CN114414521A - Milk main component measuring method based on infrared multispectral sensor - Google Patents
Milk main component measuring method based on infrared multispectral sensor Download PDFInfo
- Publication number
- CN114414521A CN114414521A CN202210066510.3A CN202210066510A CN114414521A CN 114414521 A CN114414521 A CN 114414521A CN 202210066510 A CN202210066510 A CN 202210066510A CN 114414521 A CN114414521 A CN 114414521A
- Authority
- CN
- China
- Prior art keywords
- milk
- protein
- model
- fat
- infrared
- 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.)
- Pending
Links
- 235000013336 milk Nutrition 0.000 title claims abstract description 114
- 210000004080 milk Anatomy 0.000 title claims abstract description 114
- 239000008267 milk Substances 0.000 title claims abstract description 112
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000005259 measurement Methods 0.000 claims abstract description 61
- 102000004169 proteins and genes Human genes 0.000 claims abstract description 48
- 108090000623 proteins and genes Proteins 0.000 claims abstract description 48
- 238000010561 standard procedure Methods 0.000 claims abstract description 8
- 238000012549 training Methods 0.000 claims description 23
- 235000019625 fat content Nutrition 0.000 claims description 13
- 238000012360 testing method Methods 0.000 claims description 10
- 235000019624 protein content Nutrition 0.000 claims description 8
- 238000011156 evaluation Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 5
- 238000001228 spectrum Methods 0.000 claims description 5
- 238000012795 verification Methods 0.000 claims description 4
- 238000007696 Kjeldahl method Methods 0.000 claims description 3
- 238000000691 measurement method Methods 0.000 claims description 3
- 239000002245 particle Substances 0.000 claims description 3
- 230000017525 heat dissipation Effects 0.000 claims description 2
- 230000000149 penetrating effect Effects 0.000 claims description 2
- 238000007781 pre-processing Methods 0.000 claims description 2
- 235000018102 proteins Nutrition 0.000 abstract 3
- 102000014171 Milk Proteins Human genes 0.000 abstract 1
- 108010011756 Milk Proteins Proteins 0.000 abstract 1
- 230000007613 environmental effect Effects 0.000 abstract 1
- 235000021243 milk fat Nutrition 0.000 abstract 1
- 235000021239 milk protein Nutrition 0.000 abstract 1
- 239000003925 fat Substances 0.000 description 29
- 239000000126 substance Substances 0.000 description 9
- 235000013365 dairy product Nutrition 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 6
- 238000001514 detection method Methods 0.000 description 6
- 238000002790 cross-validation Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000006872 improvement Effects 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 4
- 241000283690 Bos taurus Species 0.000 description 3
- 238000004497 NIR spectroscopy Methods 0.000 description 3
- 238000010521 absorption reaction Methods 0.000 description 3
- 238000007796 conventional method Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- QAOWNCQODCNURD-UHFFFAOYSA-N Sulfuric acid Chemical compound OS(O)(=O)=O QAOWNCQODCNURD-UHFFFAOYSA-N 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 239000003153 chemical reaction reagent Substances 0.000 description 2
- 239000000470 constituent Substances 0.000 description 2
- 230000001066 destructive effect Effects 0.000 description 2
- 238000002329 infrared spectrum Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012845 near infrared spectroscopy analysis Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 241000282414 Homo sapiens Species 0.000 description 1
- 238000004566 IR spectroscopy Methods 0.000 description 1
- 238000000862 absorption spectrum Methods 0.000 description 1
- 238000003556 assay Methods 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 238000009395 breeding Methods 0.000 description 1
- 230000001488 breeding effect Effects 0.000 description 1
- 238000009614 chemical analysis method Methods 0.000 description 1
- 231100000481 chemical toxicant Toxicity 0.000 description 1
- 238000011109 contamination Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000002366 mineral element Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000000053 physical method Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 239000003440 toxic substance Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/3577—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing liquids, e.g. polluted water
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/20—Identification of molecular entities, parts thereof or of chemical compositions
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Theoretical Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Crystallography & Structural Chemistry (AREA)
- Immunology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Pathology (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a milk main component measuring method based on an infrared multispectral sensor, which comprises the steps of obtaining the characteristics of a milk sample, establishing a measuring model and measuring the contents of protein and fat in milk. The method comprises the steps of collecting wavelength characteristics of different types of milk samples through a multispectral sensor, determining the contents of protein and fat in the milk samples by combining a standard method, constructing a milk main component measuring model, and using the trained measuring model for measuring the contents of the protein and the fat in the milk. The multi-wavelength light intensity characteristics corresponding to the samples are obtained based on the multi-spectral sensor, so that a single sample corresponds to a plurality of effective characteristics, the measurement precision of milk protein and fat is greatly improved, the rapid nondestructive prediction of the milk sample is realized, and the method has the advantages of simplicity, rapidness, low cost and environmental protection.
