CN113740394A - Qualitative identification method of doped bovine colostrum based on dielectric spectrum technology - Google Patents
Qualitative identification method of doped bovine colostrum based on dielectric spectrum technology Download PDFInfo
- Publication number
- CN113740394A CN113740394A CN202111209944.6A CN202111209944A CN113740394A CN 113740394 A CN113740394 A CN 113740394A CN 202111209944 A CN202111209944 A CN 202111209944A CN 113740394 A CN113740394 A CN 113740394A
- Authority
- CN
- China
- Prior art keywords
- bovine colostrum
- doped
- model
- dielectric
- dielectric spectrum
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 241000283690 Bos taurus Species 0.000 title claims abstract description 164
- 235000021277 colostrum Nutrition 0.000 title claims abstract description 150
- 210000003022 colostrum Anatomy 0.000 title claims abstract description 150
- 238000000034 method Methods 0.000 title claims abstract description 90
- 238000001453 impedance spectrum Methods 0.000 title claims abstract description 82
- 238000005516 engineering process Methods 0.000 title claims abstract description 24
- 235000013336 milk Nutrition 0.000 claims abstract description 62
- 210000004080 milk Anatomy 0.000 claims abstract description 62
- 239000008267 milk Substances 0.000 claims abstract description 62
- 238000007781 pre-processing Methods 0.000 claims abstract description 41
- 238000009499 grossing Methods 0.000 claims abstract description 34
- 230000008030 elimination Effects 0.000 claims abstract description 25
- 238000003379 elimination reaction Methods 0.000 claims abstract description 25
- 238000004458 analytical method Methods 0.000 claims abstract description 18
- 238000005259 measurement Methods 0.000 claims abstract description 18
- 238000012360 testing method Methods 0.000 claims abstract description 18
- 230000000694 effects Effects 0.000 claims abstract description 15
- 238000001228 spectrum Methods 0.000 claims abstract description 15
- 238000000513 principal component analysis Methods 0.000 claims abstract description 6
- 239000000523 sample Substances 0.000 claims description 61
- 238000012937 correction Methods 0.000 claims description 15
- 230000009466 transformation Effects 0.000 claims description 14
- 238000012706 support-vector machine Methods 0.000 claims description 12
- 238000010239 partial least squares discriminant analysis Methods 0.000 claims description 10
- 230000006651 lactation Effects 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 5
- 230000009467 reduction Effects 0.000 claims description 4
- 230000002411 adverse Effects 0.000 claims description 3
- 238000001566 impedance spectroscopy Methods 0.000 claims description 3
- 230000032696 parturition Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 239000004575 stone Substances 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 2
- 238000010187 selection method Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 19
- 230000008901 benefit Effects 0.000 abstract description 3
- 235000013305 food Nutrition 0.000 abstract description 2
- 239000002245 particle Substances 0.000 description 15
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 9
- 102000004169 proteins and genes Human genes 0.000 description 7
- 108090000623 proteins and genes Proteins 0.000 description 7
- 235000013365 dairy product Nutrition 0.000 description 6
- 235000020247 cow milk Nutrition 0.000 description 5
- GUBGYTABKSRVRQ-QKKXKWKRSA-N Lactose Natural products OC[C@H]1O[C@@H](O[C@H]2[C@H](O)[C@@H](O)C(O)O[C@@H]2CO)[C@H](O)[C@@H](O)[C@H]1O GUBGYTABKSRVRQ-QKKXKWKRSA-N 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 235000013861 fat-free Nutrition 0.000 description 3
- 239000008101 lactose Substances 0.000 description 3
- 235000016709 nutrition Nutrition 0.000 description 3
- 238000002203 pretreatment Methods 0.000 description 3
- 230000035945 sensitivity Effects 0.000 description 3
- 239000007787 solid Substances 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000010348 incorporation Methods 0.000 description 2
- 239000004615 ingredient Substances 0.000 description 2
- 108060003951 Immunoglobulin Proteins 0.000 description 1
- 241000124008 Mammalia Species 0.000 description 1
- 241001465754 Metazoa Species 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000009395 breeding Methods 0.000 description 1
- 230000001488 breeding effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000009614 chemical analysis method Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000008367 deionised water Substances 0.000 description 1
- 229910021641 deionized water Inorganic materials 0.000 description 1
- 238000009585 enzyme analysis Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000036039 immunity Effects 0.000 description 1
- 102000018358 immunoglobulin Human genes 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002329 infrared spectrum Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010238 partial least squares regression Methods 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 230000010287 polarization Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000004879 turbidimetry Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N27/00—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means
- G01N27/02—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means by investigating impedance
- G01N27/22—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means by investigating impedance by investigating capacitance
- G01N27/221—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means by investigating impedance by investigating capacitance by investigating the dielectric properties
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Chemical & Material Sciences (AREA)
- Physics & Mathematics (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
- Electrochemistry (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Artificial Intelligence (AREA)
Abstract
The invention discloses a qualitative identification method of doped bovine colostrum based on a dielectric spectrum technology, and belongs to the field of rapid detection of food. The method adopts a dielectric property measurement system to measure dielectric parameters of a batch of bovine colostrum and doped bovine colostrum. The method comprises the steps of preprocessing a sample dielectric spectrum to improve model prediction accuracy, wherein a noise elimination preprocessing method is obviously superior to a scattering elimination preprocessing method, and the effect of combining Savitzky-Golay smoothing and a second derivative is optimal. And then, taking the full spectrum and the dielectric spectrum data extracted by the principal component analysis as model input, and respectively establishing 2 linear models and 2 nonlinear models. The linear discriminant analysis model established based on the full spectrum is an optimal model, and the identification accuracy of the test set is 97.37%. The model has 93.18% of identification accuracy on verified milk samples from other sources. The method qualitatively identifies the doped bovine colostrum based on the dielectric spectrum technology, and has the advantages of low cost, high precision, rapid detection, capability of being used for field detection and the like.
