A method and system for detecting the production quality of hot halogen products

By analyzing the overlap of absorption peaks in the infrared absorption spectrum, obtaining bias weights and performing weighted principal components analysis, the problem of overlapping absorption peaks caused by complex components in the detection of partially heated halogen products is solved, and the detection accuracy and efficiency are improved.

CN118782189BActive Publication Date: 2024-11-19HUNAN FANXIANGQING FOOD CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411260377.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-10
Publication Date
2024-11-19
Estimated Expiration
2044-09-10

AI Technical Summary

Technical Problem

In the quality detection of partially heated halogen products, due to the complex components, the overlap of absorption peaks of multiple components leads to difficulty in identifying key feature peaks, which affects the detection accuracy.

Method used

By obtaining the wave number subinterval of the infrared absorption spectrum, analyzing the overlap of the absorption peaks, screening out the severe overlap subinterval, and obtaining the bias weight based on the overlap between the overlap subinterval and the preset standard wave number interval, combined with the change trend of the absorption peak curve, the bias weight is obtained. , weighted principal component analysis of infrared absorption spectrum is performed to reduce dimensionality.

Benefits of technology

The accuracy and detection efficiency of component analysis of partially heated halogen products are improved, the spectral characteristics of the main components are retained, and the accuracy of quality detection is enhanced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118782189B_ABST
    Figure CN118782189B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of spectral analysis technology, and specifically to a spectral detection method and system for the production quality of hot halogen products. The present invention obtains the infrared absorption spectrum of the sample to be tested, and divides it to obtain all wave number sub-intervals; further obtains the degree of overlap of the absorption peaks in each wave number sub-interval, thereby screening out the heavily overlapping sub-intervals; further obtains the bias weight of each absorption peak curve in each heavily overlapping sub-interval; according to the bias weight of each absorption peak curve in each heavily overlapping sub-interval, the infrared absorption spectrum is subjected to dimensionality reduction analysis to evaluate the production quality of the sample to be tested. The present invention analyzes and evaluates the overlap of absorption peaks in the infrared absorption spectrum of hot halogen products, determines the bias weight during dimensionality reduction analysis, and thus performs targeted dimensionality reduction, thereby improving the efficiency of component analysis while retaining the main spectral characteristics of the main components, so as to improve the quality detection accuracy of hot halogen products.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of spectral analysis, and in particular to a method and system for spectral detection of production quality of hot halogen products. Background Art

[0002] "Hot-stewed marinated products" refer to marinated foods that have undergone specific processing, and various spices are usually added during the processing of the raw materials. Infrared spectroscopy can be used to analyze the ingredients of hot-stewed marinated products. Specific chemical substances in hot-stewed marinated products such as proteins, fatty acids and spices will appear as characteristic absorption peaks in the infrared spectrum. By analyzing the changing characteristics of the characteristic absorption peaks, food safety issues such as adulteration or spoilage of raw materials in hot-stewed marinated products can be detected and evaluated.

[0003] However, in the process of analyzing characteristic absorption peaks, due to the complex ingredients in hot marinated products, the absorption peaks of various ingredients will overlap. For example, esters, aldehydes, ketones and other compounds all contain carbon-oxygen double bonds (C=O functional groups), which will have C=O absorption peaks in the infrared spectrum. When analyzing the ingredients of hot marinated products, the overlap of such absorption peaks will interfere with the identification of key characteristic peaks and affect the ingredient analysis. In the prior art, the principal component analysis algorithm (Principal Component Analysis, PCA) is usually used to reduce the dimension of spectral data. However, the traditional PCA algorithm will equally consider all variables, but the identification of certain key components such as proteins is crucial to judging food quality, but these key components may be masked by other non-key components in the PCA results, resulting in low accuracy in quality detection of hot marinated products. Summary of the invention

[0004] In order to solve the technical problem that the main component analysis in the quality detection process of hot halogen products in the prior art is inaccurate, resulting in low quality detection accuracy, the purpose of the present invention is to provide a production quality spectral detection method and system for hot halogen products, and the technical scheme adopted is as follows:

[0005] The present invention provides a method for spectral detection of production quality of hot halogen products, the method comprising:

[0006] Obtaining an infrared absorption spectrum of the sample to be tested, wherein the infrared absorption spectrum contains absorption peak curves of all components in the sample to be tested; dividing the total wavenumber interval of the infrared absorption spectrum to obtain all wavenumber sub-intervals;

[0007] In each of the wave number sub-intervals, according to the fluctuation characteristics of all the absorption peak curves, combined with the total number of the absorption peak curves, the degree of overlap of the absorption peaks in each of the wave number sub-intervals is obtained; and according to the degree of overlap, a heavily overlapped sub-interval is screened out from all the wave number sub-intervals;

[0008] According to the overlap between each of the heavily overlapping sub-intervals and the preset standard wavenumber interval, combined with the change trend of each absorption peak curve in the heavily overlapping sub-interval and the neighboring wavenumber sub-interval, the bias weight of each absorption peak curve in each of the heavily overlapping sub-intervals is obtained; according to the bias weight of each absorption peak curve in each of the heavily overlapping sub-intervals, the infrared absorption spectrum is subjected to dimensionality reduction analysis.

