural indices show that increasing smoking time significantly influences TPA (P < 0.05). A smoking time of 15 minutes showed significant differences in hardness, cohesion, and resilience compared to a 5-minute smoking time, but no significant differences compared to a 25-minute smoking time. Compared to a 15-minute smoking time, a 25-minute smoking time showed significant increases in hardness and cohesion, but a significant decrease in cohesion and resilience (P < 0.05). This suggests that a 25-minute smoking time results in a firmer texture overall.

[0037] like Figure 1 As shown, 15min smoking group L * and a * The color of the 15-min group was significantly higher than that of the 5-min group and the 25-min group (P<0.05), and the 15-min group showed the brightest color. This may be due to the denaturation of surface proteins and the initial products of the Maillard reaction in the early stage of smoking, as well as the color deterioration caused by excessive caramelization or lipid oxidation of the surface caused by long-term smoking. * It increases significantly with the increase of smoking time, indicating that the color-forming substances (such as nitrosylmyoglobin derivatives) are continuously generated during the smoking process. Overall, smoking for 15 minutes has a significant effect on improving the color of Zhucheng barbecued meat, making it have the brightest red color.

[0038] Table 1 Effects of different smoking times on TPA of Zhucheng Shaorou

[0039]

[0040]

[0041] Table 2 shows the effects of different amounts of lotus root starch added (1.5%, 2%, 2.5%) on the indicators of Four Joy Meatballs (TPA). In terms of hardness, the hardness of the Four Joy Meatballs with 2% lotus root starch added was significantly higher than that with 1.5% and 2.5% additions (P<0.05), indicating that a moderate increase in the amount of lotus root starch can increase the hardness, but excessive addition (2.5%) will reduce the hardness. In terms of elasticity, as the amount of lotus root starch added increased from 1.5% to 2.5%, there was a downward trend, indicating that the addition of lotus root starch would weaken the elasticity of the meatballs. Cohesiveness increased with the increase in addition amount, which means that lotus root starch can enhance the internal bonding force of the meatballs. The adhesiveness was the highest at a 2% addition amount, reflecting that the meatballs had a stronger ability to resist deformation due to external forces at this addition amount. There was no significant difference in chewiness at different addition amounts, while resilience was reached at 2.5%, which was slightly higher than the other two groups.

[0042] like Figure 2 As shown in the figure, when the amount of lotus root starch increased from 1.5% to 2.0%, L * Significantly decreased, a * 、b * The results showed that the color of the meat was significantly increased (P<0.05), indicating that lotus root starch significantly deepened the color of the meat by promoting the Maillard reaction and caramelization reaction at this stage; and when the addition amount was further increased to 2.5%, L * Significantly increased, a * 、b * The color stability of lotus root starch decreased significantly (P<0.05). Excessive lotus root starch may cause protein aggregation or increased lipid oxidation, resulting in decreased color stability.

[0043] Table 2 Effects of different amounts of lotus root starch added on TPA of Four Joy Meatballs

[0044]

[0045] As shown in Table 3, the effects of different ginger addition amounts (24%, 30%, 36%) on the TPA index of Taishan Fried Chicken. In terms of hardness, the hardness of Taishan Fried Chicken with 30% ginger addition was significantly lower than that with 24% and 36% addition amounts (P<0.05), and the adhesion index increased significantly with the increase in addition amount (P<0.05), indicating a significant enhancement in the internal bonding strength of the sample (P<0.05). The elastic parameter reached a peak at 36% addition, indicating that the sample's ability to recover after deformation was significantly improved. The viscosity dimension showed a nonlinear change characteristic, and the adhesion of the 30% treatment group was significantly higher than that of the 24% group. The chewiness of the 30% ginger addition was significantly lower than that of the other two groups (P<0.05).

[0046] according to Figure 3 As shown in the figure, different ginger addition amounts had a significant effect on the three key indicators of Shandong fried chicken, among which the a * Significantly higher than 24% and 36% of the supplemented groups, L * and b* The results showed that the addition of 30% ginger had a significant advantage in optimizing the color of Shandong fried chicken, which provided an important theoretical support for the precise regulation of the addition amount of ginger in actual production and helped to improve the color quality of Shandong fried chicken (P<0.05).