Description
Technical Field
The invention belongs to the field of rapid detection of foods, and particularly relates to a qualitative identification method of doped bovine colostrum based on a dielectric spectrum technology.
Background
Bovine colostrum refers to the milk secreted by healthy cows within 3 days after delivery. Bovine colostrum is rich in nutritional ingredients such as protein and fat, and immunoglobulin for improving human immunity, and has higher nutritional value, so that the bovine colostrum is popular among consumers. However, the production of bovine colostrum is very low and therefore much more expensive than normal milk. In order to obtain a high profit, some cow milk vendors often add a certain amount of milk to the cow colostrum. This behavior not only seriously affects the quality and nutritional value of bovine colostrum and its dairy products, but also infringes the legitimate rights and interests of consumers. Therefore, the detection of common milk adulteration in the bovine colostrum has important significance.
Compared with normal milk, the contents of fat, protein and the like in the bovine colostrum are higher, the particle size distribution is larger, the identification difficulty of the colostrum is further increased by the strong heterogeneous characteristic, and the quality of the bovine colostrum is difficult to accurately judge only by the physical characteristics of color, density, viscosity and the like. Although the detection precision can be improved by the chemical analysis methods commonly used in laboratories, such as the radioimmunodiffusion method, the enzyme analysis method, the immunotransmission turbidimetry method and the like, the problems of time and labor waste in detection, high cost, high technical requirements on operators and the like exist. The near infrared spectrum technology has certain possibility in the aspect of evaluating the quality of the bovine colostrum, but the technology is easily influenced by fat, protein and other large particles in the bovine colostrum, and has a strong scattering effect, so that the detection precision is difficult to satisfy. Therefore, the exploration of a method which has low cost, high precision and rapid detection and can be used for on-site detection has important significance for the guarantee of the quality of the bovine colostrum and the product thereof.
The dielectric spectroscopy technology is a technology for acquiring the electrical characteristics of a substance in a wider frequency range, and can acquire more composition and structure information of a sample. In addition, the dielectric spectroscopy technique has the advantage of being fast, and applicable to both field and on-line detection. Therefore, the method is widely applied to the detection of milk quality, such as the detection of fat, protein, lactose and water in milk. In addition, the dielectric spectrum has a larger wavelength range and a deeper penetration depth, is not easily influenced by scattering of large particles such as fat globules and the like, and has certain advantages for heterogeneous milk detection. However, no identification method for bovine colostrum doping based on the dielectric spectrum technology is available at present, so that a qualitative identification method for doped bovine colostrum based on the dielectric spectrum technology is needed to be developed, the key point is to select a proper pretreatment and modeling method to improve the identification accuracy, and a qualitative identification technology which is low in cost, high in precision, rapid in detection and capable of being used for field detection is provided for bovine colostrum doping.
Disclosure of Invention
Aiming at the defects and shortcomings in the prior art, the invention aims to provide a qualitative identification method of doped bovine colostrum. The dielectric spectrum measuring instrument is used for collecting a batch of bovine colostrum and the dielectric spectrum doped with the bovine colostrum, and a noise elimination method and a scattering elimination method are selected for preprocessing according to the characteristic of strong heterogeneity of the bovine colostrum so as to eliminate noise interference generated in the dielectric spectrum measuring process or scattering influence generated by the heterogeneous characteristic of the bovine colostrum. After the cow milk sample is preprocessed by a preferred method through a dielectric spectrum, a linear or nonlinear model for qualitatively identifying the doped cow colostrum is further established, the model identification accuracy is contrasted and analyzed, and the optimal model is determined. And substituting the dielectric spectrum data of the pretreated unknown bovine colostrum sample into the optimal model to qualitatively identify the unknown bovine colostrum sample.