[0009] Furthermore, the method for obtaining the degree of overlap includes:

[0010] In each of the wave number sub-intervals, according to the slope change and periodic characteristics of the curve sub-segments corresponding to all the absorption peak curves, the information change parameter of each of the wave number sub-intervals is obtained;

[0011] According to the information change parameter and the total number of the absorption peak curves, the overlap degree of the absorption peaks in each of the wave number sub-intervals is obtained; the information change parameter and the total number of the absorption peak curves are both positively correlated with the overlap degree.

[0012] Furthermore, the method for obtaining the information change parameter includes:

[0013] In each of the wave number sub-intervals, the curve sub-segment corresponding to each of the absorption peak curves is divided to obtain all local sub-segments in each of the curve sub-segments, and each of the local sub-segments is transformed in the frequency domain to obtain a frequency domain local sub-segment, and the period of each of the frequency domain local sub-segments is obtained;

[0014] According to the slope change difference between the adjacent local sub-segments in each of the curve sub-segments, a first change parameter of each of the curve sub-segments is obtained; according to the period change difference between the adjacent local sub-segments in each of the curve sub-segments, a second change parameter of each of the curve sub-segments is obtained;

[0015] According to the first change parameter and the second change parameter of all the curve subsegments, the information change parameter of each wave number sub-interval is obtained; the first change parameter and the second change parameter are both positively correlated with the information change parameter.

[0016] Furthermore, the method for obtaining the heavily overlapping subintervals includes:

[0017] The wave number sub-intervals whose overlapping degree is greater than a preset threshold are regarded as heavily overlapping sub-intervals.

[0018] Furthermore, the method for obtaining the preset standard wave number interval includes:

[0019] The spectral information of the target reference component in the sample to be tested is analyzed, and the wavenumber interval corresponding to the union of the absorption peaks corresponding to the wavenumber intervals of all the target reference components is used as the preset standard wavenumber interval.

[0020] Furthermore, the method for obtaining the bias weight includes:

[0021] If there is no intersection between the heavily overlapping sub-interval and the preset standard wavenumber interval, the bias weight of each absorption peak curve in the heavily overlapping sub-interval is set to 0, otherwise the minimum wavenumber difference between the two interval endpoints of the heavily overlapping sub-interval and the interval median of the preset standard wavenumber interval is negatively correlated to obtain the first weight of each absorption peak curve in each heavily overlapping sub-interval;

[0022] Obtaining a second weight of each of the absorption peak curves in each of the heavily overlapping subintervals according to the slope changes of all the local subsegments of each of the absorption peak curves in each of the heavily overlapping subintervals and the amplitude levels in the neighborhood wavenumber subintervals corresponding to the heavily overlapping subintervals;

[0023] According to the first weight and the second weight, the bias weight of each absorption peak curve in each heavily overlapping sub-interval is obtained, and the first weight and the second weight are both positively correlated with the bias weight.

[0024] Furthermore, the method for obtaining the second weight includes:

[0025] Taking the average absolute value of the slopes of all the local sub-segments of each absorption peak curve in the heavily overlapping sub-interval as the first change trend parameter;

[0026] The sum of the amplitude means of each absorption peak curve in a preset number of neighboring wave number subintervals of the heavily overlapping subinterval is used as a second variation trend parameter;

[0027] The first change trend parameter and the second change trend parameter of all the absorption peak curves are comprehensively considered to obtain a second weight; the first change trend parameter is positively correlated with the second weight, and the second change trend parameter is negatively correlated with the second weight.

[0028] Furthermore, the method for performing dimensionality reduction analysis on the infrared absorption spectrum includes:

[0029] In each of the heavily overlapping sub-intervals, the corresponding absorption peak curve is weighted using the bias weight, and based on the weighted principal component analysis algorithm, the dimension reduction results in all the heavily overlapping sub-intervals in the infrared absorption spectrum are obtained.

[0030] Furthermore, the preset threshold is 0.9.

[0031] The present invention also proposes a spectral detection system for the production quality of hot-steep halogen products. The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the spectral detection method for the production quality of hot-steep halogen products are implemented.

