A method and system for detecting the quality of a brewed foodstuff

Abstract

This invention relates to the field of food testing technology, and in particular to a method and system for quality testing of brewed foods. The method includes the following steps: obtaining a reference spectral template of the target food additive; performing liquid chromatography separation on the sample, and acquiring multi-wavelength spectral data of the eluted substances in real time during the separation process; calculating the spectral correlation between the multi-wavelength spectral data and the reference spectral template to obtain a spectral correlation value; generating a purified signal spectrum of the target food additive based on the spectral correlation value; and performing quantitative analysis on the target food additive signal in the purified signal spectrum. By introducing spectral correlation calculation and purified signal spectrum generation, this application can accurately extract the pure signal of the target food additive from complex chromatographic data.

Description

Technical Field

[0001] This invention relates to the field of food testing technology, and in particular to a method and system for quality testing of brewed foods. Background Technology

[0002] In the field of food quality testing, the accurate determination of key components in brewed foods (such as wine, soy sauce, and vinegar) is a core element in ensuring food safety, protecting consumer rights, and ensuring compliant product market entry. Due to the complexity of their raw materials and the variety of fermentation processes, brewed foods often contain large amounts of matrix components such as proteins, polysaccharides, and pigments, posing significant challenges to the extraction and detection of target analytes. In particular, the determination of food additive content, such as preservatives like sulfur dioxide and potassium sorbate, must be strictly controlled within regulatory limits. Exceeding these limits or making misjudgments can cause significant losses to production enterprises and pose a potential threat to public health.

[0003] Traditional methods such as high-performance liquid chromatography (HPLC) are reliable when processing samples with simple components. However, with the increasing complexity of winemaking processes and the diversification of matrices, components such as proteins, fats, polysaccharides, natural pigments, and spice extracts in these complex matrices are prone to overlapping with the signals of target additives during extraction and chromatographic separation, leading to peak identification interference. For example, in wine, high concentrations of phenolic compounds and sugars may co-elute with the chromatographic peaks of sulfur dioxide or potassium sorbate, making the peak shapes of these target components irregular or even completely masked.

[0004] This signal overlap caused by the complex matrix makes it impossible for traditional peak area calculation methods to accurately distinguish and quantify target components, which may lead to serious biases in the test results, such as false positives. A wine product that is normally qualified may be wrongly judged to have excessive sulfur dioxide or potassium sorbate, causing huge economic losses and reputational damage to the production company.

[0005] Faced with this matrix interference problem, the traditional solution is to improve the sample pretreatment steps, that is, to remove the interfering substances before injection into the chromatograph. For example, a "purification" step can be added, commonly using solid-phase extraction (SPE). However, this method has many drawbacks in practical applications. First, for each new and complex food matrix, purification conditions may need to be re-explored and optimized, a process that is inherently time-consuming and labor-intensive. Second, the purification operation significantly extends the total detection time for a single sample, which can severely impact detection efficiency for testing centers that need to process hundreds or thousands of samples daily. Furthermore, the additional use of consumables such as SPE columns directly increases the cost per test. Therefore, while this solution, which relies on complex pretreatment, can solve the problem, it falls short of the practical requirements of pursuing high efficiency and low cost.

[0006] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method and system for quality testing of brewed foods.

[0008] In a first aspect, the present invention provides a method for quality testing of brewed foods, applied to the detection of food additives, the method comprising the following steps: Obtain the reference spectral template of the target food additive; The sample was separated by liquid chromatography, and multi-wavelength spectral data of the eluted substances were acquired in real time during the separation process; The spectral correlation between the multi-wavelength spectral data and the reference spectral template is calculated to obtain the spectral correlation value. Based on the spectral correlation values, a purification signal spectrum of the target food additive is generated; Quantitative analysis was performed on the target food additive signal in the purified signal spectrum.

[0009] Secondly, a quality testing system for brewed foods is provided, the system comprising: The reference spectral template acquisition module is used to acquire the reference spectral template of the target food additive. The spectral data acquisition module is used for liquid chromatography separation of samples and to acquire multi-wavelength spectral data of the eluted substances in real time during the separation process. The spectral correlation calculation module is used to perform spectral correlation calculation between the multi-wavelength spectral data and the reference spectral template to obtain a spectral correlation value. A purification signal spectrum generation module is used to generate a purification signal spectrum of the target food additive based on the spectral correlation value. The quantitative analysis module is used to perform quantitative analysis on the target food additive signal in the purified signal spectrum.

[0010] Compared with the prior art, the present invention has the following beneficial effects: By introducing spectral correlation calculation and purified signal spectrum generation, this application can accurately extract the pure signal of the target food additive from complex chromatographic data. Even in the presence of severe co-elution interference, it can achieve accurate identification and quantification of the target, avoiding the "false positive" misjudgment caused by peak overlap in traditional methods. Attached Figure Description

[0011] Figure 1 This is a flowchart of the method of the present invention.

[0012] Figure 2This is a schematic diagram of the system structure of the present invention.

[0013] In the diagram: 201, Reference Spectrum Template Acquisition Module; 202, Spectrum Data Acquisition Module; 203, Spectral Correlation Calculation Module; 204, Purified Signal Spectrum Generation Module; 205, Quantitative Analysis Module. Detailed Implementation

[0014] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0015] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0016] To better understand this application, the key terms used herein are explained.

[0017] "Reference spectral template" refers to the pre-acquired and stored standard spectral data of the target food additive under specific conditions. It contains information on the absorbance or emission intensity of the target substance at different wavelengths, serving as the basis for identifying and quantifying the target substance. For wine, this might refer to the standard absorption spectrum of high-purity sulfur dioxide or potassium sorbate under specific pH and solvent conditions. "Liquid chromatography separation" is a technique that separates different components in a sample based on their differences in physicochemical properties. Through the action of the chromatographic column and mobile phase, each component in the sample elutes at different time points. In wine analysis, liquid chromatography can effectively separate matrix components such as phenols, sugars, and organic acids, creating conditions for the detection of target additives. "Multi-wavelength spectral data" refers to the spectral information of the eluent substance acquired in real time at multiple wavelengths during liquid chromatography separation. This data records the spectral characteristics of the substance in the form of a time series. For example, a DAD detector can continuously acquire spectral data of the eluent in the range of 200-800 nm. "Spectral correlation calculation" refers to using mathematical algorithms to evaluate the similarity between multi-wavelength spectral data and a reference spectral template, usually expressed as a correlation coefficient or similarity index, used to determine whether the eluted substance is the target food additive. "Purified signal chromatogram" refers to a chromatogram extracted and reconstructed from the original multi-wavelength spectral data based on the spectral correlation value, containing only the signal of the target food additive, effectively removing matrix interference signals. "Quantitative analysis" refers to integrating or otherwise mathematically processing the target food additive signal in the purified signal chromatogram to determine the content or concentration of the target food additive.