A qualitative identification method of doped bovine colostrum based on a dielectric spectrum technology is characterized by comprising the following steps:
the method comprises the following steps: collecting milk samples of different cows from different regions, seasons and feeding conditions; the bovine colostrum sample is milk of a cow in 3 days of parturition, and the common milk sample is milk of a cow in a normal lactation period; storing bovine colostrum and normal milk samples at room temperature, and mixing the bovine colostrum samples with the normal milk samples according to the proportion of 10%, 20%, 30%, 40% and 50% of the doped mass fraction before the test to obtain the doped bovine colostrum samples; dividing each bovine colostrum sample and the prepared doped bovine colostrum sample into 3 parts for later use;
step two: the network analyzer was preheated for 1h before testing and then sequentially calibrated for open circuit, short circuit and 50 Ω load. Setting instrument measurement parameters which mainly comprise a frequency measurement range and measurement points, and finally calibrating a coaxial probe with an open circuit at the tail end; before measuring the sample dielectric spectrum, the prepared sample is placed on an oscillator and shaken for about 2 min, so that the milk components are uniformly distributed. Each sample was measured 3 times in duplicate, with the average of the 3 results being the final result. During the measurement, the temperature of the sample is 25 +/-1 ℃;
step three: according to the characteristic of strong heterogeneity of bovine colostrum, a noise elimination method, namely Savitzky-Golay smoothing, a second derivative and a combination method thereof, and a scattering elimination method, namely standard normal variable transformation, multivariate scattering correction and a combination method thereof are selected for preprocessing, so that noise interference generated in the dielectric spectrum measurement process or scattering influence generated due to the heterogeneous characteristic of bovine colostrum are eliminated; dividing the pretreated bovine colostrum and the doped bovine colostrum at each ratio into a correction set and a test set by a classical sample division method Kennard-Stone according to a ratio of 3: 1;
step four: based on a noise elimination type and scattering elimination type preprocessing method, comparing the accuracy of linear partial least squares discriminant analysis and nonlinear support vector machine model identification of bovine colostrum and doped bovine colostrum, and preferably selecting a dielectric spectrum preprocessing method combining Savitzky-Golay smoothing and second derivative;
step five: original dielectric spectrum data are processed by a preferred Savitzky-Golay smoothing and second derivative combined preprocessing method, full spectrum and dielectric spectrum data subjected to principal component analysis dimensionality reduction extraction are used as model input, and two linear models of partial least squares discriminant analysis and linear discriminant analysis and two nonlinear models of a support vector machine and a multilayer perceptron are established. Based on the recognition accuracy of the linear model and the nonlinear model, comparing and determining the optimal model as a linear discriminant analysis model based on a full spectrum;
step six: and for unknown bovine colostrum samples from different sources, completing the collection of dielectric spectrums according to the second step, and substituting collected dielectric spectrum data of the unknown bovine colostrum samples into the linear discriminant analysis model determined in the fifth step after combined pretreatment of Savitzky-Golay smoothing and second-order derivative determined in the fourth step to quickly and accurately identify the samples.
The noise elimination pretreatment in the fourth step is obviously superior to the scattering elimination pretreatment, which shows that the adverse effect of doped bovine colostrum is identified as noise interference based on a dielectric spectrum method, rather than scattering caused by heterogeneous characteristics of the bovine colostrum; and fifthly, the prediction effect of a linear model established by identifying the doped bovine colostrum based on the dielectric spectrum technology is obviously superior to that of a nonlinear model.
The pretreatment method and the model selection method provided by the invention do not exclude qualitative identification of doping of colostrum of mammals with the common milk, wherein the colostrum and the common milk are from the same animal.
The invention has the following beneficial technical effects:
(1) the detection is rapid, the operation is simple, and the on-line measurement is convenient. The method provided by the invention only needs to measure the dielectric spectrum of the unknown bovine colostrum sample, and can identify the dielectric spectrum through the corresponding model after preprocessing the dielectric spectrum. The method provided by the invention is beneficial to developing a special detection instrument and realizes the online rapid detection of the doped bovine colostrum;
(2) the identification accuracy is high. After the noise reduction pretreatment is carried out on the dielectric spectrum of the bovine colostrum sample, the accuracy of a correction set and the accuracy of a test set of the established linear discriminant analysis model doped with the qualitative identification of the bovine colostrum are respectively 99.14 percent and 97.37 percent. The model has 93.18% of identification accuracy on verified milk samples from other sources. Therefore, the qualitative identification method of the doped bovine colostrum based on the invention can obtain higher identification precision and has stronger practicability.
Drawings
FIG. 1 is a flow chart of a qualitative identification method of doped bovine colostrum based on dielectric spectrum technology;
FIG. 2 is a schematic diagram of a dielectric property measurement system;
FIG. 3 is a graph of the mean dielectric spectrum of bovine colostrum spiked at different ratios used in the experiment;
FIG. 4 is a graph of the results of a preferred noise cancellation type method on sample dielectric spectrum preprocessing;
FIG. 5 is the identification result of unknown cow's milk samples based on the optimal model established by the dielectric spectrum technique;
reference numerals:
1. the system comprises a network analyzer, 2, a coaxial probe, 3, a thermometer, 4, a water bath, 5, a lifting device and 6, a computer.
Detailed Description
The method has good adaptability to qualitative identification of the doped bovine colostrum of different varieties of dairy cows; due to more varieties of dairy cows, the method takes bovine colostrum produced by 'Holstein' dairy cows within 3 days of parturition and normal milk samples produced by dairy cows in a normal lactation period as an example, and qualitative identification of other dairy cows mixed with bovine colostrum can be carried out by referring to the method of the example. Specifically, according to the measured bovine colostrum sample, a dielectric spectrum pretreatment method and a modeling method are reasonably selected, so that the bovine colostrum sample can be rapidly and accurately judged.
The qualitative identification method of the doped bovine colostrum based on the dielectric spectrum technology is further explained by combining the drawings and the embodiments of the specification given by the inventor.