[0032] The present invention has the following beneficial effects:

[0033] The present invention obtains an infrared absorption spectrum of a sample to be tested, wherein the infrared absorption spectrum contains absorption peak curves of all components in the sample to be tested; divides the total wavenumber interval of the infrared absorption spectrum to obtain all wavenumber sub-intervals, thereby facilitating subsequent analysis of the absorption peak overlap in each wavenumber interval; in each wavenumber sub-interval, according to the fluctuation characteristics of all absorption peak curves, combined with the total number of absorption peak curves, obtains the degree of overlap of the absorption peaks in each wavenumber sub-interval; according to the degree of overlap, a heavily overlapped sub-interval is screened out from all wavenumber sub-intervals; according to the overlap of each heavily overlapped sub-interval with a preset standard wavenumber interval, combined with the change trend of each absorption peak curve in the heavily overlapped sub-interval and the neighboring wavenumber sub-interval, obtains the bias weight of each absorption peak curve in each heavily overlapped sub-interval, wherein the bias weight indirectly reflects the overlap influence in the heavily overlapped sub-interval, so as to determine the consideration weight in the subsequent dimensionality reduction analysis; and according to the bias weight of each absorption peak curve in each heavily overlapped sub-interval, performs dimensionality reduction analysis on the infrared absorption spectrum. The present invention analyzes and evaluates the absorption peak overlap in the infrared absorption spectrum of the hot-leaf halogen products, determines the bias weight during dimensionality reduction analysis, and thus performs targeted dimensionality reduction. While improving the efficiency of component analysis, the main spectral characteristics of the main components are retained, thereby improving the quality detection accuracy of the hot-leaf halogen products. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0035] Figure 1 A flow chart of a method for spectral detection of production quality of hot halogen products provided by one embodiment of the present invention;

[0036] Figure 2 A flowchart of a method for obtaining an overlap degree provided by an embodiment of the present invention;

[0037] Figure 3A flowchart of a method for obtaining bias weights provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the production quality spectral detection method and system of a hot halogen product proposed by the present invention, its specific implementation method, structure, characteristics and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0039] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0040] The specific scheme of the spectral detection method and system for production quality of hot halogen products provided by the present invention is described in detail below with reference to the accompanying drawings.

[0041] See also Figure 1 , which shows a flow chart of a method for spectral detection of production quality of hot halogen products provided by an embodiment of the present invention, specifically comprising:

[0042] Step S1, obtaining an infrared absorption spectrum of the sample to be tested, wherein the infrared absorption spectrum contains absorption peak curves of all components in the sample to be tested; dividing the total wavenumber interval of the infrared absorption spectrum to obtain all wavenumber sub-intervals.

[0043] In one embodiment of the present invention, any hot halogenated product to be inspected for quality is used as a sample to be tested, which is cut into small pieces. If necessary, the sample to be tested can be subjected to homogenization treatment such as grinding, and then the processed sample to be tested is evenly loaded on a salt crystal window, an infrared spectrometer is started for scanning, and the output infrared absorption spectrum is recorded; the horizontal axis of the infrared absorption spectrum is the wave number, and the vertical axis is the absorbance, and the infrared absorption spectrum contains the absorption peak curves of all components in the sample to be tested.

[0044] It should be noted that the process of obtaining the infrared absorption spectrum of the sample to be tested is a well-known technology and will not be described in detail here.

[0045] Due to the complex composition of the sample to be tested, the absorption peak curves in the infrared absorption spectrum may overlap greatly, making it impossible to accurately determine and analyze the absorption characteristics of each component, and thus unable to accurately evaluate the quality of the sample to be tested. In addition, because the absorption peaks in the infrared spectrum curve usually reflect the absorption of infrared light by the functional groups or chemical bonds of the molecular structure corresponding to each component of the sample to be tested, different components may contain the same or similar chemical bonds, which will show absorption characteristics within a specific wavelength range, that is, within a specific wavenumber range, that is, the absorption peaks corresponding to different components overlap, which makes the component quality analysis of the sample to be tested complicated.

[0046] Therefore, in one embodiment of the present invention, the total wave number interval of the infrared absorption spectrum is first divided to obtain all wave number sub-intervals, so as to facilitate the subsequent analysis of the absorption peak overlap in each wave number interval; since the wavelength range of infrared light is usually 2.5 μm to 25 μm, the total wave number interval in the infrared absorption spectrum is 4000 Up to 400 , then the total wave number interval is divided into 500 wave number units as a division point, that is, 4000 -3500 , 3500 -3000 And so on, the interval length of the last wave number sub-interval may be inconsistent with the interval lengths of other wave number sub-intervals; in other embodiments, the implementer may also adopt other division means or division lengths, which will not be repeated here.