[0018] like Figure 1 The method shown is a quality testing method for brewed foods, applied to the detection of food additives. The method includes the following steps: S101. Obtain the reference spectral template of the target food additive; It should be noted that this reference spectral template can be obtained in several ways. For example, a high-purity standard of the target food additive (such as a high-purity sulfur dioxide aqueous solution or potassium sorbate crystals) can be prepared in advance and scanned using a spectrometer under specific conditions to obtain its characteristic spectral data at different wavelengths, which can then be stored as a reference spectral template. Alternatively, the reference spectral data of the target food additive can be directly obtained from an existing standard database.

[0019] S102. Perform liquid chromatography separation on the sample and collect multi-wavelength spectral data of the eluting substances in real time during the separation process; It should be noted that, in practice, the food sample to be tested (e.g., wine) can be injected into the liquid chromatograph after simple pretreatment (e.g., dissolution, filtration). The liquid chromatograph separates the different components in the sample. Simultaneously, a multi-wavelength detector (e.g., diode array detector, DAD) is configured at the outlet of the chromatographic column to acquire spectral data of the eluted substances at multiple wavelengths in real time. These data are recorded in time series form, forming a three-dimensional data matrix containing time, wavelength, and absorbance information.

[0020] S103. Calculate the spectral correlation between the multi-wavelength spectral data and the reference spectral template to obtain the spectral correlation value; It should be noted that, in practice, the multi-wavelength spectral data (i.e., a spectral snapshot) collected at each time point during the liquid chromatography separation process can be compared with a pre-acquired reference spectral template. This comparison can be achieved through various mathematical methods such as calculating spectral similarity, correlation coefficient, or Euclidean distance. For example, the Pearson correlation coefficient between each spectral snapshot and the reference spectral template can be calculated. This coefficient typically ranges from -1 to 1; the closer the value is to 1, the more similar the spectral shapes are, thus indicating the degree of similarity between the eluted substance at that time point and the target food additive. When sulfur dioxide is eluted from a wine sample, the Pearson correlation coefficient between its spectral snapshot and the sulfur dioxide reference spectral template will significantly increase.

[0021] S104. Based on the spectral correlation values, generate the purification signal spectrum of the target food additive; It should be noted that after obtaining the spectral correlation values ​​at each time point, these values ​​can be used to construct a purified signal spectrum. For example, the spectral correlation value at each time point can be used as the signal intensity in the purified signal spectrum at that time point. When the spectral correlation value is high, it indicates that the eluting substance at that time point is highly similar to the target food additive, and its corresponding signal intensity will be high in the purified signal spectrum; conversely, when the spectral correlation value is low, it indicates that the eluting substance at that time point is dissimilar to the target food additive, and its corresponding signal intensity will be low in the purified signal spectrum, or even close to zero. In this way, the signal of the target food additive can be effectively "purified" from complex matrix interference, forming a clear chromatogram that only reflects the peak of the target substance.

[0022] S105. Quantitative analysis of the target food additive signal in the purified signal spectrum.

[0023] It should be noted that after generating the purified signal chromatogram, conventional chromatographic quantitative methods can be used for analysis. For example, the peak region of the target food additive signal in the purified signal chromatogram can be identified, and the quantitative result can be obtained by integrating the peak area or measuring the peak height. Since the purified signal chromatogram has effectively removed matrix interference, the obtained quantitative results will be more accurate and reliable.

[0024] The proposed method for quality detection of brewed foods effectively solves the problems of signal overlap and inaccurate quantification encountered by traditional methods in detecting food additives in complex matrices by introducing spectral correlation calculation and purified signal spectrum generation. Traditional methods often rely on sample pretreatment to remove interfering substances, which is not only time-consuming and labor-intensive but also increases detection costs and operational complexity. In contrast, the method in this application directly utilizes multi-wavelength spectral data for in-depth analysis after liquid chromatography separation. Through spectral feature matching and purification, it achieves accurate identification and quantification of target food additives.

[0025] Specifically, the core innovation of this application lies in its approach: instead of relying solely on chromatographic retention time or signal intensity at a single wavelength to determine and quantify target substances, it fully utilizes the rich information contained in multi-wavelength spectral data. By calculating the spectral correlation between real-time acquired multi-wavelength spectral data and a preset reference spectral template, it can effectively distinguish target food additives from matrix interferences, even if their retention times on the chromatographic column are similar or overlap. The resulting purified signal spectrum directly reflects the spectral purity of the target substance, thereby significantly improving the accuracy of quantitative analysis. This method avoids cumbersome sample pretreatment steps, significantly shortens the detection cycle, reduces detection costs, and improves the reliability of detection results, providing an efficient, accurate, and economical solution for the field of food quality testing.

[0026] As one embodiment of the present invention, the step of calculating the spectral correlation between multi-wavelength spectral data and a reference spectral template includes: Identify suspected signal peaks of target food additives in multi-wavelength spectral data; It should be noted that this specifically refers to locating chromatographic peak regions that may contain the target food additive from multi-wavelength spectral data acquired in real-time during liquid chromatography separation, through preliminary screening or preset conditions. For example, a preliminary judgment can be made based on the known retention time window or absorption characteristics at specific wavelengths of the target food additive to narrow down the scope of subsequent detailed analysis. The aim is to focus the analysis on potential positive regions and improve processing efficiency. When detecting sulfur dioxide in wine, preliminary identification can be made based on its specific absorption peaks in the ultraviolet region and its typical retention time range.

[0027] Analyze the spectral characteristics within the suspected signal peak to determine the region where the spectral characteristics of the target object dominate within the suspected signal peak; It should be noted that this can be understood as conducting in-depth spectral purity analysis on identified suspected signal peaks. Specifically, this involves a detailed examination of spectral snapshots taken at different time points within the peak, assessing the contribution of the target food additive's spectral characteristics in each snapshot. By comparing the similarity between the spectra at different time points and the reference spectra of the pure substance, or analyzing characteristics such as the shape and peak position of the spectra, the time period or region within the suspected signal peak where the target food additive's spectral characteristics are the purest and least affected by interference can be determined. The aim is to find the region that best represents the true spectral characteristics of the target substance, providing high-quality input for the subsequent construction of dynamic reference spectra.