The method according to the embodiment of the invention comprises the following steps:
the method comprises the following steps: the method comprises the steps of collecting cow milk samples of different cows from different regions, seasons and feeding conditions, wherein the cow colostrum sample is milk of the cows in 3 days of delivery, and the normal milk sample is milk of the cows in normal lactation period. In the embodiment, the milk samples for establishing the model are all collected from 'Holstein' cows fed in three different milk farms in the Yangling area of Shaanxi province, and the milk samples for verifying the model are collected from another different milk farm in the Yangling area.
The indexes of the main components of bovine colostrum and normal milk are shown in table 1. The contents of protein and non-fat milk solid of the bovine colostrum are far larger than normal milk, the fat content is slightly larger than normal milk, but the lactose content and the water content are smaller than normal milk. The higher particle concentration of bovine colostrum enhances its heterogeneous properties. It should be noted that, because of differences in breeding conditions, cow individuals, lactation time and the like, the components of the bovine colostrum and the normal milk sample used for the test have a wide range, which also indicates that the test bovine milk sample has better representativeness.
TABLE 1 test of the content of the principal ingredients of bovine colostrum and milk samples, (g/100 g)
| Milk sample | Fat | Protein | Non-fat solids | Lactose | Water (W) |
| Colostrum of cow | 4.20±1.03 | 9.70±1.45 | 14.63±1.40 | 3.49±0.29 | 80.35±1.39 |
| Regular milk | 4.06±0.52 | 3.61±0.27 | 9.11±0.21 | 4.71±0.16 | 86.70±0.95 |
The particle size distribution of the bovine colostrum and the normal milk is shown in Table 2, the average values of D (0.5) and D [3,2] of the bovine colostrum are respectively 2.435 and 1.710 μm, which are slightly larger than the normal milk; the average values of D (0.9) and D (4,3) of the bovine colostrum are 7.394 and 3.270 mu m respectively, which are obviously higher than that of normal milk. This indicates a larger particle size distribution range in bovine colostrum and a stronger heterogeneous character.
TABLE 2 particle size distribution of bovine colostrum and normal milk
| Particle size distribution parameter (. mu.m) | d(0.5) | d(0.9) | D[4,3] | D[3,2] |
| Colostrum of cow | 2.435±0.225 | 7.394±0.796 | 3.270±0.471 | 1.710±0.102 |
| Regular milk | 2.425±0.055 | 5.459±0.169 | 2.689±0.150 | 1.700±0.034 |
The symbols in table 2 illustrate: d (0.5) and d (0.9) represent the particle sizes corresponding to the cumulative percent particle size distribution of the sample at 50% and 90%, respectively; d4, 3 and D3, 2 respectively represent the volume moment of the particles and the surface volume mean diameter.
By analyzing the main components and the particle size distribution of the bovine colostrum and the normal milk, the contents of protein, fat, non-fat solid and other components in the bovine colostrum are higher, the water content is lower, and the standard deviation of the contents of the components is far greater than that of the normal milk; on the other hand, the particle size distribution parameters D (0.9) and D4, 3 in the bovine colostrum are far larger than that of normal colostrum, and both of the parameters influence the dielectric property of the bovine colostrum, thereby providing a strong proof for the feasibility of applying the dielectric spectrum technology to qualitatively identify the doped bovine colostrum.
In the embodiment, the doped bovine colostrum samples are prepared from different bovine colostrum and normal milk samples according to the doped mass fraction of 10%, 20%, 30%, 40% and 50%. Each time 3 parts of bovine colostrum and 3 parts of normal milk were collected, the original bovine colostrum sample was mixed with the normal milk sample in the proportions of 10%, 20%, 30%, 40% and 50% by doping mass fraction to obtain 15 doped bovine colostrums, wherein each 3 replicates at each doping level. To obtain more bovine colostrum samples, 3 different cows of the original bovine colostrum samples were mixed in equal amounts per trial to obtain 4 new mixed bovine colostrum samples, i.e. a total of 7 bovine colostrum samples per trial. The tests for establishing the model were carried out 7 times in total, and finally 154 samples were obtained. The test for verifying the model was performed 2 times in total, and finally 44 samples were obtained. Preserving bovine colostrum and normal milk samples at room temperature, and completing sample configuration and dielectric spectrum acquisition within 5 h; each bovine colostrum sample and the formulated adulterated bovine colostrum sample were divided into 3 portions for use.
Step two: in this example, the dielectric parameters of the samples were measured using a grid analyzer, coaxial probe and 85070 software from Agilent technology, Malaysia. Fig. 2 shows a schematic diagram of a dielectric parameter measurement system, which is preheated for 1h before testing and then subjected to open circuit, short circuit and 50 Ω load calibration in sequence. After 85070 software is started, the measuring frequency range is set to be 20-4500 MHz, and the number of measuring frequency points is 201. And finally, carrying out open circuit, short circuit and 25-DEG C deionized water calibration on the coaxial probe. Before measuring the dielectric spectrum, the prepared sample is placed on an oscillator and shaken for about 2 min, so that the milk components are uniformly distributed. Each sample was measured 3 times in duplicate, with the average of 3 measurements being the final result. During the measurement, the sample temperature was 25. + -. 1 ℃.