[0047] Step S2, in each wave number sub-interval, according to the fluctuation characteristics of all absorption peak curves, combined with the total number of absorption peak curves, obtain the degree of overlap of the absorption peaks in each wave number sub-interval; according to the degree of overlap, select the heavily overlapped sub-interval from all wave number sub-intervals.

[0048] Taking into account that the absorption peak usually presents an obvious peak within a specific wavenumber range, if the fluctuation change of the absorption peak curve within a certain wavenumber sub-range is more obvious, that is, when it approximately presents a peak characteristic, and if the total number of absorption peak curves is itself large, then it means that the possibility of absorption peak overlap and the degree of overlap are greater.

[0049] Therefore, in a preferred embodiment of the present invention, the method for obtaining the degree of overlap includes:

[0050] See also Figure 2 , which shows a flow chart of a method for obtaining an overlap degree provided by an embodiment of the present invention, including:

[0051] Step S201 , in each wave number sub interval, according to the slope change and periodic characteristics of the curve sub segments corresponding to all absorption peak curves, obtain the information change parameter of each wave number sub interval.

[0052] Considering that the absorption peak curve corresponds to a curve subsegment in each wave number sub-interval, if the curve subsegment corresponds to an absorption peak, it should have a relatively obvious slope change, and the changes of different absorption peaks should have different change characteristics.

[0053] Among them, in a preferred embodiment of the present invention, in order to further analyze the slope change and change characteristics in each curve sub-segment, the curve sub-segment is first divided for analysis; then, in order to analyze the change characteristics of each curve sub-segment, it is further transformed into the frequency domain to evaluate its periodic characteristics; therefore, the method for obtaining the information change parameter includes:

[0054] In each wave number sub-interval, the curve sub-segment corresponding to each absorption peak curve is divided to obtain all local sub-segments in each curve sub-segment, and each local sub-segment is transformed in the frequency domain to obtain a frequency domain local sub-segment, and the period of each frequency domain local sub-segment is obtained;

[0055] According to the slope variation difference between adjacent local subsegments in each curve subsegment, a first variation parameter of each curve subsegment is obtained; according to the period variation difference between adjacent local subsegments in each curve subsegment, a second variation parameter of each curve subsegment is obtained;

[0056] According to the first change parameters and the second change parameters of all the curve subsegments, the information change parameter of each wave number subinterval is obtained; the first change parameter and the second change parameter are both positively correlated with the information change parameter.

[0057] As an example, first, every 10 wave number points are taken as a local sub-segment to obtain all local sub-segments, and the local sub-segments are further subjected to Fourier transform to obtain the corresponding frequency domain local sub-segments, and then the maximum frequency component is analyzed, and the frequency value corresponding to the maximum frequency component is inversely calculated to obtain the period value; it should be noted that the implementer may also divide the local sub-segments into other lengths, and may also adopt other frequency domain transformation methods, which and Fourier transform are existing technologies well known to technicians in this field and will not be elaborated here.

[0058] The calculation formula of the first variation parameter is: ;in, is the serial number of the wave number subinterval; For the The first wave number subinterval The first variation parameter of an absorption peak curve; For the The serial number of the local sub-segment within the wave number sub-interval; For the The total number of local subsegments within a wave number subinterval; For the The slope value of each local sub-segment; For the The slope value of each local sub-segment is calculated using the two-point formula, which will not be described in detail. is the absolute value symbol.

[0059] The calculation formula of the second variation parameter is: ;in, For the The first wave number subinterval The second variation parameter of an absorption peak curve; is the sequence number of the local sub-segment; is the total number of local sub-segments; For the The period value of a local sub-segment; For the The period value of a local sub-segment; is the absolute value symbol.

[0060] In the above two formulas, the greater the difference in slope changes between adjacent local sub-segments or the greater the difference in periodic characteristic values, the more obvious the fluctuation characteristics of the absorption peak curve, the greater the possibility that it corresponds to an absorption peak, and the greater the possibility of overlapping with other absorption peaks.

[0061] Then, in each wave number sub-interval, the first change parameter and the second change parameter of each absorption peak curve are added and combined to obtain the change parameter of each absorption peak curve, the change parameters of all absorption peak curves are accumulated, and then normalized, specifically using linear normalization, to obtain the information change parameter of the absorption peak in the wave number sub-interval.

[0062] In other examples, the implementer may also combine the first change parameter and the second change parameter by using basic mathematical operations such as multiplication or weighted summation or related mapping means, or may use other normalization means, which will not be described in detail here.

[0063] Step S202, obtaining the degree of overlap of absorption peaks in each wave number sub-interval according to the information change parameter and the total number of absorption peak curves; the information change parameter and the total number of absorption peak curves are positively correlated with the degree of overlap.