Description
Technical Field
The invention relates to the technical field of infrared multispectral analysis, in particular to a milk main component measuring method based on an infrared multispectral sensor.
Background
Milk is one of the oldest natural dairy products, is known as 'white blood', and has considerable importance to human beings. According to statistics, the milk yield in the North America in 2020 is close to 1.11 million tons, and is increased by 2.1% compared with the milk yield in 2019 (the milk yield in the United states is 1.01 million tons, and is increased by 2.2%), and the main reason is the improvement of the population number of the dairy cows and the milk yield per unit. Therefore, in the process of purchasing and production management of milk, the content of the main components of the milk needs to be detected accurately and quickly, reference basis can be provided for quality analysis and production process quality control of the milk, and scientific guidance can be provided for excellent breeding of the dairy cows.
The main component content of milk is the most important factor for determining the quality of milk. In most countries, the milk trade is based on the protein and fat content of milk, and therefore its quality is of great economic importance. Thus, there is a need in the dairy industry for a fast, reliable method to determine the concentration of major components, such as protein and fat. The traditional measurement method is still mainly based on chemical and physical methods, and not only has complex operation, needs professional technicians, is long in time consumption and high in cost, but also needs to use harmful or even toxic chemical reagents in the detection process, and also needs to store expensive analytical reagents by a specific method. The need for batch tests and real-time on-line assays, for example, the commonly used methods for protein and fat determination are kjeldahl and bobo, respectively, which separate proteins and fats by adding sulfuric acid. These standard analytical methods, although the most mature, are time consuming and prone to chemical contamination.
Ultrasonic methods and near-infrared spectroscopy are rapid, reagent-free and non-destructive analysis techniques that are advantageous over conventional methods and are increasingly being used to quantitatively analyze milk for its component content. Although the ultrasonic detection method can make up for many problems of the chemical analysis method, the method is easily influenced by factors such as temperature and poor modeling method. Near infrared spectroscopy has been used in the dairy industry for over 40 years and is a method that does not require the use of chemicals or glassware and only requires a few minutes to produce results. A great deal of basic research is carried out on near infrared spectroscopic analysis at the kansai university in japan and the laval university in canada, so that the near infrared spectroscopic analysis has made great progress in analyzing milk components and the development of green chemical analysis has been promoted. Therefore, in the past, there are numerous papers on the application of near infrared to milk, such as researchers detecting common adulterants in milk by using infrared spectroscopy to guarantee the quality of milk; the infrared spectrum technology is proved to be reliable and accurate, so that some researchers use the technology to test mineral elements such as K, Ca, P, Na and the like contained in the milk. Meanwhile, the infrared spectrum technology plays an important role in genetic improvement of the dairy cows.
However, the current method for measuring the content of milk ingredients has some problems. Firstly, the near infrared spectroscopy instrument is expensive, large in size, complex in operation and maintenance, and is mostly applied to professional laboratories or large scientific research institutions. Secondly, due to the overlapping of near infrared absorption peaks of substances, it is difficult to measure the component content by a single wavelength, and the measurement accuracy is low. Meanwhile, researches show that a milk quality detection method becomes one of the biggest problems to be solved urgently in the current dairy industry.