FIG. 3 is a graph of the mean dielectric spectrum of bovine colostrum spiked at different ratios used in the experiment; FIG. 3(a) shows the average dielectric spectrum for the sampleε′Curve of the bovine colostrum sample over most of the frequency band investigated (above about 50 MHz)ε′The values increase with increasing milk doping level. Water is a polar molecule, bovineOf colostrumε′Mainly dominated by water. With the incorporation of normal milk, free water molecules in the sample increase, and the molecular orientation polarization effect is gradually enhanced and appears in a dielectric spectrumε′The value of (a) increases. FIG. 3(b) is the average dielectric spectrum corresponding to 20% and 50% of the bovine colostrum and the normal milk mixed thereinε"Curve line.ε"The change occurred at about 2000 MHz with increasing levels of incorporation of normal milk, decreasing first and increasing second.
Step three: the raw dielectric spectrum curve is preprocessed. In order to find out main influence factors of heterogeneous characteristics on the identification of the doped bovine colostrum based on the dielectric spectrum method, eliminate noise interference generated in the dielectric spectrum measurement process or scattering influence due to the heterogeneous characteristics of the bovine colostrum and improve the model prediction performance, and by combining the heterogeneous characteristics of the bovine colostrum, two preprocessing methods of a noise elimination method, namely Savitzky-Golay smoothing, a second derivative method and a combination method thereof, and a scattering elimination method, namely standard normal variable transformation, multivariate scattering correction and a combination method thereof are adopted. Meanwhile, the influence of combined preprocessing of Savitzky-Golay smoothing and standard normal variable transformation on the recognition effect is contrastively analyzed.
Wherein the Savitzky-Golay smoothing is Savitzky-Golay, the second derivative is second derivative, the standard normal variable is transformed to a standard normal variable, and the multivariate scattering is corrected to a multivariate scatter correction. Savitzky-Golay smoothing and second derivative preprocessing can reduce random noise caused by electric signal interference and the like and improve the signal-to-noise ratio. The influence of scattering on spectral lines caused by uneven distribution of sample particles and different particle sizes can be eliminated by standard normal variable transformation and multivariate scattering correction preprocessing.
Dividing the pretreated bovine colostrum and the doped bovine colostrum at each ratio into a correction set and a test set by a classical sample division method Kennard-Stone method according to a ratio of 3: 1.
Step four: based on different dielectric spectrum preprocessing, the accuracy of recognizing the bovine colostrum and the doped bovine colostrum by comparing partial least squares discriminant analysis and a support vector machine model, and the dielectric spectrum preprocessing method combining Savitzky-Golay smoothing and a second derivative is preferably selected. Table 3 lists the results of linear partial least squares discriminant analysis and nonlinear support vector machine model identification after 154 sample dielectric spectrum data are respectively preprocessed by noise elimination preprocessing, Savitzky-Golay smoothing, second derivative and combination thereof, and scattering elimination preprocessing method, standard normal variable transformation, multivariate scattering correction and combination thereof, and 7 preprocessing modes of Savitzky-Golay smoothing and standard normal variable transformation combination, and lists the prediction results without dielectric spectrum preprocessing as comparison. As can be seen from table 3, when the sample division ratio is 3:1, the improvement effect of the model precision by three preprocessing, namely Savitzky-Golay smoothing, second-order derivative and combination of Savitzky-Golay smoothing and second-order derivative based on the noise elimination method is better than that by standard normal variable transformation, multivariate scattering correction and combination preprocessing of standard normal variable transformation and multivariate scattering correction based on the scattering elimination method, and the recognition accuracy of the two models, namely linear partial least squares discriminant analysis and nonlinear support vector machine, is improved, which indicates that the main influences of the heterogeneous characteristics of bovine colostrum on the adverse influence of the dielectric spectrum prediction are noise, baseline drift and the like, rather than the scattering effect caused by the heterogeneous characteristics. This phenomenon is particularly pronounced in non-linear support vector machine models. In addition, the effect of Savitzky-Golay smoothing and standard normal variable transformation, which are combined preprocessing of the noise elimination method and the scattering elimination method, is superior to that of standard normal variable transformation, and further shows that the elimination of noise interference in the dielectric spectrum can improve the model prediction performance. Meanwhile, as can be seen from table 3, the preprocessing method is that the accuracy of the test set of the Savitzky-Golay smoothing and second derivative combined time-biased least squares discriminant analysis model is the highest, reaching 97.37%. Therefore, a dielectric spectrum preprocessing method combining Savitzky-Golay smoothing with second derivative is preferred.
TABLE 3 accuracy of two models under different combination methods
The symbols in table 3 illustrate:
untreated means that the dielectric spectrum has not been pre-processed; S-G represents Savitzky-Golay smoothing preprocessing; SD represents second derivative preprocessing; S-G + SD represents the combined preprocessing of Savitzky-Golay smoothing and second derivative; SNV represents standard normal variable transformation preprocessing; MSC represents multivariate scatter correction preprocessing; SNV + MSC represents the pretreatment of the combination of standard normal variable transformation and multivariate scattering correction; S-G + SNV represents Savitzky-Golay smoothing combined with standard normal variable transformation preprocessing.