[0028] Based on the region where the spectral characteristics of the target object are dominant, the signal-to-noise ratio and spectral purity of each spectral snapshot in the region are quantified. The weight of each spectral snapshot is determined based on the signal-to-noise ratio and spectral purity. Then, the reference spectral template and the spectral data in the region are weighted and averaged according to the weight to construct a dynamic reference spectrum that adapts to the current matrix environment. It should be noted that this specifically refers to adjusting and optimizing the original reference spectral template using the spectral information of the region where the spectral characteristics of the target analyte dominate within the suspected signal peak. For example, the average spectrum or the most representative spectrum of this dominant region can be extracted as a benchmark, and combined with the characteristics of the original reference spectral template, a new reference spectrum that more closely resembles the actual spectral performance of the target analyte in the current sample matrix can be generated through algorithms (such as weighted averaging, spectral correction, or feature fusion).

[0029] Specifically, when constructing a dynamic reference spectrum, it is first necessary to quantify the signal-to-noise ratio (SNR) and spectral purity of each spectral snapshot within the region where the target analyte's spectral characteristics dominate. The SNR is quantified as follows: calculate the signal peak height H_signal of the spectral snapshot in the target wavelength band (i.e., the wavelength band where the main absorption peak of the target food additive is located), and the root mean square noise RMS_noise of the baseline region on both sides of the suspected signal peak of the spectral snapshot (the time period without a clear target analyte signal). Then, SNR = H_signal / RMS_noise. When SNR > 3, it indicates that the target analyte signal in the spectral snapshot is significantly higher than the noise level, making it suitable for constructing a dynamic reference spectrum.

[0030] Spectral purity P is evaluated as follows: the ratio of absorbance A_target at the target characteristic absorption peak to absorbance A_adjacent at the adjacent non-characteristic absorption peak in the spectral snapshot is calculated, i.e., P = A_target / A_adjacent. A larger P value indicates that the spectral characteristics of the target food additive are more dominant, and the spectrum at that time point is purer. When P > 1.5, the spectral snapshot is considered to have sufficient spectral purity (the typical threshold for P is 1.5-3.0, determined based on the spectral characteristics of the target food additive). The consecutive time periods containing spectral snapshots that simultaneously satisfy SNR > 3 and P > 1.5 are identified as the regions where the spectral characteristics of the target substance are dominant.

[0031] When constructing a dynamic reference spectrum using a weighted averaging method, the weights are determined based on the signal-to-noise ratio (SNR) and spectral purity. For the i-th spectral snapshot within the region where the target analyte's spectral characteristics dominate, its weight w_i = α·SNR_norm_i + β·P_norm_i, where SNR_norm_i and P_norm_i are the normalized SNR and spectral purity values ​​(linearly normalizing the SNR and P values ​​of each spectral snapshot to the [0,1] interval), and α and β are weighting coefficients with α + β = 1. In typical applications, α is typically 0.5–0.7, and β is 0.3–0.5. The dynamic reference spectrum S_dynamic = Σ(w_i·S_i) / Σw_i, where S_i is the spectral vector of the i-th spectral snapshot. Through this weighted averaging method, spectral snapshots with higher SNR and better spectral purity are given greater weight during fusion, thus enabling the dynamic reference spectrum to more accurately reflect the true spectral characteristics of the target food additive under the current matrix environment. The purpose is to ensure that the reference spectrum used for correlation calculations can accurately reflect the true spectral characteristics of the target food additive under specific sample conditions, thereby improving the accuracy of matching.

[0032] The spectral correlation between multi-wavelength spectral data and dynamic reference spectrum is calculated to obtain the spectral correlation value.

[0033] It should be noted that this specifically refers to using the constructed dynamic reference spectrum to evaluate the spectral similarity of the entire multi-wavelength spectral data (or a specific region). This calculation can employ various spectral similarity algorithms, such as Pearson correlation coefficient, cosine similarity, or Euclidean distance, to quantify the degree of matching between the spectrum at each time point and the dynamic reference spectrum. The aim is to obtain a precise numerical indicator reflecting the presence and purity of the target food additive, providing a reliable basis for subsequent purification signal spectrum generation.

[0034] The proposed method first identifies potential signal peaks of the target food additive, focusing the analysis on regions where the target substance may be present. Subsequently, by analyzing the spectral characteristics within these potential signal peaks, the dominant region of the target substance's spectral characteristics can be precisely determined, effectively avoiding the potential influence of co-eluting substances or matrix interference on the target substance's spectrum. Based on the spectral information of this dominant region, combined with the original reference spectral template, a dynamic reference spectrum adapted to the current matrix environment is constructed. This dynamic adjustment mechanism allows the reference spectrum used for spectral correlation calculations to more accurately reflect the true spectral characteristics of the target food additive in the actual sample, significantly improving the accuracy and specificity of spectral correlation calculations and effectively addressing the limitations of fixed reference spectra in complex matrices.

[0035] Through the above technical solutions, this application can significantly improve the accuracy and reliability of food quality testing. By constructing a dynamic reference spectrum adapted to the current matrix environment, the interference of sample matrix complexity or co-eluting substances on spectral analysis is effectively overcome, making the spectral correlation calculation results more accurate. This reduces the incidence of false positives or false negatives, improves the confidence level of target food additive identification and quantitative analysis, and is particularly suitable for complex samples or trace detection scenarios.

[0036] In some preferred embodiments, a specific example is given below. Suppose that sulfur dioxide needs to be detected in a wine sample. During liquid chromatography separation, multi-wavelength spectral data of the effluent are acquired in real time. First, the system identifies a chromatographic peak in the multi-wavelength spectral data that is suspected to contain sulfur dioxide, based on the known retention time window and preliminary spectral characteristics of sulfur dioxide. Next, a detailed analysis is performed on spectral snapshots at different time points within this suspected signal peak. For example, by calculating the similarity of each snapshot to a pure sulfur dioxide reference spectrum, or analyzing the shape and position of its characteristic absorption peak (e.g., 200 nm), the region within the peak where the sulfur dioxide spectral characteristics are purest and least disturbed is determined. For example, the spectral purity may be highest near the peak apex, and this region is identified as the region where the target analyte's spectral characteristics dominate. Subsequently, using the actual spectral information of this dominant region, combined with the original sulfur dioxide reference spectral template, a dynamic sulfur dioxide reference spectrum adapted to the current wine matrix environment is constructed. For example, if a slight shift in the actual peak position is detected, the dynamic reference spectrum will be adjusted accordingly. Finally, the spectral correlation of the entire multi-wavelength spectral data with this dynamically constructed sulfur dioxide reference spectrum is calculated to obtain the spectral correlation value at each time point. In this way, sulfur dioxide can be more accurately identified and quantified even in complex matrices, avoiding the spectral mismatch problem caused by matrix effects.