Disclosure of Invention
In order to solve the problems in the background technology, the invention provides a milk main component measuring method based on an infrared multispectral sensor.
In order to achieve the purpose, the invention provides the following scheme:
the method for measuring the main components of the milk based on the infrared multispectral sensor comprises the following steps:
acquiring multi-wavelength light intensity characteristic data of different types of milk based on an infrared multispectral sensor, and measuring the contents of protein and fat in the milk sample by combining a standard method;
taking the multi-wavelength light intensity characteristic data as input, taking the protein and fat content as output, constructing a milk main component measurement model, training and perfecting model parameters;
and inputting the multi-wavelength light intensity characteristic data of the milk sample to be measured into the measurement model, and obtaining the measurement result by the output protein and fat contents.
Preferably, the establishing step of the milk main component measurement model comprises the following steps:
acquiring characteristic data, and acquiring multi-wavelength light intensity data of a broadband near-infrared light source penetrating through a milk sample by using a multi-wavelength spectrum sensor to serve as the characteristic data of a measurement model;
obtaining label data, and measuring the content of protein and fat in the milk sample by using a standard method as the label data of a measurement model;
dividing a data set, preprocessing the multi-wavelength light intensity data, and dividing the preprocessed multi-wavelength light intensity data into a training set and a test set;
training a model, namely taking characteristic data in a training set as input of the measurement model, taking label data in the training set as output, training the measurement model by using a leave-out method, and detecting the stability of the measurement model by using repeated verification and five-fold cross verification;
establishing a measurement model, and measuring the protein and fat contents of the samples in the test set through the measurement model;
and evaluating the measurement model, carrying out grid search, and determining final parameters to obtain the milk main component measurement model.
Preferably, the milk major constituent measurement model is evaluated based on a decision coefficient, a mean absolute error, a mean square error, and a symmetrical mean absolute percentage as evaluation indexes.
Preferably, multi-wavelength light intensity data of a sample to be measured is acquired through an infrared multispectral sensor and input into the trained measurement model to obtain the content of protein and fat in the milk sample.
Preferably, the infrared multispectral sensor is in the model AS7263, and the infrared multispectral sensor is provided with six near-infrared channels, and the wavelengths of the six near-infrared channels are 610nm, 680nm, 730nm, 760nm, 810nm and 860nm respectively.
Preferably, the data acquisition device comprises: the constant light intensity driving circuit comprises a constant light intensity driving circuit, a light source, a heat dissipation plate, a sample groove and an infrared multispectral sensor, wherein the light source is a broadband near-infrared LED and is used for enabling emitted broadband infrared light to be incident into a milk sample and absorbed by protein and fat particles, and then the multi-wavelength light intensity characteristics of the milk sample are obtained through the infrared multispectral sensor.
Preferably, the milk major constituent measurement model is a gradient lifting regression tree model GBRT.
Preferably, the protein and fat content of the milk sample is measured based on standard methods for preparing label data of the milk measurement model, and the measurement process comprises the following steps: and (3) determining the content of fat in the milk sample according to a Gerber method, and determining the content of protein according to a Kjeldahl method.
The invention has the beneficial effects that:
by the established protein and fat detection method, the chemical pollution can be reduced during measurement, the dependence on complex operation during measurement can be reduced, and the rapid nondestructive prediction of the milk sample is realized. In addition, the multispectral acquisition system has high transportability, can be distributed on a milk component content measurement site in a pasture, and can rapidly measure the milk component content in real time.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of a method of an embodiment of the present invention;
FIG. 2 is a diagram illustrating the results of 100 training sessions of a model according to an embodiment of the present invention, wherein (a) is a protein model and (b) is a fat model;
FIG. 3 is a graph showing measurement results according to an embodiment of the present invention; wherein (a) is a protein model and (b) is a fat model.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
It is well known that milk is composed of 87% water, 3.7% fat, 3.4% protein and other substances. According to the Lambert-beer law, as the content of the components of the sample changes, the spectral characteristics of the sample also change. This is the theoretical basis for the analytical method of the present invention. Fat, protein and other components in the milk have an absorption effect on near infrared light, so that the absorption spectrum of the milk can be used for determining the component content of the milk. Meanwhile, the multispectral sensor is innovatively used for simultaneously acquiring the characteristic data of a plurality of wavelengths corresponding to the sample, so that one sample corresponds to a plurality of characteristic values, and the accurate measurement of the content of fat and protein in the milk can be realized by combining a machine learning model. Meanwhile, the types of milk samples are more, the corresponding labels are fewer, and the precision of the traditional light intensity fitting method is lower. In order to verify the effectiveness of the method provided by the invention, the method provided by the invention is compared and researched with other machine learning methods, and finally the prediction precision of the gradient lifting regression algorithm is found to be highest and the error is minimal.