FIG. 4 is a diagram of the results of a preferred noise cancellation-like approach to sample dielectric spectrum preprocessing, i.e., Savitzky-Golay smoothing, second derivative, and samples preprocessed by a combination of Savitzky-Golay smoothing and second derivativeε′Andε"curve line. FIG. 4 (a) shows that Savitzky-Golay smoothing removes the noise effects present in the original dielectric spectrum, resulting in a smoother, more stationary dielectric spectrum curve; the dielectric spectrum after the second derivative processing in fig. 4 (b) is more obvious in characteristic, and meanwhile, the interferences such as noise, baseline drift and the like in the original dielectric spectrum are removed. In FIG. 4 (c), Savitzky-Golay smoothing and second derivative have good complementarity, and the combined processing of the Savitzky-Golay smoothing and the second derivative reduces noise interference in the original dielectric spectrum, highlights component difference information in the dielectric spectrum and enhances the signal-to-noise ratio of the dielectric spectrum.
Symbolic illustration in fig. 4: S-G represents Savitzky-Golay smoothing; SD represents the second derivative; S-G + SD represents a Savitzky-Golay smoothing combined with a second derivative.
Step five: original dielectric spectrum data are processed by a preferred Savitzky-Golay smoothing and second derivative combined preprocessing method, the characteristics of dielectric spectra of bovine colostrum and doped bovine colostrum are combined, full spectra and dielectric spectrum data extracted by principal component analysis dimensionality reduction are used as input, two linear models of partial least squares discriminant analysis and linear discriminant analysis and two nonlinear models of a support vector machine and a multilayer perceptron are established, and the optimal model for qualitatively identifying the bovine colostrum and the doped bovine colostrum is determined. The partial least squares discriminant analysis is partial least squares discriminant analysis, and is a discriminant analysis method based on partial least squares regression. The linear discriminant analysis is linear discriminant analysis, and the principle is that the distance between two types of samples after projection is as large as possible by finding out the classified effective projection direction. The support vector machine is a support vector machine, and is a classification method based on a risk minimization idea, and the principle is to establish an optimal classification hyperplane so as to maximize blank areas among different samples. The multi-layer perceptron is a multi-layer perceptron, a feedforward artificial neural network model that maps multiple input datasets onto a single output dataset. The qualitative identification of the spiked bovine colostrum is shown in table 4.
And comprehensively evaluating the recognition performance of the linear and nonlinear models by sensitivity, specificity and accuracy. The sensitivity refers to the proportion,%, of the bovine colostrum sample which is correctly judged as the bovine colostrum sample; the specificity refers to the proportion,%, of the doped bovine colostrum sample correctly judged as the doped bovine colostrum sample; the accuracy rate is the ratio of the bovine colostrum and the adulterated bovine colostrum sample which are correctly judged in percent.
By comparing the qualitative identification accuracy of the doped bovine colostrum of the linear model and the nonlinear model, the identification accuracy of the linear model and the nonlinear model which are established based on the full frequency is higher than that of the model which is established based on the principal component analysis, namely, the accuracy of the model which is established by analyzing and extracting the principal component of the principal component shows that the variables related to the bovine colostrum sample information are removed in the process of analyzing and extracting the principal component of the principal component, and the noise interference, the baseline drift and the like are removed from the full spectrum data through Savitzky-Golay smoothing and second derivative preprocessing, so that a better identification effect is obtained. The recognition accuracy of the full-spectrum-based linear model test set is 97.37 percent, which is higher than 94.74 percent of the full-spectrum-based nonlinear model. This indicates that the linear model is more favorable for the identification of adulterated bovine colostrum. The linear discriminant analysis model based on the full spectrum has the best recognition performance, and the accuracy rates of a correction set and a test set are respectively 99.14% and 97.37%. Thus, for this example, the best model for qualitatively identifying adulterated bovine colostrum was a full spectrum based linear discriminant analysis model. The result shows that the reasonable selection of the dielectric spectrum pretreatment method and the model has important significance for improving the accuracy of the dielectric spectrum technology-based identification of the doped bovine colostrum model.
TABLE 4 qualitative discrimination of doped bovine colostrum by linear and nonlinear models
The symbols in table 4 illustrate: PLS-DA, LDA, SVM and MLP respectively represent partial least square discriminant analysis, linear discriminant analysis, support vector machine and multilayer perceptron model; FS represents the full spectrum, PCA represents the characteristic wavelength extracted from the full spectrum by principal component analysis.
Step six: and for unknown bovine colostrum samples from different sources, completing the collection of dielectric spectrums according to the second step, and substituting collected dielectric spectrum data of the unknown bovine colostrum samples into the linear discriminant analysis model determined in the fifth step after combined pretreatment of Savitzky-Golay smoothing and second-order derivative determined in the fourth step to quickly and accurately identify the samples.
The unknown samples used for model verification in the embodiment are 44 samples, wherein 14 bovine colostrums are used, and 30 doped bovine colostrums are used for further verifying the performance of the established dielectric spectrum technology-based identification doped bovine colostrums model. FIG. 5 shows the identification result of unknown cow milk samples based on the optimal model established by the dielectric spectrum technique. The sensitivity, specificity and identification accuracy of the established linear discrimination model to samples in unknown samples are respectively 85.71%, 96.67% and 93.18%, which shows that the method has good identification performance and applicability to milk samples from different sources. The accuracy of the model established for the identification of unknown samples is lower than the accuracy of the identification of the samples used for modeling, mainly because of the different sample sources and acquisition times.