[0037] As one embodiment of the present invention, the step of analyzing the spectral characteristics within a suspected signal peak and determining the region where the spectral characteristics of the target analyte dominate within the suspected signal peak includes: Spectral feature analysis was performed on the spectral snapshots inside the suspected signal peaks to obtain the spectral features of each snapshot; It should be noted that spectral feature analysis of spectral snapshots within suspected signal peaks refers to treating the multi-wavelength spectral data acquired in real-time at each time point within the time period corresponding to the suspected signal peak as an independent spectral snapshot during liquid chromatography separation. Feature analysis of these snapshots may include, but is not limited to, extracting parameters such as the position, height, area, and half-maximum width of the main absorption peaks, or performing multivariate statistical methods such as spectral deconvolution and principal component analysis to identify and quantify the chemical component information contained in each spectral snapshot. The aim is to obtain detailed spectral fingerprints of substances at each time point, providing fundamental data for subsequent regional determination.

[0038] Based on the distribution characteristics of the spectral features of the spectral snapshots, the region where the target analyte's spectral features dominate is determined. This region is defined as a continuous time period with a signal-to-noise ratio (SNR) greater than 3 and a spectral purity greater than 1.5. The SNR is the ratio of the signal peak height in the target wavelength range to the root mean square (RMS) of the noise in the baseline region. Spectral purity is the ratio of the absorbance at the target characteristic absorption peak to the absorbance at adjacent non-characteristic absorption peaks in the spectral snapshot. Specifically, this refers to analyzing the trends and distribution patterns of these features over the entire time axis of the suspected signal peak after acquiring the spectral features of each spectral snapshot within the suspected signal peak. For example, by comparing the similarity, purity value, or relative intensity changes of specific absorption peaks of the spectral features of each spectral snapshot with the reference spectral template of the target food additive, the region with the purest characteristic spectral signal and the least interference from other co-eluted substances can be identified. This region typically represents a continuous time period where the target analyte's characteristic spectral signal intensity is highest and its spectral shape is closest to the reference spectral template. The purpose is to accurately pinpoint the actual elution region of the target food additive in the suspected signal peak, eliminate impurities or matrix interference, and ensure the accuracy of subsequent analysis.

[0039] This application's approach, through detailed spectral feature analysis of each spectral snapshot within the suspected signal peak, obtains spectral information of the target food additive and its potential co-eluting impurities at different time points during chromatographic elution. It is precisely this in-depth analysis of the spectral features of these snapshots, and further examination of their distribution characteristics throughout the suspected signal peak, that enables the system to identify the region with the purest and least disturbed characteristic spectral signal of the target food additive. This method effectively distinguishes the spectral contributions of the target analyte and co-eluting impurities, thereby more accurately determining the region where the target analyte's spectral features dominate, laying a solid foundation for subsequently constructing a dynamic reference spectrum adapted to the current matrix environment.

[0040] By employing the aforementioned technical solution, this application overcomes the problem of traditional methods where, in complex sample matrices, the signals of target food additives and impurities overlap due to co-elution, making it difficult to accurately identify the pure spectral region of the target analyte. By meticulously analyzing the spectral characteristics of the spectral snapshot and examining its distribution properties, the identification accuracy and reliability of the dominant spectral region of the target analyte can be effectively improved. This ensures the accuracy of subsequent spectral correlation calculations and the generation of purified signal maps, significantly enhancing the specificity of food quality testing and the accuracy of quantitative analysis.

[0041] As one embodiment of the present invention, the step of quantitatively analyzing the target food additive signal in the purified signal spectrum includes: Identifying peak regions of the target food additive signal in the purified signal spectrum; specifically, using conventional peak detection algorithms, such as thresholding, derivative methods, or wavelet transform, to initially locate the range where the target food additive signal may exist. This peak region is typically a continuous time period where the signal intensity is significantly higher than the background noise.

[0042] Analyzing the signal fluctuation characteristics of the baseline regions on both sides of the peak region yields the baseline mean and baseline noise standard deviation. This can be understood as performing statistical analysis on the signals before and after the peak region to obtain information such as the average baseline level, noise level, and whether there is a trend drift. For example, the root mean square noise (RMS noise) or standard deviation of the baseline region can be calculated to quantify its fluctuation degree.

[0043] Based on the signal fluctuation characteristics and the signal change rate in the peak region, the starting and ending points of the target food additive signal are determined. The starting point is defined as when the signal intensity is above the sum of the baseline mean and k times the baseline noise standard deviation for N consecutive data points, and the first derivative of the signal reaches a slope threshold. The ending point is defined as when the signal intensity falls below the sum of the baseline mean and k times the baseline noise standard deviation for N consecutive data points, and the absolute value of the first derivative of the signal is below the slope threshold. Here, N ≥ 3, k is 3 to 5, and the slope threshold is m times the standard deviation of the first derivative of the signal in the baseline region, where m is 2 to 4. This aims to accurately define the boundary of the integration. The signal change rate can be obtained by calculating the first or second derivative of the signal. The change rate increases significantly when the signal rises or falls from the baseline. Specifically, the signal statistical characteristics of the baseline regions on both sides of the peak region are first calculated to obtain the baseline mean μ_baseline and the baseline noise standard deviation σ_baseline. The baseline region is selected within a time period of 5-20 data points before and after the peak region. For determining the starting point: When the signal strength of N consecutive data points (N≥3) is higher than the sum of the baseline mean and k times the standard deviation of the baseline noise (i.e., signal strength > μ_baseline + k·σ_baseline), and the first derivative of the signal (i.e., the difference in signal strength between adjacent data points) reaches the slope threshold T_slope, the first data point of this continuous segment is determined as the starting point. Here, k is the noise judgment factor, ranging from 3 to 5. A k value of 3 indicates higher sensitivity, suitable for weak peak detection, while a k value of 5 indicates higher specificity, suitable for strong matrix interference scenarios. The slope threshold T_slope = m·σ_slope, where σ_slope is the standard deviation of the first derivative of the signal in the baseline region, and m is the slope factor coefficient, ranging from 2 to 4. For determining the ending point: When the signal strength falls below μ_baseline + k·σ_baseline, and the absolute value of the first derivative of the signal of N consecutive data points (N≥3) is lower than T_slope, the last data point of this continuous segment is determined as the ending point. For example, in a typical application scenario for detecting sulfur dioxide in wine, when k=3 and m=2, the starting point determination condition is that the signal strength of more than three consecutive data points is greater than μ_baseline + 3σ_baseline and the slope is greater than 2σ_slope; when k=5 and m=4, the determination condition is more stringent, which can effectively eliminate small fluctuations caused by co-elution impurities. Through the above clearly defined parameter determination rules, different technicians can obtain consistent detection results by following the same rules.