[0064] As an example, the information change parameter and the total number of absorption peak curves are multiplied and combined, and the product is normalized, specifically using linear normalization, to obtain the degree of overlap; in other examples, implementers can also use basic mathematical operations such as addition or weighted summation or related mapping methods to combine the two, and further use other normalization methods, which are not repeated here.

[0065] After obtaining the degree of overlap, in a preferred embodiment of the present invention, the wavenumber sub-intervals with an overlap degree greater than a preset threshold are further taken as heavily overlapping sub-intervals; wherein the preset threshold is 0.9. In other embodiments of the present invention, implementers may also set it by themselves, which will not be elaborated here.

[0066] Step S3, according to the overlap between each heavily overlapping sub-interval and the preset standard wavenumber interval, combined with the change trend of each absorption peak curve in the heavily overlapping sub-interval and the subsequent wavenumber interval, obtain the bias weight of each absorption peak curve in each heavily overlapping sub-interval; according to the bias weight of each absorption peak curve in each heavily overlapping sub-interval, perform dimensionality reduction analysis on the infrared absorption spectrum.

[0067] Taking into account that there are many components of the sample to be tested in the heavily overlapping sub-intervals of the absorption peaks, but in the process of production quality detection and analysis of the sample to be tested, analyzing whether the main components are adulterated or deteriorated is more important for evaluating the production quality; therefore, the embodiment of the present invention first analyzes the target reference component of the sample to be tested, the target reference component is part of the main components of the sample to be tested, obtains the preset standard wavenumber interval, and further analyzes and evaluates the bias weight of each absorption peak curve in the heavily overlapping sub-interval, the bias weight indirectly reflects the possibility that the corresponding absorption peak of each absorption peak curve in the heavily overlapping sub-interval is the corresponding absorption peak of the functional group of the target reference component, thereby determining the subsequent dimensionality reduction weight to increase its influence in the dimensionality reduction analysis results, thereby accurately evaluating the quality of the sample to be tested.

[0068] Preferably, in one embodiment of the present invention, the method for obtaining the preset standard wave number interval includes:

[0069] The spectral information of the target reference component in the sample to be tested is analyzed, and the wavenumber interval corresponding to the union of the absorption peaks of all the target reference components is used as the preset standard wavenumber interval.

[0070] As an example, the target reference components in the sample to be tested include at least protein, fatty acid, and methyl syringate; analyzing the spectral wavenumber band of the target reference components of the hot halogen product, the following can be obtained:

[0071] The main infrared absorption peak of protein is mainly manifested as: Amide I vibration, corresponding to the wave number range of 1700-1600 ; Amide II vibration, corresponding to the wave number range 1600-1500 ;

[0072] The main infrared absorption peaks of fatty acids are mainly manifested as: CH vibration, corresponding to the wave number range of 3000-2850 ; C=O vibration, corresponding to the wave number range 1735-1750 ;

[0073] The main infrared absorption peak of methyl syringate is mainly manifested as: C=O vibration, corresponding to the wave number range 1735-1750 ; CH vibration, corresponding to the wave number range 3000-2850 .

[0074] Based on the corresponding wavenumber bands of the main infrared absorption peaks of all target reference components in the sample to be tested, the preset standard wavenumber interval can be obtained as follows: .

[0075] In other examples, implementers may also analyze the types of target reference components and the union of the wavenumber bands corresponding to the corresponding main infrared absorption peaks according to actual needs, thereby obtaining other preset standard wavenumber intervals, which will not be elaborated here.

[0076] Preferably, in one embodiment of the present invention, the method for obtaining the bias weight includes:

[0077] See also Figure 3 , which shows a flow chart of a method for obtaining bias weights provided by an embodiment of the present invention, including:

[0078] Step S301: If there is no intersection between the heavily overlapping sub-interval and the preset standard wavenumber interval, the bias weight of each absorption peak curve in the corresponding heavily overlapping sub-interval is set to 0; otherwise, the minimum wavenumber difference between the two interval endpoints of the heavily overlapping sub-interval and the interval median of the preset standard wavenumber interval is negatively correlated to obtain the first weight of each absorption peak curve in each heavily overlapping sub-interval.

[0079] Considering that if the heavily overlapping sub-interval intersects with the wavenumber interval corresponding to the target reference component of the sample to be tested, that is, with the preset standard wavenumber interval, it means that the target reference component exists in the heavily overlapping sub-interval, then when performing dimensionality reduction analysis within the heavily overlapping sub-interval, it is necessary to increase the consideration weight to avoid deviations in the detection and analysis of production quality.