Referring to the attached figure 1, the method for determining the main components of the milk based on the infrared multispectral sensor specifically comprises the following steps:
acquiring multi-wavelength light intensity characteristic data of different types of milk based on an infrared multispectral sensor, and measuring the contents of protein and fat in the milk sample by combining a standard method;
taking the multi-wavelength light intensity characteristic data as input, taking the protein and fat content as output, constructing a milk main component measurement model, training and perfecting model parameters;
and inputting the multi-wavelength light intensity characteristic data of the milk sample to be measured into the measurement model, and obtaining the measurement result by the output protein and fat contents.
In this example, 71 domestic and imported milks were selected, and the samples covered all the commercially available pure milks, with the protein content of the samples ranging from 2.4/100ml to 4.0g/100ml and the fat content ranging from 0/100ml to 4.6g/100 ml.
According to international standard ISO 2446: 2008 (Bob method) and ISO 8968-1: 2014 (Kjeldahl method) respectively determining the contents of fat and protein in the milk and using the contents as labels of milk measurement models.
And collecting the multichannel light intensity characteristics of each kind of milk by using a data acquisition device to serve as the input characteristics of the milk measurement model. The data acquisition device includes: the light source is a broadband infrared LED, emitted broadband infrared light enters a milk sample and is absorbed by protein and fat particles, and then the multi-wavelength light intensity characteristics of the milk sample are obtained through the multi-wavelength spectrum sensor. As the content of the milk components changes, the characteristics of the milk components also change. In the embodiment, the multi-wavelength spectrum sensor integrates 6 independent photosensitive elements and optical filters, so that the phenomenon of overlapping of absorption peaks when milk components are measured by single wavelength is solved, and the design of an optical channel and a signal processing unit is omitted.
Constructing a data set based on the label data and the wavelength characteristics of different types of milk, and randomly dividing the data set into a training set and a testing set; the training set is used for adjusting the hyper-parameters in the model and training the model, and the testing set is used for detecting the generalization ability of the model.
And (3) taking the intensities of 6 channel wavelengths of 610nm, 680nm, 730nm, 760nm, 810nm and 860nm as input, and taking the contents of protein and fat as output of the model to construct a milk measurement model, wherein the milk measurement model is a GBRT model.
The process of constructing the GBRT model is as follows:
the model is first initialized, which is a tree with only root nodes whose output values are c, which makes the loss function extremely small:
wherein f is0(x) For the initial learner, L () is a loss function.
Secondly, m regression trees are built:
(a) calculating the residual error rmi
Wherein f ism-1For the model after iteration m-1 times, f (x)i) Is the predicted value of the ith data.
(b) Fitting residual rmiLearning a regression tree to obtain the mth leaf node region Rmj(ii) a Calculating each region RmjOutput value y ofmi：
(c) Updating a model
Then, after M iterations, the final model f is outputM(x)：
Wherein I () is an indicative function, γmjFor the leaf node region R in the decision tree jmjThe output value of (d);
the performance of the GBRT model is evaluated using a coefficient of determination (R) between predicted and measured values2) Mean Square Error (MSE) and Mean Absolute Error (MAE) to evaluate the performance of the model. R2MSE, MAE is calculated as follows:
where m is the number of samples,
is the true value, yiIs a predicted value of the number of the frames,
is the predicted mean. R2Is a percentage-based indicator, ideally R2As close to 100% or 1 as possible. The MAE can better reflect the actual situation of the error of the predicted value. MSE reflects the degree of difference between the predicted value and the true value, and the smaller the values of MAE and MSE, the better the prediction capability of the model.