[0044] The signal of the target food additive between the starting and ending points is integrated to obtain a quantitative result. Specifically, this involves using numerical integration methods, such as the trapezoidal method or Simpson's method, to calculate the area or height under the precisely defined peak region, thereby obtaining quantitative information about the target food additive.

[0045] This application's method first identifies the peak region of the target food additive signal, providing a preliminary range for subsequent precise quantitative analysis. Then, by analyzing the signal fluctuation characteristics of the baseline regions on both sides of this peak region, the background noise and baseline drift of the current local environment can be accurately assessed, avoiding the errors caused by using a globally fixed baseline. Simultaneously, by combining the signal change rate of the peak region, the precise turning point where the target signal begins to separate from the baseline and eventually returns to it can be sensitively captured. It is precisely because of the comprehensive consideration of the dynamic characteristics of the baseline and the morphological changes of the signal itself that the starting and ending points of the target food additive signal can be more accurately determined, thereby ensuring the accuracy of subsequent integration operations and effectively solving the problem of inaccurate quantification caused by unstable baselines or complex peak shapes in traditional methods.

[0046] Through the above technical solutions, this application can significantly improve the accuracy and reliability of quantitative analysis of food additives. By analyzing the baseline signal fluctuation characteristics on both sides of the peak region, it can effectively address the complex baseline drift and noise interference in actual samples, avoiding errors in determining the integration start and end points due to baseline instability. Simultaneously, by combining the signal change rate to determine the integration boundary, even in cases of irregular peak shapes or the presence of shoulder peaks, the true signal range of the target food additive can be accurately defined, thereby obtaining more precise quantitative results and providing more reliable data support for food quality testing.

[0047] As one embodiment of the present invention, the step of identifying suspected signal peaks of a target food additive in multi-wavelength spectral data includes: The similarity between the spectral snapshot at each time point in the multi-wavelength spectral data and the reference spectral template is calculated to obtain the similarity at each time point; Based on the similarity at each time point, regions whose similarity at multiple consecutive time points is higher than a preset recognition threshold are identified; Analyze multi-wavelength spectral data within the region to determine the local noise level and signal strength. Based on the local noise level and the signal strength in the region, the starting and ending points where the signal strength is higher than the local noise level are identified; The region between the start and end points is identified as the suspected signal peak of the target food additive.

[0048] Specifically, when calculating the similarity between the spectral snapshots at each time point in multi-wavelength spectral data and the reference spectral template, the similarity can be evaluated using various spectral similarity algorithms, such as cosine similarity, Pearson correlation coefficient, or Euclidean distance. By calculating the similarity between the spectral snapshots at each time point and the reference spectral template of the known target food additive, it is possible to preliminarily determine at which time points the signal of the target food additive may be present. Specifically, when identifying regions where the similarity at multiple consecutive time points is higher than a preset identification threshold, the specific method for setting the preset identification threshold is as follows: First, calculate the statistical distribution of the similarity at all time points throughout the entire chromatographic analysis process to determine the background similarity level, i.e., the mean similarity μ_bg and standard deviation σ_bg of the regions without the target substance; the preset identification threshold T_id is set as T_id = μ_bg + k_id·σ_bg, where k_id is the signal-to-noise ratio coefficient, with a value ranging from 3 to 5. When k_id is 3, the detection sensitivity is high, suitable for screening trace targets; when k_id is 5, the specificity is high, reducing false positives. Furthermore, when using the Pearson correlation coefficient as a similarity index, the preset identification threshold can be directly set to 0.70-0.90, preferably 0.85. The more complex the matrix, the lower the threshold can be (e.g., to 0.75) to avoid missed detections; when the matrix is ​​simpler, the threshold can be increased to 0.90 to enhance specificity. The purpose is to screen out preliminary regions with a high probability of containing the target food additive signal. This threshold should be high enough to exclude most background noise, but not too high to avoid missing true signals. In practical applications, when analyzing multi-wavelength spectral data within an area to determine the local noise level and signal intensity, the local noise level can be determined by statistically analyzing the baseline fluctuations within or near the region, such as calculating the standard deviation. Signal intensity can be measured based on the height or area of ​​the spectral peak. The purpose is to distinguish true signals from background interference, providing a basis for subsequent accurate peak identification. Further, based on the local noise level and the signal intensity of the region, the starting and ending points where the signal intensity is higher than the local noise level are identified. This means that only when the signal intensity is significantly higher than the background noise is it considered a valid signal. Determining the start and end points aims to precisely define the boundaries of the target food additive signal, ensuring the accuracy of subsequent analyses. Therefore, the region between the start and end points is identified as a potential signal peak for the target food additive. This potential signal peak is a preliminarily identified chromatographic peak region that may contain the target food additive, laying the foundation for subsequent purity assessment and quantitative analysis.

[0049] The proposed solution first calculates the similarity between the spectral snapshots at each time point and a reference spectral template, thus initially screening out regions similar to the spectral characteristics of the target food additive. Subsequently, by setting a preset identification threshold and identifying regions with consecutive similarities exceeding the threshold, transient noise interference can be effectively eliminated, focusing on continuous time periods that may contain the target substance's signal. Furthermore, by analyzing the local noise level and signal intensity within these regions, and using this as a basis to determine the starting and ending points where the signal intensity exceeds the noise level, the proposed solution can accurately define the boundaries of suspected signal peaks of the target food additive. This step-by-step, refined identification method makes the identification process of suspected signal peaks more robust and accurate, avoiding the misidentification of noise or baseline fluctuations as target signals.

[0050] As one embodiment of the present invention, the step of identifying the starting point and the ending point where the signal strength is higher than the local noise level includes: The purity of the target food additive was evaluated by analyzing the multi-wavelength spectral data at each time point within the suspected signal peak, and the purity value at each time point was obtained. Analyze the changes in purity values ​​over time to identify the boundaries of regions where purity values ​​transition from low to high levels, as well as the boundaries of regions where purity values ​​transition from high to low levels. By combining the regional boundaries of purity value transitions and the relative relationship between signal strength and local noise levels, the starting and ending points of the target food additive signal are determined.