[0047] Table 3 Effects of different ginger addition amounts on TPA of Taishan fried chicken

[0048]

[0049] Example 2: Flavor Quality Analysis of Luwei Meat Dishes

[0050] To optimize the stir-frying process, Shandong stir-fried chicken was analyzed by GC-MS. As shown in Table 4, a total of 36 volatile flavor compounds were identified, including 2 aldehydes, 1 ketone, 3 alcohols, 4 esters, and 1 acid. Other volatile flavor compounds were primarily chlorinated compounds and sulfides. Shandong stir-fried chicken is dominated by fat oxidation products, hexanal and nonanal, and alcohols, exhibiting a greasy, grassy, ​​and slightly earthy aroma, likely related to the high-temperature stir-frying process. The flavor of Shandong stir-fried chicken is primarily composed of fat oxidation products and alcohols. Alcohols are produced by fat oxidation, amino acid metabolism, or carbohydrate metabolism. Generally, saturated alcohols are produced during heating through the oxidative decomposition of fat or by the reduction of carbonyl compounds, resulting in higher thresholds. Unsaturated alcohols, on the other hand, have lower thresholds and can affect the flavor of aquatic products. Compared to Four Joy Meatballs, stir-fried chicken contains fewer aldehydes and alcohols, but has a higher content of 1-pentanol, possibly related to the use of cooking wine or fermented seasonings. Four Joy Meatballs are dominated by fatty acids, offering a rich, fruity, and meaty aroma. Their rich, layered flavor embodies the fusion of fat and spices found in traditional meatballs. Stir-fried chicken, on the other hand, relies more heavily on aldehydes and alcohols for its flavor.

[0051] Table 4 Volatile flavor substances of Taishan fried chicken (mg / kg)

[0052]

[0053]

[0054] Table 5 shows that the main aroma compounds detected in barbecued meat products by solid-phase microextraction and gas chromatography-mass spectrometry (GC-MS) are alcohols, aldehydes, and heterocyclic compounds. Among them, aldehydes have the highest content and a very low threshold, contributing most to the barbecued meat aroma. Furfural has the greatest impact, followed by 5-methylfurfural. Heterocyclic compounds, such as 1-(2-furyl)ethanone and furfurylmethanol, have the second highest content and a lower threshold, contributing significantly to the barbecued meat aroma. Alcohols, while lower in content than aldehydes and heterocyclic compounds, contribute significantly to the barbecued meat aroma even with higher thresholds. Esters, acids, ethers, and ketones are present in very low amounts and have higher thresholds, contributing little to the barbecued meat aroma. Compared to Four Joy Meatballs and Shandong Fried Chicken, Zhucheng barbecued meat has a higher content of aldehydes, likely due to the high-temperature cooking process, which results in the Maillard reaction products furfural and methylfurfural. These compounds impart a pronounced caramel and toasted aroma, but may also impart a bitter taste. Aldehydes and alcohols are key flavors shared by all three dishes.

[0055] Table 5 Volatile flavor substances of Zhucheng smoked meat (mg / kg)

[0056]

[0057]

[0058] As shown in Table 6, the processing technology for Four Joy Meatballs was optimized using GC-MS (gas chromatography-mass spectrometry) and volatile compounds were analyzed. While food flavor often originates naturally, added substances can also contribute to it. The interactions between these substances during the cooking process produce flavor compounds such as aldehydes and alcohols, which impart a unique flavor to the meatballs. As shown in Table 1, a total of 32 volatile flavor compounds were identified, including 10 aldehydes, 6 terpenes, 5 alcohols, and 3 esters. Other volatile flavor compounds included acids, hydrocarbons, and sulfides. Oleic acid and n-hexadecanoic acid were found in particularly high concentrations, dominating the oil and fat flavors. Compounds such as n-hexane and methylcyclopentane may have a lesser impact on the flavor. Studies have shown that aldehydes and alcohols contribute significantly to the flavor of Four Joy Meatballs, particularly decanal, octanal, nonanal, and phenylethanol. Aldehydes often have nutty and fatty notes and contribute significantly to the flavor of meat products. This is primarily due to their high concentrations and low olfactory thresholds, resulting in a fatty aroma and being a key component of meat flavor. Aldehydes such as hexanal, nonanal, octanal, dodecanal, citral, and decanal were detected during the production of Four Joy Meatballs. Octanal and nonanal primarily originate from the oxidation of oleic acid, imparting an oily aroma. Ginger also contains significant amounts of octanal. Hexanal primarily originates from the oxidation of ω-6 unsaturated fatty acids. At low concentrations, it has a grassy aroma, but at high concentrations, it can produce an unpleasant odor. Alcohols are primarily produced by the oxidative decomposition of fats upon heating. Hexanal, primarily produced by the oxidation of linoleic acid and arachidonic acid, has a rich vanilla flavor. Terpenes such as camphene, eucalyptol, and linalool, likely derived from added spices or ingredients like pepper and star anise, impart woody, floral, or medicinal aromas. Cyclopentadecanone is present in high concentrations, potentially contributing to the unique aroma of Four Joy Meatballs. Esters often have a sweet, fruity aroma, primarily derived from the interaction between alcohols produced by oxidation of pork lipids during heating and free fatty acids, giving the meat its characteristic aroma. However, as Table 1 shows, esters are relatively rare in the sample, and due to their high threshold, their impact on flavor is minimal. Therefore, their contribution to the overall flavor of the Four Joy Meatballs is not significant. However, the extremely high content of palmitic acid glyceride may be related to the oily flavor. In summary, aldehydes, terpenes, and high levels of fatty acids are the core of the Four Joy Meatballs' flavor, with citral, cyclopentadecanone, and 1-ethynylcyclododecanol likely contributing to the distinctive flavor.

[0059] Table 6 Volatile flavor substances of Sixi meatballs (mg / kg)

[0060]

[0061]

[0062] Example 3: Research on the Construction of Fingerprints of Luwei Meat Dishes

[0063] The software Labspec6.1.1 was used for spectral evaluation and processing. Polynomial fitting was used to perform baseline calibration on the original spectrum to eliminate background interference and improve signal authenticity, so as to better reflect the peak information of the sample. The spectral data was then smoothed by the SG convolution polynomial, and finally area normalization was performed so that the spectra of different samples were in the same range, which was convenient for direct comparison or statistical analysis. PCA is used to reduce the dimension of spectral data, eliminate noise interference and redundant information, improve the efficiency and generalization ability of subsequent machine learning models, and draw scatter plots intuitively. At the same time, LDA clustering was combined with the LDA model to observe the distribution, clustering and outliers of the data, and to make judgments based on the resulting confusion matrix. Finally, the regression coefficient VIP score was analyzed to find the degree of contribution of the substances corresponding to different peaks to the realization of discrimination.

[0064] The principal component analysis (PCA) diagram shows the distribution of near-infrared spectral data of Taishan Fried Chicken under different conditions (24d, 36d, b1) in PC1 and PC2 dimensions. Different colored points represent each group of samples, and the ellipse defines the data distribution range. It can be seen that the sample distribution is both overlapping and discrete, reflecting the similarities and differences in the chemical composition in the principal component space. The linear discriminant analysis (LDA) results show that the overall accuracy of the model is 88.24%. The confusion matrix and sensitivity and specificity data show that the sensitivity of the 24d group is 60.00% and the specificity is 100.00%; the sensitivity and specificity of the 36d group are 100.00% and 90.91% respectively; the sensitivity and specificity of the b1 group are both 100.00% and 90.91%, indicating that the LDA model has a good ability to distinguish samples under different conditions, but there is a certain amount of misjudgment in the 24d group. In the VIP diagram of the partial least squares discriminant analysis (PLS-DA), each Raman shift (such as 7652.2cm -1 、7650.2cm -1 The VIP values ​​of 7778.9cm and 7778.9cm are much higher than 1, indicating that these bands are key variables for distinguishing samples under different conditions and make outstanding contributions to model discrimination. In the regression coefficient diagram of partial least squares regression (PLSR), the positive and negative coefficients and their magnitude reflect the direction and strength of the correlation between the corresponding Raman shift and the predicted variable, such as 7778.9cm -1 The positive coefficient indicates that the signal enhancement is positively correlated with the predictor variable, 9686.7cm -1 Negative coefficients indicate negative correlation, and the bands with large absolute values ​​of these coefficients are crucial for prediction ( Figure 4 ).

[0065] The principal component analysis (PCA) diagram shows the distribution of near-infrared spectral data of Zhucheng barbecued meat at different processing times (5m, 15m, and 25m) in the PC1 and PC2 dimensions. The different colored points represent each group of samples, and the ellipse represents the data distribution range. This shows that the spectral characteristics of each group of samples have both overlap (similarity) and dispersion (difference), reflecting the distribution pattern of chemical components in the principal component space. The linear discriminant analysis (LDA) results show that the overall accuracy of the model is 74.07%. The confusion matrix shows that 7 samples in the 5m group were correctly predicted (sensitivity 77.78%, specificity 77.78%), 6 samples in the 15m group were correctly predicted (sensitivity 66.67% and specificity 88.89%), and 7 samples in the 25m group were correctly predicted (sensitivity 77.78% and specificity 94.44%). This shows that LDA has a certain ability to distinguish samples with different processing times, but there are some misjudgments. In the VIP diagram of the partial least squares discriminant analysis (PLS-DA), the Raman shifts (such as 785.7cm -1 、781.9cm -1 The VIP values ​​of the three bands (e.g., 785.7cm) are significantly higher than 1 (red dotted line), indicating that these bands are key spectral features for distinguishing samples with different processing times and make outstanding contributions to model discrimination. In the regression coefficient diagram of partial least squares regression (PLSR), the sign and magnitude of the coefficient reflect the direction and strength of the correlation between the corresponding Raman shift and the predicted variable. -1 、781.9cm -1 ) indicates that signal enhancement is positively correlated with the predictor variable, and negative coefficients (such as 546.4cm -1 、547.1cm -1 ) are negatively correlated, and the bands with larger absolute values ​​of these coefficients contribute significantly to the prediction, revealing the intrinsic connection between the near-infrared spectrum and the characteristics of Zhucheng barbecue samples (such as changes in composition related to processing time). Taken together, these analyses reveal the near-infrared spectral characteristics of Zhucheng barbecue at different processing times from different dimensions: PCA shows data distribution differences, LDA evaluates classification effects, PLS-DA and PLSR determine key spectral variables, providing a spectroscopic basis for in-depth exploration of the changes in Zhucheng barbecue ingredients with processing time, and helping to analyze the relationship between its chemical properties and processing time at the molecular level ( Figure 5 ).

[0066] Judging from the results of Raman spectroscopy analysis of Four Joy Meatballs, the c1.5, c2, and c2.5 groups exhibit the following characteristics: In the PCA diagram, there is significant overlap between these three groups of data points, indicating that their Raman spectral features have a high degree of similarity, and are difficult to clearly distinguish in the comprehensive spectral dimensions represented by the PC1 and PC2 principal components, reflecting that the chemical composition or molecular structure of the three groups of samples differ slightly in these dimensions. LDA analysis shows that the sensitivity of the c1.5, c2, and c2.5 groups is 0.00%, and the model is completely unable to correctly identify them. In the confusion matrix, these three groups of samples are largely misclassified as other groups, indicating that based on the current Raman spectral data, the LDA model is difficult to capture the features that effectively distinguish these three groups, and the spectral differences between the groups fail to form a significant distinction in the discriminant model. In the VIP diagram of PLS-DA, although there are high VIP value Raman shifts that make an important contribution to group differentiation (such as 1535.1cm -1 、420.6cm -1 etc.), but combined with the LDA results, it is speculated that these key variables may not form sufficient differences between the c1.5, c2, and c2.5 groups or there may be cross-changes in the change trends, resulting in the model being unable to effectively distinguish. The PLSR regression coefficient diagram shows the direction and strength of the association between each Raman shift and the predicted variable. However, given that LDA failed to identify these three groups, it is speculated that the spectral response differences of the c1.5, c2, and c2.5 groups in the key Raman shift are small, and their differences cannot form an effective basis for discrimination in the model, making it difficult for PLSR analysis based on Raman spectroscopy to accurately associate with specific groups ( Figure 6 ).

[0067] Judging from the results of the near-infrared spectrum analysis of the Four Joy Meatballs, the c1.5, c2, and c2.5 groups exhibit the following characteristics in terms of spectral features and model discrimination: In the PCA diagram, there is obvious overlap between the three groups of data points, indicating that their near-infrared spectral features have high similarity, and it is difficult to clearly distinguish them in the comprehensive spectral dimensions represented by the PC1 and PC2 principal components, reflecting that the chemical composition or molecular structure of the three groups of samples are relatively subtle in these dimensions. LDA analysis shows that the sensitivity of the c1.5, c2, and c2.5 groups is 0.00%, and the model is completely unable to correctly identify them. In the confusion matrix, these three groups of samples are largely misclassified as other groups, indicating that based on the current near-infrared spectral data, the LDA model is difficult to capture the features that effectively distinguish these three groups, and the spectral differences between the groups fail to form a significant distinction in the discriminant model. In the VIP diagram of PLS-DA, although there are high VIP value bands that make an important contribution to group differentiation (such as 7172.0cm -1 、7181.6cm -1etc.), but combined with the LDA results, it is speculated that these key variables may not form sufficient differences among the c1.5, c2, and c2.5 groups or there is an intersection in the change trends, resulting in the model being unable to effectively distinguish. The PLSR regression coefficient diagram shows the direction and strength of the association between each band and the predictor variable. However, given that LDA failed to identify these three groups, it is speculated that the spectral response differences of the c1.5, c2, and c2.5 groups in the key bands are small, and their differences fail to form an effective basis for discrimination in the model, making it difficult for the PLSR analysis based on near-infrared spectroscopy to be accurately associated with a specific group. In summary, the c1.5, c2, and c2.5 groups are highly similar in near-infrared spectral characteristics, and the existing LDA model cannot effectively distinguish them. Although PLS-DA identified important spectral variables, these variables failed to construct a significant discrimination boundary between the three groups, suggesting that the differences in chemical composition or structure of the three groups of samples are relatively subtle, and further exploration is needed in combination with more sophisticated experimental design or deep spectral analysis ( Figure 7 ).

[0068] First, a principal component analysis (PCA) plot illustrates the distribution characteristics of Raman spectral data from Zhucheng barbecued meat at different processing times (5, 15, and 25 minutes). PC1 and PC2 are the first and second principal components, respectively. Different colored points represent sample groups, and the ellipse represents the data distribution range. The data points within each group exhibit some overlap and dispersion, indicating that the spectral characteristics of samples processed at different times exhibit both similarities and differences, reflecting the distribution patterns of their chemical components within the principal component space. Linear discriminant analysis (LDA) results show an overall model accuracy of 72.22%. In the confusion matrix, all six samples in the 5-minute group were correctly predicted, achieving a sensitivity of 100.00% and a specificity of 66.67%. The 15-minute group achieved a sensitivity of 50.00% (three correct predictions, two misclassifications as 5-minute samples, and one misclassification as 25-minute samples), with a specificity of 100.00%. The 25-minute group achieved a sensitivity of 66.67% (four correct predictions, two misclassifications as 5-minute samples), with a specificity of 91.67%. This shows that the LDA model accurately identifies the 5m group, but there are some misjudgments in the 15m and 25m groups. Overall, it has some ability to distinguish samples with different processing times. In the VIP diagram of the partial least squares discriminant analysis (PLS-DA), the Raman shifts (such as 1101.0cm -1 、1093.2cm -1 The VIP values ​​of the Raman shifts (e.g., 1631.0 cm) are all well above 1 (red dashed line), indicating that these bands play a key role in distinguishing samples with different processing times in the model and are important spectral features reflecting sample differences. The regression coefficient plot of the partial least squares regression (PLSR) shows that the sign and magnitude of the coefficients of different Raman shifts reflect the direction and strength of their association with the predicted variables. Positive coefficients (e.g., 1631.0 cm) -1 、1632.7cm -1) indicates that the corresponding band signal enhancement is positively correlated with the predicted variable, and the negative coefficient (such as 1101.0cm -1 、1629.2cm -1 ) are negatively correlated. These bands with larger absolute values ​​of coefficients contribute significantly to the prediction, which helps to reveal the intrinsic connection between Raman spectroscopy and the characteristics of Zhucheng roast meat samples (such as composition changes related to processing time). Figure 8 ).

[0069] Taken together, these multivariate statistical analyses revealed the Raman spectral characteristics of Zhucheng barbecued pork at different processing times from different dimensions: PCA showed data distribution differences, LDA evaluated classification effects, and PLS-DA and PLSR determined key spectral variables. These analyses provided a spectroscopic basis for in-depth exploration of the changes in the composition of Zhucheng barbecued pork with processing time, and helped to analyze the relationship between its chemical properties and processing time from the molecular vibration level.

[0070] The principal component analysis (PCA) diagram shows the distribution of Raman spectral data of Taishan Fried Chicken under different conditions (24d, 36d, b1) in PC1 and PC2 dimensions. The points of different colors represent the samples of each group, and the ellipse represents the data distribution range. The samples of each group have both overlap and dispersion, reflecting that their chemical compositions have both similarities and differences in the principal component space. The results of linear discriminant analysis (LDA) showed that the overall accuracy of the model was 81.25%. The confusion matrix and sensitivity and specificity indicators showed that the sensitivity of the 24d group was 75.00% and the specificity was 91.67%; the sensitivity of the 36d group was 100.00% and the specificity was 90.91%; the sensitivity of the b1 group was 71.43% and the specificity was 88.89%, indicating that the LDA model has a certain ability to distinguish samples under different conditions, but there are some misjudgments in the b1 group. In the VIP diagram of partial least squares discriminant analysis (PLS-DA), the Raman shifts (such as 528.2cm -1 、523.9cm -1 The VIP values ​​of the bands (e.g., 528.2cm) are all much higher than 1, indicating that these bands are key variables for distinguishing samples under different conditions and contribute significantly to model discrimination. In the regression coefficient diagram of partial least squares regression (PLSR), the positive and negative coefficients and their magnitude reflect the direction and strength of the correlation between the corresponding Raman shift and the predicted variable, such as 528.2cm -1 The positive coefficient indicates that the signal enhancement is positively correlated with the predictor variable, 1183.5cm -1 Negative coefficients indicate negative correlation, and the bands with larger absolute values ​​of these coefficients play an important role in prediction ( Figure 9 ).

[0071] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent replacements, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for constructing an intelligent characteristic spectrum recognition system for typical Luwei meat dishes, characterized in that: The construction method comprises the following steps: (1) Process typical Luwei meat dishes and determine the optimal processing parameters; (2) detecting the sample processed in step (1) using near-infrared spectroscopy, Raman spectroscopy, or a combination of the two, preprocessing the collected near-infrared spectroscopy data and Raman spectroscopy data, extracting data analysis features, and establishing a classification and discrimination model based on near-infrared spectroscopy, Raman spectroscopy, or a combination of the two; (3) Evaluate the accuracy and stability of the model obtained in step (2), establish a digital fingerprint of typical Shandong-style meat dishes, and complete the construction of an intelligent characteristic spectrum recognition system for typical Shandong-style meat dishes.

2. The construction method according to claim 1, characterized in that In step (1), the typical Shandong-style meat dishes tested are Shandong fried chicken, Zhucheng roast pork and four-happiness meatballs.

3. The construction method according to claim 1, wherein In step (1), the process includes: selecting, cutting, marinating and cooking typical Shandong-flavored meat dishes test samples.

4. The construction method according to claim 1, characterized in that In step (2), the preprocessing includes: cleaning, smoothing, and normalizing the data.

5. A digital fingerprint of typical Luwei meat dishes established based on the construction method of claim 1.

6. Application of the digital fingerprint of typical Shandong-style meat dishes according to claim 5 in the following (1) or (2): (1) Achieve a comprehensive evaluation of the quality of Luwei meat dishes; (2) To classify and identify the quality of Shandong-style meat dishes; (3) Realize the digital display of Shandong-style meat dishes.

Images