In order to ensure the strong generalization ability of the model, the deviation and the variance of the model are coordinated with each other by adjusting the parameters of the model, so that the accuracy of the model is improved. First, no parameter is set, and default values are adopted. Then, after grid search is carried out on the key parameters, the obtained optimized protein model parameter combination is that the maximum depth (max _ depth) of the tree is 14, the iteration times (n _ estimators) are 90, and the learning rate (learning _ rate) is 0.11; the fat model has a parameter combination max _ depth of 11, n _ estimators of 110, and a learning _ rate of 0.1. The overall predicted effect pairs of GBRT before and after parameter optimization are shown in table 1.
TABLE 1
As can be seen from table 1 above, after the parameters are optimized, although the goodness of fit of the fat model is improved a little and is only improved by 0.8%, MSE and MAE are respectively reduced by 69.68% and 80.17%; protein model R on the basis of original2The improvement is 19.11 percent, the MSE is reduced by 81.80 percent, and the MAE is reduced by 85.97 percent. The result shows that after the key parameters of the GBRT are optimized, the mean square error and the mean absolute error of the prediction result can be greatly reduced on the basis of keeping or slightly improving the original goodness of fit, so that the overall prediction precision is improved.
In this example, in order to evaluate the GBRT model performance, the present invention selects different machine learning methods (SVR, BPNN, RNN, and KNN) as comparison experiments in the regression prediction part. The evaluation indexes of the obtained milk measurement model are shown in table 2.
TABLE 2
As can be seen from table 2, the GBRT method proposed by the present invention has a higher goodness of fit and a lowest root mean square error compared to the other 4 conventional methods, which indicates that the method proposed herein can indeed obtain better prediction accuracy compared to the other conventional methods.
After determining the parameters of the fat and protein prediction model, random sampling was repeated 100 times on the data set of the model in order to demonstrate the effect of the model under various training and validation data. The training results are shown in fig. 2. The predicted value is a measurement result obtained by inputting the acquired wavelength characteristics of the milk to be measured into the trained GBRT model; the observed values are values of protein (GB 5009.5-2010) and fat (GB 5009.6-2016) content of milk measured by national standards. In 100 protein training models (FIG. 2(a)), the R between the predicted value and the true value was evaluated comprehensively20.9875, Root Mean Square Error (RMSE) 0.0695, MAE 0.0165, i.e., predicted and actual values 0.0165g/100ml apart, and Symmetric Mean Absolute Percent Error (SMAPE) 0.493%. Due to the random selection of the data set and the randomness of the neural network, the performance index of the neural network is changing constantly, but the error is not more than 1%. In 100 training sessions of the fat model (see FIG. 2(b)), the index R is determined20.9994, RMSE 0.095, MAE 0.0236 and SMAPE 1.1632%. Compared with a protein model, the fat model has better evaluation indexes in all aspects and belongs to an excellent prediction model.
In a five-fold cross validation experiment, a data set is randomly divided into 5 subsets with approximately equal sample number, 4 of the subsets are combined into a training set in turn, the rest 1 of the subsets are used as a test set, and finally after k times of experiments, the average value of evaluation indexes is used as the generalization ability of the model. Table 3 shows the performance index results of 5-fold cross validation, which further confirms the reliability of the data set and the model. The five-fold cross validation obtains effective information as much as possible from the limited data set, so that the overfitting risk of the model can be relieved, the local minimum value can be effectively avoided, and the highest utilization rate of the sample is reflected.
TABLE 3
The five-fold cross validation index of the protein and fat model is within the variation range of the repeated validation index, thereby proving the validity and reliability of the model and the data set.
The test set data was input into the parameter adjusted milk measurement model, and the results are shown in fig. 3. In this example, the predicted Root Mean Square Error (RMSE) of the protein model is 0.00069, the Mean Absolute Error (MAE) is 0.00034, and according to fig. 3(a), the mean measured error of all samples is within ± 0.0043g/100ml, wherein the maximum error is 0.0042g/100ml, the error of more than about 88% is within 0.001g/100ml, and most of the errors are concentrated around 0, which indicates that the protein model has strong accuracy for strange data of samples involved in training. The fat model had an RMSE of 0.00004 and an MAE of 0.00003. In FIG. 3(b), where the maximum error in the sample is 0.00008g/100ml, the fat model has a high generalization ability in terms of the predicted indices and results.
The invention provides a method suitable for predicting the content of main components of milk in multiple scenes based on the combination of a multispectral sensor and a GBRT algorithm. The multispectral acquisition system is used for acquiring the wavelength intensity of the milk, various machine learning models are compared through data processing, and the most appropriate model is selected to realize the measurement of the content of the milk component. By the established protein and fat detection method, chemical pollution can be reduced during measurement, dependence on complex operation during measurement is reduced, and rapid non-destructive prediction of a milk sample is realized. In addition, the multispectral acquisition system has high transportability, can be distributed on a milk component content measurement site in a pasture, and can rapidly measure the milk component content in real time.
The final result shows that the performance effect of the milk component content prediction model combining the multi-wavelength spectrum data with the GBRT algorithm is good, and the errors of the actually measured data of the protein and the fat of the milk are less than 1%. Therefore, the measurement model can provide a better reference for the measurement of the content of the milk component by combining the multispectral sensor, is beneficial to improving the speed and the precision of detection, and can also be applied to the measurement of other substances.
The above-described embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solutions of the present invention can be made by those skilled in the art without departing from the spirit of the present invention, and the technical solutions of the present invention are within the scope of the present invention defined by the claims.
Claims (8)
1. The method for measuring the main components of the milk based on the infrared multispectral sensor is characterized by comprising the following steps:
acquiring multi-wavelength light intensity characteristic data of different types of milk based on an infrared multispectral sensor, and measuring the contents of protein and fat in the milk sample by combining a standard method;
taking the multi-wavelength light intensity characteristic data as input, taking the protein and fat content as output, constructing a milk main component measurement model, training and perfecting model parameters;
and inputting the multi-wavelength light intensity characteristic data of the milk sample to be measured into the measurement model, and obtaining the measurement result by the output protein and fat contents.
2. The method for measuring the milk main component based on the infrared multispectral sensor as recited in claim 1, wherein the step of establishing the milk main component measurement model comprises the following steps:
acquiring characteristic data, and acquiring multi-wavelength light intensity data of a broadband near-infrared light source penetrating through a milk sample by using a multi-wavelength spectrum sensor to serve as the characteristic data of a measurement model;
obtaining label data, and measuring the content of protein and fat in the milk sample by using a standard method as the label data of a measurement model;
dividing a data set, preprocessing the multi-wavelength light intensity data, and dividing the preprocessed multi-wavelength light intensity data into a training set and a test set;
training a model, namely taking characteristic data in a training set as input of the measurement model, taking label data in the training set as output, training the measurement model by using a leave-out method, and detecting the stability of the measurement model by using repeated verification and five-fold cross verification;
establishing a measurement model, and measuring the protein and fat contents of the samples in the test set through the measurement model;
and evaluating the measurement model, carrying out grid search, and determining final parameters to obtain the milk main component measurement model.
3. The infrared multispectral sensor-based milk principal component measurement method according to claim 2, wherein the milk principal component measurement model is evaluated based on a decision coefficient, a mean absolute error, a mean square error, and a symmetric mean absolute percentage as evaluation indexes.
4. The method for measuring the main components of the milk based on the infrared multispectral sensor as claimed in claim 3, wherein the multi-wavelength light intensity data of the sample to be measured is obtained by the infrared multispectral sensor and is input into the trained measurement model to obtain the protein and fat contents of the milk sample.
5. The method for measuring the milk main component based on the infrared multispectral sensor AS claimed in claim 4, wherein the infrared multispectral sensor is AS7263, the infrared multispectral sensor is provided with six near-infrared channels, and the wavelengths of the six near-infrared channels are 610nm, 680nm, 730nm, 760nm, 810nm and 860nm respectively.
6. The method for measuring the main components of the milk based on the infrared multispectral sensor as recited in claim 1, wherein the data acquisition device comprises: the constant light intensity driving circuit comprises a constant light intensity driving circuit, a light source, a heat dissipation plate, a sample groove and an infrared multispectral sensor, wherein the light source is a broadband near-infrared LED and is used for enabling emitted broadband infrared light to be incident into a milk sample and absorbed by protein and fat particles, and then the multi-wavelength light intensity characteristics of the milk sample are obtained through the infrared multispectral sensor.
7. The infrared multispectral sensor-based milk principal component measurement method according to claim 2, wherein the milk principal component measurement model is a gradient boosting regression tree model (GBRT).
8. The method for measuring the main components of the milk based on the infrared multispectral sensor as claimed in claim 1, wherein the measurement of the protein and fat contents in the milk sample based on a standard method is used for making label data of the milk measurement model, and the measurement process comprises the following steps: and (3) determining the content of fat in the milk sample according to a Gerber method, and determining the content of protein according to a Kjeldahl method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210066510.3A CN114414521A (en) | 2022-01-20 | 2022-01-20 | Milk main component measuring method based on infrared multispectral sensor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210066510.3A CN114414521A (en) | 2022-01-20 | 2022-01-20 | Milk main component measuring method based on infrared multispectral sensor |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114414521A true CN114414521A (en) | 2022-04-29 |
Family
ID=81275455
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210066510.3A Pending CN114414521A (en) | 2022-01-20 | 2022-01-20 | Milk main component measuring method based on infrared multispectral sensor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114414521A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117470801A (en) * | 2023-11-02 | 2024-01-30 | 上海贝高医疗科技有限公司 | Breast milk component rapid analysis device based on multispectral simultaneous detection |
| CN117723514A (en) * | 2024-02-07 | 2024-03-19 | 山西品东智能控制有限公司 | Multi-index coal quality analyzer based on sodium light source and intelligent algorithm and detection method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108572155A (en) * | 2017-03-08 | 2018-09-25 | 康师傅饮品控股有限公司 | Protein, fatty near-infrared quantitative system and its method for building up in milk beverage imitating flavour of coffee |
| CN109726700A (en) * | 2019-01-07 | 2019-05-07 | 武汉南博网络科技有限公司 | A kind of identifying pest method for early warning and device based on multiple features |
| CN111781154A (en) * | 2020-08-10 | 2020-10-16 | 西北农林科技大学 | Low-cost milk component analysis method and device based on multispectral sensor |
| US20210172800A1 (en) * | 2019-12-10 | 2021-06-10 | Perkinelmer Health Sciences Canada, Inc. | Systems and Methods for Analyzing Unknown Sample Compositions Using a Prediction Model Based On Optical Emission Spectra |
| CN113607676A (en) * | 2021-08-04 | 2021-11-05 | 淮北师范大学 | Milk component rapid detection method based on special-shaped cuvette and near-infrared image |
-
2022
- 2022-01-20 CN CN202210066510.3A patent/CN114414521A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108572155A (en) * | 2017-03-08 | 2018-09-25 | 康师傅饮品控股有限公司 | Protein, fatty near-infrared quantitative system and its method for building up in milk beverage imitating flavour of coffee |
| CN109726700A (en) * | 2019-01-07 | 2019-05-07 | 武汉南博网络科技有限公司 | A kind of identifying pest method for early warning and device based on multiple features |
| US20210172800A1 (en) * | 2019-12-10 | 2021-06-10 | Perkinelmer Health Sciences Canada, Inc. | Systems and Methods for Analyzing Unknown Sample Compositions Using a Prediction Model Based On Optical Emission Spectra |
| CN111781154A (en) * | 2020-08-10 | 2020-10-16 | 西北农林科技大学 | Low-cost milk component analysis method and device based on multispectral sensor |
| CN113607676A (en) * | 2021-08-04 | 2021-11-05 | 淮北师范大学 | Milk component rapid detection method based on special-shaped cuvette and near-infrared image |
Non-Patent Citations (1)
| Title |
|---|
| 周真 等: "基于多波长的牛奶成分光谱测量方法的研究", 《哈尔滨理工大学学报》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117470801A (en) * | 2023-11-02 | 2024-01-30 | 上海贝高医疗科技有限公司 | Breast milk component rapid analysis device based on multispectral simultaneous detection |
| CN117470801B (en) * | 2023-11-02 | 2024-04-12 | 上海贝高医疗科技有限公司 | Breast milk component rapid analysis device based on multispectral simultaneous detection |
| CN117723514A (en) * | 2024-02-07 | 2024-03-19 | 山西品东智能控制有限公司 | Multi-index coal quality analyzer based on sodium light source and intelligent algorithm and detection method |
| CN117723514B (en) * | 2024-02-07 | 2024-05-17 | 山西品东智能控制有限公司 | Multi-index coal quality analyzer based on sodium light source and intelligent algorithm and detection method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114414521A (en) | Milk main component measuring method based on infrared multispectral sensor | |
| CN108875913A (en) | A kind of matsutake Fast nondestructive evaluation system and method based on convolutional neural networks | |
| CN109540836A (en) | Near infrared spectrum pol detection method and system based on BP artificial neural network | |
| CN106841083A (en) | Sesame oil quality detecting method based on near-infrared spectrum technique | |
| CN108846203A (en) | The method and device of fruit non-destructive testing | |
| CN106596456A (en) | Changeable moving window based selection method of near infrared spectral region of solution | |
| CN106950192A (en) | A kind of method of Contents of Main Components quick detection in vegetable protein beverage based on near-infrared spectral analysis technology | |
| CN109187443A (en) | Water body bacterial micro-organism based on multi-wavelength transmitted spectrum accurately identifies method | |
| CN115993344A (en) | Quality monitoring and analyzing system and method for near infrared spectrum analyzer | |
| Sheng et al. | Analysis of protein and fat in milk using multiwavelength gradient-boosted regression tree | |
| CN109001182B (en) | Raman spectrum nondestructive testing method for alcohol content in closed container | |
| Chang et al. | Detection of sugar content in food based on the electrochemical method with the assistance of partial least square method and deep learning | |
| Qi et al. | Rapid and non-destructive determination of soluble solid content of crown pear by visible/near-infrared spectroscopy with deep learning regression | |
| CN113267466B (en) | Fruit sugar degree and acidity nondestructive testing method based on spectral wavelength optimization | |
| CN113324940A (en) | Spectrum grading method for super-high-quality milk, high-protein special milk, high-milk-fat special milk and common milk | |
| CN105259136A (en) | Measuring-point-free temperature correction method of near-infrared correction model | |
| Zhou et al. | Detection of chemical oxygen demand in water based on UV absorption spectroscopy and PSO-LSSVM algorithm | |
| CN116310788A (en) | Plant leaf nitrogen content hyperspectral modeling method based on map fusion | |
| CN113916817B (en) | Spectrum method chromaticity online measurement method for urban living drinking water | |
| CN113252624B (en) | Apple flavone content nondestructive testing method based on fluorescence spectrum | |
| CN113418889B (en) | Real-time detection method for water content and total colony count of dried vegetables based on deep learning | |
| CN104374740A (en) | Detection method of bee product | |
| CN113933247A (en) | Construction method of fruit maturity detection model | |
| Fu et al. | Nondestructive prediction modeling of S‐ovalbumin content in stored eggs based on hyperspectral fusion information | |
| García-Jares et al. | Prediction of some physico-chemical parameters in red wines from ultraviolet–visible spectra using a partial least-squares model in latent variables |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220429 |