[0080] Therefore, as an example, if there is no intersection, the bias weight of each absorption peak curve in the corresponding heavily overlapping sub-interval is directly set to 0, otherwise further analysis is performed; since the preset standard wavenumber interval is the union of two intervals, and the interval length of the heavily overlapping sub-interval is 500 wavenumbers, the heavily overlapping sub-interval may be divided into one of them that is partially or completely located in the preset standard wavenumber interval, so the absolute value of the minimum wavenumber difference between the two interval endpoints of the heavily overlapping sub-interval and the interval median of the preset standard wavenumber interval is specifically obtained, which indirectly reflects the degree of overlap between the heavily overlapping sub-interval and the preset standard wavenumber interval. The smaller the value, the closer any endpoint of the heavily overlapping sub-interval is to the center of the preset standard wavenumber interval, and the greater the degree of overlap between the two intervals. After adding the anti-zero positive parameter 1, an inverse operation is performed, and the negative correlation mapping adjustment logic is performed to obtain the first weight of each absorption peak curve in each heavily overlapping sub-interval.

[0081] In other examples, the implementer may directly evaluate the size of the wavenumber interval corresponding to the intersection to obtain the first weight; or the absolute value of the minimum wavenumber difference may be used as an exponential function with the natural constant e as the base. In , to perform negative correlation mapping, other means may also be used, which will not be described here.

[0082] Step S302, obtaining the second weight of each absorption peak curve in each heavily overlapping subinterval according to the slope change of all local subsegments of each absorption peak curve in each heavily overlapping subinterval and the amplitude level in the neighborhood wavenumber subinterval of the corresponding heavily overlapping subinterval.

[0083] Considering that in each heavily overlapping sub-interval, if the absorption peak curve undergoes a large slope change, it is more likely that an absorption peak is generated in the heavily overlapping sub-interval, and at the same time, if the amplitude level of the absorption peak curve in the adjacent wavenumber interval before and after the heavily overlapping sub-interval is low, it means that most of the absorption peaks generated are located in the heavily overlapping sub-interval, then the overlapping influence is greater, and the consideration weight during dimensionality reduction analysis should also be greater.

[0084] In a preferred embodiment of the present invention, the method for obtaining the second weight includes:

[0085] The absolute mean of the slopes of all local subsegments of each absorption peak curve in the heavily overlapping subinterval is taken as the first change trend parameter; the sum of the amplitude means of each absorption peak curve in a preset number of neighborhood wavenumber subintervals of the heavily overlapping subinterval is taken as the second change trend parameter; the first change trend parameters and the second change trend parameters of all absorption peak curves are combined to obtain the second weight; the first change trend parameter is positively correlated with the second weight, and the second change trend parameter is negatively correlated with the second weight.

[0086] It should be noted that the preset number is 2, that is, one wave number sub-interval on each side of the heavily overlapping sub-interval is its neighborhood wave number sub-interval, and the implementer may also set other preset numbers.

[0087] As an example, the calculation formula of the second weight is: ;in, is the serial number of the heavily overlapping subinterval; is the serial number of the absorption peak curve; For the The first The second weight of an absorption peak curve; For the The serial number of the local sub-segment in the heavily overlapping sub-interval, which has the same meaning as the local sub-segment in step S201 and is not described here; For the The total number of local subsegments within heavily overlapping subintervals; For the The slope value of each local sub-segment; For the The serial number of the neighboring wavenumber subintervals of the heavily overlapping subintervals; is the preset quantity; For the The left and right sides of the heavily overlapping subintervals The first The mean amplitude of all data points of an absorption peak curve; is the first change trend parameter; is the second change trend parameter; To preset the anti-zero positive parameter, take 1 in this example.

[0088] In the above formula, the larger the numerator of the first change trend parameter is, the more likely it is that an absorption peak is generated in the heavily overlapping sub-interval, and the larger the second weight is; the smaller the second change trend parameter plus the preset anti-zero positive parameter is, the greater the possibility that most of the generated absorption peaks are located in the heavily overlapping sub-interval, and the larger the second weight is.

[0089] In other examples, the implementer may also convert the second change trend parameter into an exponential function with the natural constant e as the base. In , perform negative correlation mapping, and then multiply the negative correlation mapping result by the first change trend parameter to obtain a second weight; the implementer can also use other basic mathematical operations or related mapping methods to combine the two, which will not be repeated here.

[0090] Step S303, obtaining the bias weight of each absorption peak curve in each heavily overlapping sub-interval according to the first weight and the second weight, wherein the first weight and the second weight are both positively correlated with the bias weight.

[0091] As an example, the first weight and the second weight are multiplied and combined to obtain the bias weight of each absorption peak curve in each heavily overlapping sub-interval; in other examples, the implementer may also combine the two by using basic mathematical operations such as addition or weighted summation or related mapping methods, which will not be elaborated here.

[0092] After obtaining the bias weight of each absorption peak curve in each heavily overlapping subinterval, the infrared absorption spectrum is subjected to dimensionality reduction analysis to evaluate the production quality of the sample to be tested.

[0093] Preferably, in one embodiment of the present invention, the method for performing dimensionality reduction analysis on infrared absorption spectrum comprises:

[0094] In each heavily overlapping subinterval, the corresponding absorption peak curve is weighted by using the bias weight, and based on the weighted principal component analysis algorithm, the dimension reduction results in all heavily overlapping subintervals in the infrared absorption spectrum are obtained. It should be noted that weighted principal component analysis of infrared absorption spectra is a prior art well known to those skilled in the art and will not be described in detail here.

[0095] As an example, specifically in each heavily overlapping sub-interval, the sum of the bias weights of all absorption peak curves is used as the denominator, and the bias weight of each absorption peak curve is used as the numerator. The bias weight of each absorption peak curve is normalized, and each normalized bias weight is used to perform weighted averaging on the absorbance of the data points on each wavenumber scale of the corresponding absorption peak curve. The absorbance corresponding to the wavenumber point on each wavenumber scale is refitted to fit a new comprehensive absorption peak curve. Principal component analysis is performed on the comprehensive absorption peak curve in each heavily overlapping sub-interval, that is, principal component analysis is performed within a specific band, and the principal component analysis results in all heavily overlapping sub-intervals are combined to obtain the final downweighted result.

[0096] In one embodiment of the present invention, after obtaining the dimension reduction result of the infrared absorption spectrum, the dimension reduction result can be input into a pre-trained production quality inspection neural network, and the neural network can automatically output the production quality inspection result of the sample to be tested.

[0097] It should be noted that the training method of the production quality inspection neural network generally includes: first, obtaining high-quality hot marinated products produced based on the standard food production process, hot marinated products with different degrees of deterioration and corruption, and hot marinated products made from inferior raw materials, obtaining a large number of hot marinated products to make reference samples, and manually assigning corresponding production quality labels; using the aforementioned method to perform infrared spectroscopy detection on different reference samples and obtain their weighted dimensionality reduction results, thereby obtaining a training sample set; using the training sample set to train the convolutional neural network model to obtain a trained production quality inspection neural network; it is already a prior art well known to those skilled in the art, and implementers may also use other neural network models, which will not be repeated here.

[0098] In summary, the embodiment of the present invention obtains the infrared absorption spectrum of the sample to be tested, and the infrared absorption spectrum contains the absorption peak curves of all components in the sample to be tested; the total wave number interval of the infrared absorption spectrum is divided to obtain all wave number sub-intervals; in each wave number sub-interval, according to the fluctuation characteristics of all absorption peak curves, combined with the total number of absorption peak curves, the degree of overlap of the absorption peaks in each wave number sub-interval is obtained; according to the degree of overlap, the heavily overlapped sub-interval is screened out from all wave number sub-intervals; according to the overlap of each heavily overlapped sub-interval with the preset standard wave number interval, combined with the change trend of each absorption peak curve in the heavily overlapped sub-interval and the neighboring wave number sub-interval, the bias weight of each absorption peak curve in each heavily overlapped sub-interval is obtained; according to the bias weight of each absorption peak curve in each heavily overlapped sub-interval, the infrared absorption spectrum is subjected to dimensionality reduction analysis. The present invention analyzes and evaluates the overlap of absorption peaks in the infrared absorption spectrum of the hot halogen product, determines the bias weight during dimensionality reduction analysis, thereby reducing the dimension in a targeted manner, while improving the efficiency of component analysis, retaining the main spectral characteristics of the main components, so as to improve the quality detection accuracy of the hot halogen product.

[0099] The present invention also proposes a production quality spectral detection system for hot-steep halogen products. The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of a production quality spectral detection method for hot-steep halogen products are implemented.

[0100] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0101] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for detecting the production quality of hot halogen products by spectroscopy, characterized in that: The method comprises: Obtaining an infrared absorption spectrum of the sample to be tested, wherein the infrared absorption spectrum contains absorption peak curves of all components in the sample to be tested; dividing the total wavenumber interval of the infrared absorption spectrum to obtain all wavenumber sub-intervals; In each of the wave number sub-intervals, according to the fluctuation characteristics of all the absorption peak curves, combined with the total number of the absorption peak curves, the degree of overlap of the absorption peaks in each of the wave number sub-intervals is obtained; and according to the degree of overlap, a heavily overlapped sub-interval is screened out from all the wave number sub-intervals; According to the overlap between each heavily overlapping subinterval and the preset standard wavenumber interval, combined with the change trend of each absorption peak curve in the heavily overlapping subinterval and the neighboring wavenumber subinterval, the bias weight of each absorption peak curve in each heavily overlapping subinterval is obtained; according to the bias weight of each absorption peak curve in each heavily overlapping subinterval, the dimensionality reduction analysis of the infrared absorption spectrum is performed; The method for obtaining the degree of overlap includes: In each of the wave number sub-intervals, according to the slope change and periodic characteristics of the curve sub-segments corresponding to all the absorption peak curves, the information change parameter of each of the wave number sub-intervals is obtained; According to the information change parameter and the total number of the absorption peak curves, the degree of overlap of the absorption peaks in each of the wave number sub-intervals is obtained; the information change parameter and the total number of the absorption peak curves are both positively correlated with the degree of overlap; The method for obtaining the information change parameter includes: In each of the wave number sub-intervals, the curve sub-segment corresponding to each of the absorption peak curves is divided to obtain all local sub-segments in each of the curve sub-segments, and each of the local sub-segments is transformed in the frequency domain to obtain a frequency domain local sub-segment, and the period of each of the frequency domain local sub-segments is obtained; According to the slope change difference between the adjacent local sub-segments in each of the curve sub-segments, a first change parameter of each of the curve sub-segments is obtained; according to the period change difference between the adjacent local sub-segments in each of the curve sub-segments, a second change parameter of each of the curve sub-segments is obtained; According to the first change parameter and the second change parameter of all the curve subsegments, the information change parameter of each wave number sub-interval is obtained; the first change parameter and the second change parameter are both positively correlated with the information change parameter.

2. According to the method for detecting the production quality of hot halogen products according to claim 1, it is characterized in that: The method for obtaining the heavily overlapping subintervals includes: The wave number sub-intervals whose overlapping degree is greater than a preset threshold are regarded as heavily overlapping sub-intervals.

3. According to the method for detecting the production quality of hot halogen products according to claim 1, it is characterized in that: The method for obtaining the preset standard wave number interval includes: The spectral information of the target reference component in the sample to be tested is analyzed, and the wavenumber interval corresponding to the union of the absorption peaks corresponding to the wavenumber intervals of all the target reference components is used as the preset standard wavenumber interval.

4. A method for detecting the production quality of hot halogen products according to claim 1, characterized in that: The method for obtaining the bias weight includes: If there is no intersection between the heavily overlapping sub-interval and the preset standard wavenumber interval, the bias weight of each absorption peak curve in the heavily overlapping sub-interval is set to 0, otherwise the minimum wavenumber difference between the two interval endpoints of the heavily overlapping sub-interval and the interval median of the preset standard wavenumber interval is negatively correlated to obtain the first weight of each absorption peak curve in each heavily overlapping sub-interval; Obtaining a second weight of each of the absorption peak curves in each of the heavily overlapping subintervals according to the slope changes of all the local subsegments of each of the absorption peak curves in each of the heavily overlapping subintervals and the amplitude levels in the neighborhood wavenumber subintervals corresponding to the heavily overlapping subintervals; According to the first weight and the second weight, the bias weight of each absorption peak curve in each heavily overlapping sub-interval is obtained, and the first weight and the second weight are both positively correlated with the bias weight.

5. A method for detecting the production quality of hot halogen products according to claim 4, characterized in that: The method for obtaining the second weight includes: Taking the average absolute value of the slopes of all the local sub-segments of each absorption peak curve in the heavily overlapping sub-interval as the first change trend parameter; The sum of the amplitude means of each absorption peak curve in a preset number of neighboring wave number subintervals of the heavily overlapping subinterval is used as a second variation trend parameter; The first change trend parameter and the second change trend parameter of all the absorption peak curves are comprehensively considered to obtain a second weight; the first change trend parameter is positively correlated with the second weight, and the second change trend parameter is negatively correlated with the second weight.

6. The method for detecting the production quality of hot halogen products according to claim 1, characterized in that: The method for performing dimensionality reduction analysis on the infrared absorption spectrum comprises: In each of the heavily overlapping sub-intervals, the corresponding absorption peak curve is weighted using the bias weight, and based on the weighted principal component analysis algorithm, the dimension reduction results in all the heavily overlapping sub-intervals in the infrared absorption spectrum are obtained.

7. The method for detecting the production quality of hot halogen products according to claim 2, characterized in that: The preset threshold is 0.

9.

8. A production quality spectrum detection system for hot halogen products, characterized in that: The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of a method for spectral detection of production quality of a thermally biased halogen product as described in any one of claims 1 to 7 are implemented.