The embodiments show that the invention can rapidly and accurately carry out qualitative identification on the doped bovine colostrum by utilizing the dielectric spectrum technology.
It should be noted that the above-mentioned contents are only for illustrating one technical solution of the present invention, and not for limiting the scope of the present invention, and that the simple modifications or equivalent substitutions of the technical solution of the present invention by those skilled in the art do not exceed the scope of the present invention.
Claims (4)
1. A qualitative identification method of doped bovine colostrum based on a dielectric spectrum technology is characterized by comprising the following steps:
the method comprises the following steps: collecting milk samples of different cows from different regions, seasons and feeding conditions; the bovine colostrum sample is milk of a cow in 3 days of parturition, and the common milk sample is milk of a cow in a normal lactation period; storing bovine colostrum and normal milk samples at room temperature, and mixing the bovine colostrum samples with the normal milk samples according to the proportion of 10%, 20%, 30%, 40% and 50% of the doped mass fraction before the test to obtain the doped bovine colostrum samples; dividing each bovine colostrum sample and the prepared doped bovine colostrum sample into 3 parts for later use;
step two: preheating a network analyzer for 1h before testing, and then sequentially carrying out open circuit, short circuit and 50 omega load calibration on the network analyzer; setting instrument measurement parameters which mainly comprise a measurement frequency range and measurement point numbers, and finally calibrating a coaxial probe with an open circuit at the tail end; before measuring the dielectric spectrum of the sample, placing the prepared sample on an oscillator and shaking for about 2 min to ensure that the milk components are uniformly distributed; each sample was measured 3 times repeatedly, and the average of the 3 measurements was the final result; during the measurement, the temperature of the sample is 25 +/-1 ℃;
step three: according to the characteristic of strong heterogeneity of bovine colostrum, a noise elimination method, namely Savitzky-Golay smoothing, a second derivative and a combination method thereof, and a scattering elimination method, namely standard normal variable transformation, multivariate scattering correction and a combination method thereof are selected for preprocessing, so that noise interference generated in the dielectric spectrum measurement process or scattering influence generated due to the heterogeneous characteristic of bovine colostrum are eliminated; dividing the pretreated bovine colostrum and the doped bovine colostrum at each ratio into a correction set and a test set by a classical sample division method Kennard-Stone according to a ratio of 3: 1;
step four: based on a noise elimination type and scattering elimination type preprocessing method, comparing the accuracy of linear partial least squares discriminant analysis and nonlinear support vector machine model identification of bovine colostrum and doped bovine colostrum, and preferably selecting a dielectric spectrum preprocessing method combining Savitzky-Golay smoothing and second derivative;
step five: processing original dielectric spectrum data by a preferred Savitzky-Golay smoothing and second derivative combined preprocessing method, taking full spectrum and dielectric spectrum data subjected to principal component analysis dimensionality reduction extraction as model input, and establishing two linear models of partial least squares discriminant analysis and linear discriminant analysis and two nonlinear models of a support vector machine and a multilayer perceptron; based on the recognition accuracy of the linear model and the nonlinear model, comparing and determining the optimal model as a linear discriminant analysis model based on a full spectrum;
step six: and for unknown bovine colostrums samples from different sources, completing the collection of dielectric spectrums according to the second step, and substituting collected dielectric spectrum data of the unknown samples into the linear discriminant analysis model determined in the fifth step after the combined pretreatment of Savitzky-Golay smoothing and second-order derivative determined in the fourth step to quickly and accurately identify the samples.
2. The qualitative identification method of doped bovine colostrum based on the dielectric spectrum technique as claimed in claim 1, wherein the noise elimination type preprocessing is significantly better than the scattering elimination type preprocessing in step four, which indicates that the adverse effect of identifying the doped bovine colostrum based on the dielectric spectrum method is noise interference rather than scattering effect caused by heterogeneous nature of bovine colostrum.
3. The qualitative identification method of doped bovine colostrum based on the dielectric spectrum technology according to claim 1, wherein the linear model established by identifying the doped bovine colostrum based on the dielectric spectrum technology in the fifth step has a prediction effect significantly better than that of the non-linear model.