[0051] Specifically, the purity of the target food additive is evaluated using multi-wavelength spectral data at each time point within the suspected signal peak. This aims to quantify the degree of matching between the spectral data at that time point and the pure substance spectrum of the target food additive. The purity value can be understood as spectral purity, reflecting the relative content or the purity of the spectral characteristics of the target food additive in the eluted substance at a specific time point. This evaluation process can be implemented using various spectral purity algorithms, such as calculating by comparing the similarity, correlation, or residuals between the spectrum at the current time point and the reference spectral template. Analyzing the change in purity value over time aims to identify the purity change trend of the target food additive signal during the chromatographic peak elution process. Ideally, the purity value of the target food additive is lower near the start and end points of the peak, and reaches a higher level in the central region of the peak. By identifying the boundary of the region where the purity value transitions from low to high levels, the starting point of significant elution of the target food additive can be determined; by identifying the boundary of the region where the purity value transitions from high to low levels, the ending point of elution can be determined. These transition boundaries typically correspond to inflection points where significant changes in purity occur at the rising and falling edges of the chromatographic peaks. In practical applications, combining the regional boundaries of purity value transitions with the relative relationship between signal intensity and local noise levels allows for more precise determination of the start and end points of the target food additive signal. For example, the start point can be identified when the purity value begins to rise significantly while the signal intensity is above the local noise level; conversely, the end point can be identified when the purity value begins to fall significantly while the signal intensity drops to near the local noise level. This combined approach effectively avoids misinterpreting signals caused by co-eluting impurities or baseline fluctuations as part of the target food additive signal, thus ensuring that the determined peak boundaries more accurately reflect the true elution range of the target food additive.

[0052] This application's approach, by introducing purity assessment of multi-wavelength spectral data at each time point within a suspected signal peak and analyzing the change in purity values ​​over time, can more precisely capture the true purity information of the target food additive during chromatographic elution. Traditional methods relying solely on signal intensity may fail to accurately distinguish the boundary between the target signal and background noise or impurity signals when faced with complex sample matrices or co-elution interference. Purity assessment effectively identifies the pure regions of the target food additive and precisely locates the start and end points of the target signal based on the trend of purity value changes. This method combines spectral purity information with chromatographic signal intensity information, ensuring that the determined peak boundaries consider not only the presence of the signal but also its "purity," thus overcoming the boundary ambiguity or misjudgment problems that may arise from relying solely on signal intensity.

[0053] As one embodiment of the present invention, the step of evaluating the purity of the target food additive by analyzing the multi-wavelength spectral data at each time point within the suspected signal peak to obtain the purity value at each time point includes: Spectral feature analysis was performed on the multi-wavelength spectral data at each time point within the suspected signal peak to identify the systematic spectral shift characteristics with the reference spectral template; It should be noted that spectral feature analysis of multi-wavelength spectral data at each time point within the suspected signal peak aims to gain a deeper understanding of the spectral characteristics of the effluxed substance at that moment, in order to identify systematic spectral shifts from the reference spectral template. These systematic shifts may manifest as slight drifts in absorption peak positions, changes in peak shape, or baseline drifts. These shifts are typically caused by factors such as sample matrix effects, changes in instrument status, or fluctuations in ambient temperature. Identifying these shift characteristics is crucial to ensuring the accuracy of subsequent purity assessments. Systematic spectral shifts can be understood as regular differences between the sample spectrum and the reference spectrum across the entire spectral range or within specific key absorption peak regions, rather than random noise.

[0054] Based on the systematic spectral shift characteristics, the reference spectral template is adaptively adjusted to obtain an adaptive reference spectrum; It should be noted that the adaptive adjustment is performed in the following order: baseline correction → spectral shift → scaling → peak fitting. The specific implementation of each step is as follows: (1) Baseline correction: Baseline correction is performed on the reference spectral template and the current spectral snapshot respectively, and the baseline is fitted and subtracted using polynomials of order 1 to 3. The order of the polynomial is determined according to the complexity of the spectral baseline: when the baseline is close to horizontal, first-order linear fitting is used; when the baseline has a slow drift, second-order quadratic fitting is used; and when the baseline has a significant curvature, third-order cubic fitting is used. The purpose of prioritizing baseline correction is to eliminate the influence of background drift on the subsequent peak position and peak shape judgment. (2) Spectral shift: The offset Δλ between the current spectral snapshot and the reference spectral template is calculated by cross-correlation analysis. Specifically, cross-correlation is performed on the two baseline-corrected spectra in the target wavelength range, and the wavelength offset corresponding to the maximum value of the cross-correlation function is Δλ. The reference spectral template is shifted Δλ along the wavelength axis to align the absorption peak position. (3) Scaling: Calculate the signal amplitude ratio R = A_sample / A_ref in the target wavelength range between the translated and aligned reference spectral template and the current spectral snapshot, where A_sample is the peak area of ​​the current spectral snapshot in the target wavelength range and A_ref is the peak area of ​​the reference spectral template in the corresponding wavelength range. Multiply the ordinate of the reference spectral template by R to normalize the amplitude. (4) Peak shape fitting: Perform Gaussian-Lorentz mixed peak shape fitting on the main absorption peaks of the reference spectral template after the above correction, and adjust the peak width and symmetry parameters to make them consistent with the peak shape of the corresponding absorption peak in the current spectral snapshot. The proportion of the Gaussian component in the mixing function is usually 0.5-0.8, which is adjusted according to the actual spectral peak shape. The purpose is to enable the adjusted reference spectrum to more accurately reflect the true spectral characteristics of the target food additive in the current complex sample matrix. Through the above correction order and parameter determination method, the evaluation error caused by systematic shift can be effectively eliminated or reduced.

[0055] The purity of the multi-wavelength spectral data at each time point was evaluated by comparing it with the adaptive reference spectrum, and the purity value at each time point was obtained.

[0056] It should be noted that purity assessment can employ various spectral similarity algorithms, such as spectral angle, correlation coefficient, or Euclidean distance, with the aim of quantifying the degree of matching between the current time-point spectrum and the adaptive reference spectrum. By using an adaptive reference spectrum for assessment, the true purity of the target food additive can be reflected more accurately, thereby improving the reliability of purity assessment.

[0057] This application's solution effectively addresses the issue of decreased accuracy in traditional purity assessments due to spectral shifts by introducing the identification of systematic spectral shift characteristics and adaptive adjustment of the reference spectral template. Specifically, by analyzing the spectral features of spectral snapshots within suspected signal peaks, subtle spectral changes in the target food additive under actual testing conditions can be captured and quantified as systematic spectral shift characteristics. It is precisely because these shifts are identified that the reference spectral template can be precisely adjusted, generating an adaptive reference spectrum highly matched to the current sample matrix and instrument conditions. Subsequently, purity assessment based on this adaptive reference spectrum can minimize or reduce the impact of matrix interference and instrument drift on the purity assessment results, ensuring the accuracy and reliability of the purity values. This adaptive assessment mechanism allows purity assessment to dynamically adapt to actual testing conditions, rather than relying on a fixed reference spectrum, thus significantly improving the accuracy of target food additive signal identification.

[0058] Through the above technical solutions, this application can significantly improve the accuracy and robustness of purity assessment for target food additives. Especially when facing complex sample matrices or fluctuations in instrument conditions, dynamically adjusting the reference spectral template can effectively compensate for the effects of systematic spectral shifts, thereby obtaining more accurate purity values. This not only helps to more accurately identify the start and end points of the target food additive signal in the purification signal spectrum, avoiding misjudgments or omissions caused by spectral shifts, but also provides more reliable basic data for subsequent quantitative analysis, ultimately improving the overall accuracy and reliability of the quality testing methods for brewed foods.

[0059] As one embodiment of the present invention, the step of identifying systematic spectral shift characteristics relative to the reference spectral template includes: By comparing the position or shape of the main absorption peaks in the multi-wavelength spectral data at each time point with the position or shape of the corresponding absorption peaks in the reference spectral template, the geometric difference is obtained. Based on geometric differences, systematic spectral shift characteristics were identified.

[0060] The primary absorption peak refers to the peak of the target food additive that exhibits significant absorption characteristics within a specific wavelength range; these peaks are typically a key component of its spectral fingerprint. Position can be understood as the peak wavelength of the absorption peak, while shape can include characteristics such as peak width, symmetry, and shoulders. Geometric difference refers to the deviation quantified by comparing these spectral characteristic parameters. For example, geometric difference can be obtained by calculating the shift in peak wavelength, the rate of change of the full width at half maximum (FWHM), or the differences in peak shape fitting parameters.

[0061] This application's method, by directly comparing the geometric characteristics of the main absorption peaks in multi-wavelength spectral data with those in a reference spectral template, can accurately capture systematic spectral shifts caused by matrix effects, instrument drift, or other environmental factors. Traditional methods may rely solely on overall spectral similarity, making it difficult to distinguish signals caused by changes in target purity from those caused by systematic shifts. By quantifying the geometric differences in the position or shape of absorption peaks, these systematic changes caused by non-target substances can be identified, thus providing an accurate basis for subsequent adaptive adjustments to the reference spectral template. This allows purity assessments to more accurately reflect the true purity of the target food additive, rather than being affected by external interference.

[0062] As one embodiment of the present invention, the step of comparing the position or shape of the main absorption peak in the multi-wavelength spectral data at each time point with the position or shape of the corresponding absorption peak in the reference spectral template to obtain the geometric difference includes: Extract the characteristic parameters of the main absorption peaks from the multi-wavelength spectral data at each time point; It should be noted that the characteristic parameters of the main absorption peak may include, but are not limited to, peak position (i.e., the wavelength position of the absorption peak), peak height (i.e., the intensity of the absorption peak), peak width (e.g., the full width at half maximum or the peak width at half maximum), peak area, and peak shape (e.g., peak symmetry, the presence of shoulders, etc.). These characteristic parameters can be obtained by performing spectral analysis, curve fitting, or mathematical transformations on multi-wavelength spectral data. For example, peak-finding algorithms can be used to identify the center wavelength and maximum absorbance value of the absorption peak as the peak position and peak height; the peak width can be characterized by calculating the full width at half maximum; and the peak area can be obtained by integrating the area under the peak.

[0063] The geometric difference is obtained by comparing the characteristic parameters with the corresponding parameters of the absorption peaks in the reference spectral template. Specifically, this involves comparing the characteristic parameters of the main absorption peaks extracted from the multi-wavelength spectral data at each time point with the characteristic parameters of the corresponding absorption peaks of the target food additive pre-stored in the reference spectral template. Various mathematical or statistical methods can be used for comparison, such as calculating the absolute difference, relative difference, percentage deviation, Euclidean distance, and correlation coefficient of the parameters. Thus, the differences in absorption peak characteristics between the multi-wavelength spectral data and the reference spectral template can be quantitatively obtained; these differences constitute the aforementioned geometric difference, reflecting the degree and direction of the systematic spectral shift.

[0064] The proposed method effectively quantifies and identifies systematic spectral shifts by parameterizing key absorption peaks in spectral data and comparing them precisely with corresponding parameters of a reference template. This parameter-based comparison method offers greater specificity and robustness compared to directly comparing the entire spectral curve. By extracting features such as peak position and shape, it can more accurately capture minute spectral changes caused by matrix effects, instrument drift, or sample state variations, which may lead to shifts, broadening, or distortions in absorption peaks. Quantitative comparison of these parameters allows for precise identification of these systematic shifts, providing a reliable basis for subsequent adaptive adjustments.

[0065] The above technical solution enables a more refined and quantitative identification of systematic spectral shifts between multi-wavelength spectral data and reference spectral templates. This comparison method based on characteristic parameters improves the accuracy and sensitivity of shift identification, especially for the trace detection of target food additives in complex matrices. It can more effectively capture subtle spectral changes, thus providing more accurate input for subsequent purity assessment and quantitative analysis, ultimately enhancing the overall reliability and accuracy of quality testing methods for brewed foods.

[0066] like Figure 2 The system shown is a quality testing system for brewed foods, which includes: Reference spectral template acquisition module 201 is used to acquire the reference spectral template of the target food additive; The spectral data acquisition module 202 is used to perform liquid chromatography separation on the sample and to acquire multi-wavelength spectral data of the eluted substances in real time during the separation process. The spectral correlation calculation module 203 is used to calculate the spectral correlation between multi-wavelength spectral data and a reference spectral template to obtain spectral correlation values. The purification signal spectrum generation module 204 is used to generate a purification signal spectrum of the target food additive based on the spectral correlation value. The quantitative analysis module 205 is used to perform quantitative analysis on the target food additive signal in the purified signal spectrum.

[0067] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A method for quality testing of brewed foods, applied to the detection of food additives, characterized in that, The method includes the following steps: Obtain the reference spectral template of the target food additive; The sample was separated by liquid chromatography, and multi-wavelength spectral data of the eluted substances were acquired in real time during the separation process; The spectral correlation between the multi-wavelength spectral data and the reference spectral template is calculated to obtain the spectral correlation value. Based on the spectral correlation values, a purification signal spectrum of the target food additive is generated; Quantitative analysis was performed on the target food additive signal in the purified signal spectrum; The step of calculating the spectral correlation between the multi-wavelength spectral data and the reference spectral template includes: Identify suspected signal peaks of target food additives in multi-wavelength spectral data; Analyze the spectral characteristics within the suspected signal peak to determine the region within the suspected signal peak where the spectral characteristics of the target object are dominant; Based on the region where the spectral characteristics of the target object are dominant, the signal-to-noise ratio and spectral purity of each spectral snapshot in the region are quantified. The weight of each spectral snapshot is determined based on the signal-to-noise ratio and spectral purity. The reference spectral template and the spectral data in the region are weighted and averaged according to the weight to construct a dynamic reference spectrum adapted to the current matrix environment. The spectral correlation between the multi-wavelength spectral data and the dynamic reference spectrum is calculated to obtain a spectral correlation value. The step of quantitatively analyzing the target food additive signal in the purified signal spectrum includes: Identify the peak regions of the target food additive signal in the purified signal spectrum; Analyze the signal fluctuation characteristics of the baseline regions on both sides of the peak region to obtain the baseline mean and baseline noise standard deviation; Based on the signal fluctuation characteristics and the signal change rate in the peak region, the starting point and ending point of the target food additive signal are determined. The starting point is determined when the signal intensity is higher than the sum of the baseline mean and k times the baseline noise standard deviation for N consecutive data points, and the first derivative of the signal reaches the slope threshold. The ending point is determined when the signal intensity falls back to below the sum of the baseline mean and k times the baseline noise standard deviation, and the absolute value of the first derivative of the signal is lower than the slope threshold for N consecutive data points. N ≥ 3, k is 3 to 5, and the slope threshold is m times the standard deviation of the first derivative of the signal in the baseline region, where m is 2 to 4. The signal of the target food additive between the starting point and the ending point is integrated to obtain a quantitative result.

2. The method for quality testing of brewed food products according to claim 1, characterized in that, The step of analyzing the spectral characteristics within the suspected signal peak and determining the region within the suspected signal peak where the spectral characteristics of the target object dominate includes: Spectral feature analysis is performed on the spectral snapshots within the suspected signal peaks to obtain the spectral features of each spectral snapshot; The region where the spectral features of the target object dominate is determined based on the distribution characteristics of the spectral features of the spectral snapshot. The region where the spectral features of the target object dominate is a continuous time period with a signal-to-noise ratio greater than 3 and a spectral purity greater than 1.

5. The signal-to-noise ratio is the ratio of the signal peak height in the target wavelength range of the spectral snapshot to the root mean square noise of the baseline region. The spectral purity is the ratio of the absorbance at the target characteristic absorption peak to the absorbance at the adjacent non-characteristic absorption peak in the spectral snapshot.

3. The method for quality testing of brewed food products according to claim 1, characterized in that, The step of identifying suspected signal peaks of target food additives in multi-wavelength spectral data includes: The similarity between the spectral snapshot at each time point in the multi-wavelength spectral data and the reference spectral template is calculated to obtain the similarity at each time point; Based on the similarity at each time point, regions in which the similarity at multiple consecutive time points is higher than a preset identification threshold are identified; Analyze multi-wavelength spectral data within the region to determine the local noise level and signal strength of the region. Based on the local noise level and the signal strength of the region, identify the starting and ending points where the signal strength is higher than the local noise level; The region between the starting point and the ending point is identified as a potential signal peak for the target food additive.

4. The method for quality testing of brewed food products according to claim 3, characterized in that, The step of identifying the starting and ending points where the signal strength is higher than the local noise level includes: The purity of the target food additive was evaluated by analyzing the multi-wavelength spectral data at each time point within the suspected signal peak, and the purity value at each time point was obtained. Analyze the change of the purity value over time to identify the boundary regions where the purity value changes from low to high levels, and the boundary regions where the purity value changes from high to low levels. By combining the boundary of the purity value transition region and the relative relationship between the signal strength and the local noise level, the start and end points of the target food additive signal are determined.

5. The method for quality testing of brewed food products according to claim 4, characterized in that, The step of evaluating the purity of the target food additive by analyzing the multi-wavelength spectral data at each time point within the suspected signal peak to obtain the purity value at each time point includes: Spectral feature analysis was performed on the multi-wavelength spectral data at each time point within the suspected signal peak to identify the systematic spectral shift characteristics with the reference spectral template; Based on the systematic spectral shift characteristics, the reference spectral template is adaptively adjusted to obtain an adaptive reference spectrum; The purity of the multi-wavelength spectral data at each time point is evaluated by comparing it with the adaptive reference spectrum to obtain the purity value at each time point.

6. The method for quality testing of brewed food products according to claim 5, characterized in that, The step of identifying the systematic spectral shift characteristics relative to the reference spectral template includes: By comparing the position or shape of the main absorption peak in the multi-wavelength spectral data at each time point with the position or shape of the corresponding absorption peak in the reference spectral template, the geometric difference is obtained. Based on the geometric differences, the systematic spectral shift characteristics are identified.

7. The method for quality testing of brewed food products according to claim 6, characterized in that, The step of comparing the position or shape of the main absorption peak in the multi-wavelength spectral data at each time point with the position or shape of the corresponding absorption peak in the reference spectral template to obtain the geometric difference includes: Extract the characteristic parameters of the main absorption peaks from the multi-wavelength spectral data at each time point; By comparing the characteristic parameters with the corresponding parameters of the absorption peaks in the reference spectral template, the geometric differences are obtained.

8. A quality testing system for brewed foods, used to execute the quality testing method for brewed foods according to any one of claims 1 to 7, characterized in that, The system includes: The reference spectral template acquisition module is used to acquire the reference spectral template of the target food additive. The spectral data acquisition module is used for liquid chromatography separation of samples and to acquire multi-wavelength spectral data of the eluted substances in real time during the separation process. The spectral correlation calculation module is used to perform spectral correlation calculation between the multi-wavelength spectral data and the reference spectral template to obtain a spectral correlation value. A purification signal spectrum generation module is used to generate a purification signal spectrum of the target food additive based on the spectral correlation value. The quantitative analysis module is used to perform quantitative analysis on the target food additive signal in the purified signal spectrum.

Images