4. A qualitative identification method of doped bovine colostrum based on dielectric spectroscopy according to claim 1, wherein the pre-processing method and the model selection method do not exclude qualitative identification of doped common milk of mammalian colostrum.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111209944.6A CN113740394B (en) | 2021-10-18 | 2021-10-18 | Qualitative identification method of doped bovine coloctrum based on dielectric spectrum technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111209944.6A CN113740394B (en) | 2021-10-18 | 2021-10-18 | Qualitative identification method of doped bovine coloctrum based on dielectric spectrum technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113740394A true CN113740394A (en) | 2021-12-03 |
| CN113740394B CN113740394B (en) | 2024-03-01 |
Family
ID=78726903
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111209944.6A Active CN113740394B (en) | 2021-10-18 | 2021-10-18 | Qualitative identification method of doped bovine coloctrum based on dielectric spectrum technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113740394B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106248746A (en) * | 2016-09-25 | 2016-12-21 | 西北农林科技大学 | A kind of milk protein method for quickly detecting contents based on dielectric and magnetic technology |
| CN108318442A (en) * | 2018-02-06 | 2018-07-24 | 江苏康缘药业股份有限公司 | A kind of detection method suitable for Chinese medicine suspending system |
| CN108562622A (en) * | 2018-02-05 | 2018-09-21 | 西北农林科技大学 | A kind of fresh sheep breast fast detecting method for total number of bacterial colony based on dielectric property technology |
| AU2020101607A4 (en) * | 2019-08-28 | 2020-09-10 | Agro-environmental Protection Institute, Ministry Of Agriculture And Rural Affairs | Method for rapidly predicting nitrogen and phosphorus content in slurry movement routes of multiple different large-scale dairy farms by comprehensively integrating all factors |
| CN112730312A (en) * | 2021-02-03 | 2021-04-30 | 西北农林科技大学 | Doped bovine colostrum qualitative identification method based on near infrared spectrum technology |
-
2021
- 2021-10-18 CN CN202111209944.6A patent/CN113740394B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106248746A (en) * | 2016-09-25 | 2016-12-21 | 西北农林科技大学 | A kind of milk protein method for quickly detecting contents based on dielectric and magnetic technology |
| CN108562622A (en) * | 2018-02-05 | 2018-09-21 | 西北农林科技大学 | A kind of fresh sheep breast fast detecting method for total number of bacterial colony based on dielectric property technology |
| CN108318442A (en) * | 2018-02-06 | 2018-07-24 | 江苏康缘药业股份有限公司 | A kind of detection method suitable for Chinese medicine suspending system |
| AU2020101607A4 (en) * | 2019-08-28 | 2020-09-10 | Agro-environmental Protection Institute, Ministry Of Agriculture And Rural Affairs | Method for rapidly predicting nitrogen and phosphorus content in slurry movement routes of multiple different large-scale dairy farms by comprehensively integrating all factors |
| CN112730312A (en) * | 2021-02-03 | 2021-04-30 | 西北农林科技大学 | Doped bovine colostrum qualitative identification method based on near infrared spectrum technology |
Non-Patent Citations (1)
| Title |
|---|
| 任东;沈俊;任顺;王纪华;陆安祥;: "一种面向土壤重金属含量检测的X射线荧光光谱预处理方法研究", 光谱学与光谱分析, no. 12 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113740394B (en) | 2024-03-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP6339244B2 (en) | Method for predicting sugar content and acidity of fruit using multivariate statistical analysis of FT-IR spectrum data | |
| Balan et al. | Application of Attenuated Total Reflectance-Fourier Transform Infrared (ATR-FTIR) spectroscopy coupled with chemometrics for detection and quantification of formalin in cow milk | |
| CN106248746B (en) | A kind of milk protein method for quickly detecting contents based on dielectric and magnetic technology | |
| An et al. | Effect of spectral pretreatment on qualitative identification of adulterated bovine colostrum by near-infrared spectroscopy | |
| CN109540838B (en) | Method for rapidly detecting acidity in fermented milk | |
| Wang et al. | Effect of homogenisation on detection of milk protein content based on NIR diffuse reflectance spectroscopy | |
| Liu et al. | A new comprehensive index for discriminating adulteration in bovine raw milk | |
| CN107917897A (en) | The method of the special doctor's food multicomponent content of near infrared ray | |
| CN112730312A (en) | Doped bovine colostrum qualitative identification method based on near infrared spectrum technology | |
| Scavarda et al. | Cocoa smoky off-flavour: A MS-based analytical decision maker for routine controls | |
| Soto-Barajas et al. | Prediction of the type of milk and degree of ripening in cheeses by means of artificial neural networks with data concerning fatty acids and near infrared spectroscopy | |
| CN109374548A (en) | A method of quickly measuring nutritional ingredient in rice using near-infrared | |
| CN113310930A (en) | Spectral identification method of high-temperature sterilized milk, pasteurized milk and pasteurized milk mixed with high-temperature sterilized milk | |
| Strani et al. | Milk renneting: Study of process factor influences by FT-NIR spectroscopy and chemometrics | |
| Risoluti et al. | Assessing the quality of milk using a multicomponent analytical platform MicroNIR/chemometric | |
| He et al. | Rapid detection of adulteration of goat milk and goat infant formulas using near-infrared spectroscopy fingerprints | |
| Yavari et al. | Internet of Things milk spectrum profiling for industry 4.0 dairy and milk manufacturing | |
| CN105223140A (en) | The method for quickly identifying of homology material | |
| CN107976417B (en) | Crude oil type identification method based on infrared spectrum | |
| Spina et al. | Mid-infrared (MIR) spectroscopy for the detection of cow’s milk in buffalo milk | |
| González‐Martín et al. | Discrimination of seasonality in cheeses by near‐infrared technology | |
| CN110672578A (en) | Model universality and stability verification method for polar component detection of frying oil | |
| CN113310929A (en) | Soybean powder doped in high-temperature sterilized milk and spectral identification method of doping proportion thereof | |
| CN113324940A (en) | Spectrum grading method for super-high-quality milk, high-protein special milk, high-milk-fat special milk and common milk | |
| CN113740394B (en) | Qualitative identification method of doped bovine coloctrum based on dielectric spectrum technology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |