CN115468931A - Construction method of soybean meal quality detection model, and soybean meal quality detection method and device - Google Patents
Construction method of soybean meal quality detection model, and soybean meal quality detection method and device Download PDFInfo
- Publication number
- CN115468931A CN115468931A CN202211360269.1A CN202211360269A CN115468931A CN 115468931 A CN115468931 A CN 115468931A CN 202211360269 A CN202211360269 A CN 202211360269A CN 115468931 A CN115468931 A CN 115468931A
- Authority
- CN
- China
- Prior art keywords
- soybean meal
- sample
- model
- bean pulp
- moisture
- 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
- 235000019764 Soybean Meal Nutrition 0.000 title claims abstract description 379
- 239000004455 soybean meal Substances 0.000 title claims abstract description 379
- 238000001514 detection method Methods 0.000 title claims abstract description 88
- 238000010276 construction Methods 0.000 title claims abstract description 45
- 235000019750 Crude protein Nutrition 0.000 claims abstract description 114
- 244000046052 Phaseolus vulgaris Species 0.000 claims description 208
- 235000010627 Phaseolus vulgaris Nutrition 0.000 claims description 208
- 238000002329 infrared spectrum Methods 0.000 claims description 91
- 238000012937 correction Methods 0.000 claims description 90
- 238000000034 method Methods 0.000 claims description 66
- 238000010561 standard procedure Methods 0.000 claims description 33
- 238000012795 verification Methods 0.000 claims description 26
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 24
- 238000005259 measurement Methods 0.000 claims description 22
- 230000008569 process Effects 0.000 claims description 14
- 238000007781 pre-processing Methods 0.000 claims description 11
- 230000003595 spectral effect Effects 0.000 claims description 10
- 238000000513 principal component analysis Methods 0.000 claims description 9
- 102000004169 proteins and genes Human genes 0.000 claims description 7
- 108090000623 proteins and genes Proteins 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 230000009467 reduction Effects 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 3
- 238000012847 principal component analysis method Methods 0.000 claims description 2
- 239000003153 chemical reaction reagent Substances 0.000 abstract description 6
- 238000012360 testing method Methods 0.000 description 8
- 235000012054 meals Nutrition 0.000 description 7
- 239000002994 raw material Substances 0.000 description 4
- 238000007696 Kjeldahl method Methods 0.000 description 3
- 235000010469 Glycine max Nutrition 0.000 description 2
- 238000002835 absorbance Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 239000002537 cosmetic Substances 0.000 description 2
- 238000002790 cross-validation Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 235000013305 food Nutrition 0.000 description 2
- 235000013402 health food Nutrition 0.000 description 2
- 230000008676 import Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 244000144972 livestock Species 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 244000144977 poultry Species 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 235000019624 protein content Nutrition 0.000 description 2
- 238000004451 qualitative analysis Methods 0.000 description 2
- 238000011002 quantification Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000012031 short term test Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000004497 NIR spectroscopy Methods 0.000 description 1
- 239000003242 anti bacterial agent Substances 0.000 description 1
- 229940088710 antibiotic agent Drugs 0.000 description 1
- 230000003115 biocidal effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 229920000515 polycarbonate Polymers 0.000 description 1
- 239000004417 polycarbonate Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 210000004767 rumen Anatomy 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000004580 weight loss Effects 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
- G01N21/359—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
- 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
- G01N21/3554—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for determining moisture content
-
- 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
- G01N21/3563—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing solids; Preparation of samples therefor
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P60/00—Technologies relating to agriculture, livestock or agroalimentary industries
- Y02P60/80—Food processing, e.g. use of renewable energies or variable speed drives in handling, conveying or stacking
- Y02P60/87—Re-use of by-products of food processing for fodder production
Landscapes
- Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The application provides a construction method of a soybean meal quality detection model, a soybean meal quality detection method and a soybean meal quality detection device. The construction method of the soybean meal quality detection model comprises the following steps: collecting and dividing samples, acquiring construction model data, constructing a qualitative discrimination model, constructing a soybean meal moisture quantitative model, constructing a soybean meal crude protein quantitative model, and constructing the soybean meal quality detection model by the qualitative discrimination model, the soybean meal moisture quantitative model and the soybean meal crude protein quantitative model. The soybean meal quality detection model constructed by the construction method of the soybean meal quality detection model provided by the first aspect of the application meets the requirements of rapid, efficient and nondestructive detection on the quality of the soybean meal at present, shortens the detection time, avoids the use of chemical reagents for detection, and reduces the detection cost.
Description
Technical Field
The application relates to the technical field of soybean meal detection, in particular to a construction method of a soybean meal quality detection model, a soybean meal quality detection method and a soybean meal quality detection device.
Background
The soybean meal has wide application range, can be used as a main raw material for preparing livestock and poultry feeds, and can also be used as a raw material for preparing cake foods, health foods, cosmetics and antibiotics. The import quantity and the demand quantity of the soybean meal in China are large, so that the quality of the soybean meal is urgently required to be quickly and efficiently detected. However, the existing quality detection of soybean meal is generally a traditional laboratory detection method, such as kjeldahl method, and the traditional detection methods have the disadvantages of long detection time, high detection cost, large consumption of chemical reagents and difficulty in meeting the requirements of efficient and rapid detection.
Therefore, a new method for constructing a soybean meal quality detection model, a method and an apparatus for detecting soybean meal quality are needed.
Disclosure of Invention
The application provides a construction method of a soybean meal quality detection model, a soybean meal quality detection method and a soybean meal quality detection device.
The first aspect of the application provides a method for constructing a soybean meal quality detection model, which comprises the following steps:
collecting and dividing a sample, wherein a plurality of bean pulp samples are collected to form a bean pulp sample set, the bean pulp sample set comprises unprocessed bean pulp and processed bean pulp, and the bean pulp sample set is divided into a correction set and an independent verification set, wherein all the bean pulp samples in the correction set are unprocessed bean pulp, and the independent verification set comprises the unprocessed bean pulp and the processed bean pulp;
acquiring construction model data, wherein the construction model data comprises near infrared spectrum information of each soybean meal sample in a soybean meal sample set, a crude protein national standard method measurement value and a water national standard method measurement value;
establishing a qualitative discrimination model, carrying out centralized processing on near infrared spectrum information of each bean pulp sample in a correction set to obtain a qualitative discrimination model center, calculating the Mahalanobis distance from the near infrared spectrum information of each bean pulp sample in the correction set to the qualitative discrimination model center, and selecting a Mahalanobis distance threshold value according to a preset confidence level to obtain the qualitative discrimination model for discriminating whether the bean pulp sample is unprocessed bean pulp;
constructing a bean pulp moisture quantitative model, taking near infrared spectrum information of each bean pulp sample in a correction set as the input of the bean pulp moisture quantitative model, taking a moisture national standard method measured value of each bean pulp sample in the correction set as the output of the bean pulp moisture quantitative model, and constructing the bean pulp moisture quantitative model by using a partial least square method;
the method comprises the steps of constructing a soybean meal crude protein quantitative model, taking near infrared spectrum information of each corrected and concentrated soybean meal sample as input of the soybean meal crude protein quantitative model, taking a national standard method measured value of the crude protein of each corrected and concentrated soybean meal sample as output of the soybean meal crude protein quantitative model, constructing the soybean meal crude protein quantitative model by using a partial least square method, and constructing the soybean meal quality detection model by using a qualitative judgment model, the soybean meal moisture quantitative model and the soybean meal crude protein quantitative model.
According to the construction method of the soybean meal quality detection model provided by the first aspect of the application, the soybean meal quality detection model comprising the qualitative judgment model, the soybean meal moisture quantitative model and the soybean meal crude protein quantitative model is efficiently and accurately constructed on the basis of the near infrared spectrum analysis technology, and the qualitative judgment model in the soybean meal quality detection model can realize the qualitative analysis on whether a soybean meal sample is subjected to special process treatment, so that whether the soybean meal sample is common soybean meal (i.e. untreated soybean meal) or special soybean meal (i.e. process-treated soybean meal) can be quickly judged. The soybean meal moisture quantitative model is used for rapidly and quantitatively measuring the moisture content in the soybean meal sample so as to help judge whether the soybean meal sample is a qualified sample. The soybean meal crude protein quantitative model is used for rapidly and quantitatively measuring the crude protein content in the soybean meal sample so as to help to evaluate the quality grade of the qualified soybean meal sample. The soybean meal quality detection model constructed by the construction method of the soybean meal quality detection model provided by the first aspect of the application meets the requirements of rapid, efficient and nondestructive detection on the quality of soybean meal at present, shortens the detection time, avoids the use of chemical reagents for detection, and reduces the detection cost.
In a second aspect, the present application provides a method for detecting quality of soybean meal, comprising:
obtaining near infrared spectrum information of a soybean meal sample to be detected;
calculating a first Mahalanobis distance from the near infrared spectrum information of the bean pulp sample to be detected to the center of a qualitative judgment model of the qualitative judgment model, judging the bean pulp sample to be detected to be the processed bean pulp when the first Mahalanobis distance is greater than or equal to a Mahalanobis distance threshold value, and judging the bean pulp sample to be detected to be the unprocessed bean pulp when the first Mahalanobis distance is less than the Mahalanobis distance threshold value;
inputting the near infrared spectrum information of the bean pulp sample to be detected, which is judged to be unprocessed bean pulp, into a bean pulp moisture quantitative model, obtaining the moisture content of the bean pulp sample to be detected, which is output by the bean pulp moisture quantitative model, judging that the bean pulp sample to be detected is a sample with unqualified moisture content when the moisture of the bean pulp sample to be detected is greater than or equal to a preset moisture threshold value, and judging that the bean pulp sample to be detected is a sample with qualified moisture content when the moisture of the bean pulp sample to be detected is less than the preset moisture threshold value;
inputting the near infrared spectrum information of the soybean meal sample to be detected, which is judged to be a qualified sample of the moisture content, into a soybean meal crude protein quantitative model, obtaining the crude protein content of the soybean meal sample to be detected, which is output by the soybean meal moisture quantitative model, and evaluating the quality grade of the soybean meal sample to be detected according to a preset crude protein content standard, wherein the qualitative judgment model, the soybean meal moisture quantitative model and the soybean meal crude protein quantitative model are obtained by pre-constructing according to the construction method provided by the first aspect of the application.
The soybean meal quality detection method provided by the second aspect of the application realizes efficient, rapid, accurate and nondestructive detection on the quality of the soybean meal, and is low in detection cost.
The third aspect of the present application provides an apparatus for constructing a soybean meal quality detection model, the apparatus comprising:
the device comprises a first acquisition unit, a second acquisition unit and a third acquisition unit, wherein the first acquisition unit is used for acquiring construction model data, the construction model data comprises near infrared spectrum information, a crude protein national standard method measurement value and a water national standard method measurement value of each soybean meal sample in a soybean meal sample set, and the soybean meal sample set comprises untreated soybean meal and process-treated soybean meal;
the model building unit is used for dividing the bean pulp sample set into a correction set and an independent verification set, wherein all bean pulp samples in the correction set are unprocessed bean pulp, the independent verification set comprises unprocessed bean pulp and processed bean pulp, the model building unit is used for building a qualitative discrimination model, the near infrared spectrum information of each bean pulp sample in the correction set is processed in a centralized mode to obtain a qualitative discrimination model center, the Mahalanobis distance from the near infrared spectrum information of each bean pulp sample in the correction set to the qualitative discrimination model center is calculated, and the Mahalanobis distance threshold value is selected according to the preset reliability level to obtain the qualitative discrimination model for discriminating whether the bean pulp sample is unprocessed bean pulp;
the model construction unit is used for constructing a bean pulp moisture quantitative model, taking near infrared spectrum information of each bean pulp sample in the correction set as the input of the bean pulp moisture quantitative model, taking a moisture national standard method measured value of each bean pulp sample in the correction set as the output of the bean pulp moisture quantitative model, and constructing the bean pulp moisture quantitative model by using a partial least square method;
the model construction unit is also used for constructing a soybean meal crude protein quantitative model, the near infrared spectrum information of each soybean meal sample in correction set is used as the input of the soybean meal crude protein quantitative model, the national standard method measurement value of the crude protein of each soybean meal sample in correction set is used as the output of the soybean meal crude protein quantitative model, the partial least square method is used for constructing the soybean meal crude protein quantitative model, and the qualitative judgment model, the soybean meal moisture quantitative model and the soybean meal crude protein quantitative model jointly form the soybean meal quality detection model.
The fourth aspect of this application provides a device of short-term test dregs of beans quality, and the device includes:
the second acquisition unit is used for acquiring the near infrared spectrum information of the soybean meal sample to be detected;
the rapid measuring unit of the bean pulp quality is used for inputting the near infrared spectrum information of the bean pulp sample to be measured, which is determined as unprocessed bean pulp, into the bean pulp moisture quantitative model to obtain the moisture content of the bean pulp sample to be measured, which is output by the bean pulp quantitative model, when the moisture of the bean pulp sample to be measured is greater than or equal to a preset moisture threshold value, the bean pulp sample to be measured is determined as a sample with unqualified moisture content, and when the moisture of the bean pulp sample to be measured is less than the preset moisture threshold value, the bean pulp sample to be measured is determined as a sample with qualified moisture content; the rapid determination unit of dregs of beans quality still is used for judging as moisture content qualified sample the near infrared spectroscopy information input dregs of beans coarse protein quantitative model of the dregs of beans sample that awaits measuring, acquires dregs of beans moisture quantitative model output the coarse protein content of the dregs of beans sample that awaits measuring to divide according to predetermineeing coarse protein content standard the quality grade of the dregs of beans sample that awaits measuring, wherein, qualitative discrimination model dregs of beans moisture quantitative model and dregs of beans coarse protein quantitative model is this application third aspect the device of constructing dregs of beans quality detection model establish in advance and obtain.
Drawings
Fig. 1 is a schematic flow chart of an embodiment of a method for constructing a soybean meal quality detection model according to the first aspect of the present disclosure;
fig. 2 is a schematic flow chart of step S30 in the method for constructing the soybean meal quality detection model according to the first aspect of the present application;
fig. 3 is a schematic flow chart of step S40 in the method for constructing the soybean meal quality detection model according to the first aspect of the present application;
fig. 4 is a schematic flow chart of step S50 in the method for constructing the soybean meal quality detection model according to the first aspect of the present application;
fig. 5 is a schematic flow chart of another embodiment of the method for constructing the soybean meal quality detection model according to the first aspect of the present application;
fig. 6 is a near-infrared spectrum of each soybean meal sample set in the soybean meal sample set in example 1 of the method for constructing a soybean meal quality detection model according to the first aspect of the present application;
FIG. 7 is a schematic view of Mahalanobis distance in two-dimensional coordinates in embodiment 1 of the first aspect of the present application;
FIG. 8 is a water content scatter plot from the national standard measurement of moisture in soybean meal and the output of a model for quantifying moisture in soybean meal, for a calibrated concentrated soybean meal sample according to example 1 of the first aspect of the present application;
FIG. 9 is a scattergram of the crude protein content outputted from the national Standard test value of the soybean meal crude protein and the quantitative model of the soybean meal crude protein of the corrected and concentrated soybean meal sample according to example 1 of the first aspect of the present application;
FIG. 10 is a water content scattergram of national standard method measurement values of water content and output of a soybean meal water content quantitative model of the remaining 19 soybean meal samples in the independent test set in example 1 of the first aspect of the present application;
FIG. 11 is a scattergram of crude protein content obtained from the national standard measurement of crude protein and the quantitative model of crude protein of soybean meal for the remaining 19 soybean meal samples in the independent test set in example 1 of the first aspect of the present application;
fig. 12 is a schematic flow chart of an embodiment of a method for detecting quality of soybean meal provided by the second aspect of the present application.
Detailed Description
The present application will be described in further detail below with reference to the drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The soybean meal has wide application range, can be used as a main raw material for preparing livestock and poultry feed, and can also be used for preparing cake food, health food, cosmetics and antibiotic raw materials. The import quantity and the demand quantity of the soybean meal in China are large, so that the quality of the soybean meal is urgently required to be rapidly and efficiently detected. However, the existing quality detection of soybean meal is generally a traditional laboratory detection method, such as kjeldahl method, and the traditional detection methods have the disadvantages of long detection time, high detection cost, large consumption of chemical reagents and difficulty in meeting the requirements of efficient and rapid detection.
The inventor researches the technical field of bean pulp detection and finds that general customs and bean pulp importers detect multiple indexes of the bean pulp to detect the quality of the bean pulp, wherein the indexes which can reflect the quality of the bean pulp are moisture and crude protein. The traditional method for detecting the crude protein content of the soybean meal mainly adopts a Kjeldahl method, and the traditional method for detecting the moisture content of the soybean meal mainly adopts a drying weight loss method. The traditional detection method has the problems of long detection time consumption, high cost and large consumption of a large amount of chemical reagents, and can not meet the requirement for quickly and efficiently detecting the quality of the soybean meal.
In view of this, the present application is proposed.
As shown in fig. 1, a first aspect of the present application provides a method for constructing a soybean meal quality detection model, including:
s10: the method comprises the steps of sample collection and division, wherein a plurality of bean pulp samples are collected to form a bean pulp sample set, the bean pulp sample set comprises untreated bean pulp and processed bean pulp, the bean pulp sample set is divided into a correction set and an independent verification set, all the bean pulp samples in the correction set are untreated bean pulp, and the independent verification set comprises the untreated bean pulp and the processed bean pulp;
s20: acquiring construction model data, wherein the construction model data comprises near infrared spectrum information of each soybean meal sample in the soybean meal sample set, a crude protein national standard method measurement value and a water national standard method measurement value;
s30: establishing a qualitative discrimination model, carrying out centralized processing on near infrared spectrum information of each bean pulp sample in a correction set to obtain a qualitative discrimination model center, calculating the Mahalanobis distance from the near infrared spectrum information of each bean pulp sample in the correction set to the qualitative discrimination model center, and selecting a Mahalanobis distance threshold value according to a preset confidence level to obtain the qualitative discrimination model for discriminating whether the bean pulp sample is unprocessed bean pulp;
s40: constructing a bean pulp moisture quantitative model, taking near infrared spectrum information of each bean pulp sample in a correction set as the input of the bean pulp moisture quantitative model, taking a moisture national standard method measured value of each bean pulp sample in the correction set as the output of the bean pulp moisture quantitative model, and constructing the bean pulp moisture quantitative model by using a partial least square method;
s50: the method comprises the steps of constructing a soybean meal crude protein quantitative model, taking near infrared spectrum information of each corrected and concentrated soybean meal sample as input of the soybean meal crude protein quantitative model, taking a national standard method measured value of the crude protein of each corrected and concentrated soybean meal sample as output of the soybean meal crude protein quantitative model, constructing the soybean meal crude protein quantitative model by using a partial least square method, and constructing the soybean meal quality detection model by using a qualitative judgment model, the soybean meal moisture quantitative model and the soybean meal crude protein quantitative model.
According to the construction method of the soybean meal quality detection model provided by the first aspect of the application, the near infrared spectrum analysis technology is used as a basis to efficiently and accurately construct the soybean meal quality detection model comprising the qualitative judgment model, the soybean meal moisture quantitative model and the soybean meal crude protein quantitative model, and the qualitative judgment model in the soybean meal quality detection model can realize qualitative analysis on whether a soybean meal sample is subjected to special process treatment, so that whether the soybean meal sample is ordinary soybean meal (i.e. unprocessed soybean meal) or special soybean meal (i.e. processed soybean meal) can be quickly judged. The soybean meal moisture quantitative model is used for rapidly and quantitatively measuring the moisture content in the soybean meal sample so as to help judge whether the soybean meal sample is a qualified sample. The soybean meal crude protein quantitative model is used for rapidly and quantitatively measuring the crude protein content in the soybean meal sample so as to help to evaluate the quality grade of the qualified soybean meal sample. The soybean meal quality detection model constructed by the construction method of the soybean meal quality detection model provided by the first aspect of the application meets the requirements of rapid, efficient and nondestructive detection on the quality of the soybean meal at present, shortens the detection time, avoids the use of chemical reagents for detection, and reduces the detection cost.
In some optional embodiments of the first aspect of the present application, in the method for constructing the soybean meal quality detection model, there is no specific sequence between step S30, step S40, and step S50.
In some optional embodiments of the first aspect of the present application, in the step S20, the national standards "method for determining crude protein in feed" (GB/T6432-2018) and "method for determining water in feed" (GB/T6435-2014) are respectively used in obtaining the data of the constructed model to obtain the national standard value of crude protein and the national standard value of water for each soybean meal sample in each soybean meal sample set.
In some alternative embodiments of the first aspect of the present application, the near infrared spectral information of each individual sample of the collection of meal samples is obtained by a Brookfield near infrared spectrometer (MPA)
) Obtained with a near infrared spectral resolution of 8cm -1 The wave number range of the near infrared spectrum is 11536 cm -1 ~3952 cm -1 The near infrared spectral information for each meal sample included 949 spectral data points, with each photovoltaic data point consisting of both (wave number, absorbance).
In some alternative embodiments of the first aspect of the present application, the untreated soy meal of the first aspect of the present application is regular soy meal, which has not been subjected to a special process. The process-treated soybean meal includes special soybean meal such as rumen soybean meal and fermented soybean meal, which should not be understood as ordinary soybean meal.
In some optional embodiments of the first aspect of the present application, in the step of collecting and dividing the sample, a principal component analysis method is used to pre-divide the sample set of the soybean meal into a preliminary correction set and a preliminary independent verification set;
removing the process-treated soybean meal in the preliminary correction set to form a correction set;
and combining the process-treated soybean meal removed from the preparation correction set and the preparation independent verification set into an independent verification set.
In some examples of these embodiments, a Principal Component Analysis (PCA) method of bruker (OPUS) software was used to divide the soybean meal sample set into a preparatory correction set and a preparatory independent validation set.
In some optional embodiments of the first aspect of the present application, as shown in fig. 2, in the step S30 of constructing the qualitative judgment model, the step of centering the near infrared spectrum information of each of the corrected and concentrated soybean meal samples includes:
s31: preprocessing the near infrared spectrum information of each bean pulp sample in the correction set by adopting a multivariate scattering correction Method (MSC);
s32: selecting a first wave number range as a wave number range for correcting and concentrating the near infrared spectrum information of each soybean meal sample, wherein the first wave number range is 9400 cm -1 ~5448 cm -1 And 4602 cm -1 ~4246 cm -1 ；
S33: and (3) performing dimensionality reduction treatment on the near infrared spectrum information of each soybean meal sample in the first wave number range in the correction set by adopting Principal Component Analysis (PCA) to obtain a plurality of score data, taking the first 12 score data to enable the variance interpretation rate to be 99%, and obtaining a qualitative judgment model center based on the first 12 score data. In some examples, the average value of the first 12 score data is subtracted from the first 12 score data to obtain 12 processed score data, and then the 12 processed score data are averaged to obtain a coordinate origin, which is also the center of the qualitative discrimination model.
In some optional embodiments of the first aspect of the present application, as shown in fig. 3, the step S40 of constructing a moisture quantitative model of the soybean meal further includes:
s41: preprocessing the near infrared spectrum information of each soybean meal sample in the correction set by adopting a first derivative and multivariate scattering correction;
s42: selecting a second wave number range as a wave number range for correcting and concentrating the near infrared spectrum information of each soybean meal sample, wherein the second wave number range is 9400 cm -1 ~6096 cm -1 And 5454 cm -1 ~4600 cm -1 。
In some of these embodiments, the near infrared spectral information of each of the meal samples in the calibration set is preprocessed using the first derivative, the standard forward-to-theta transform, the multivariate scattering correction, and a combination of at least two of the three methods, and the preprocessed near infrared spectral information is regression modeled using partial least squares with the national moisture standard values after selecting the corresponding second wave number range. And (4) performing regression screening by adopting a leave-one-out method for cross validation, and taking the model with the minimum error as a soybean meal moisture quantitative model. The model with the least error corresponds to a pre-processing method that is a combination of both first derivative and multivariate scatter correction.
In some optional embodiments of the first aspect of the present application, as shown in fig. 4, the step S50 of constructing a quantitative model of crude protein of soybean meal further comprises:
s51: preprocessing the near infrared spectrum information of each bean pulp sample in the correction set by adopting a minimum and maximum normalization method;
s52: selecting a third wave number range as a wave number range for correcting and concentrating the near infrared spectrum information of each soybean meal sample, wherein the third wave number rangeIn the range of 8448 cm -1 ~7496 cm -1 、6104 cm -1 ~5448 cm -1 And 4600 cm -1 ~4248 cm -1 。
In some of these embodiments, the near infrared spectral information of each of the meal samples in the calibration set is preprocessed using the first derivative, the standard forward-to-tai transformation, the multivariate scattering correction, the min-max normalization, and a combination of at least two of the four methods, and the preprocessed near infrared spectral information is regression modeled using partial least squares with the crude protein national standard measurements after selecting the corresponding third range of wavelengths. And the regression screening adopts leave-one-out cross validation, and a model with the minimum error is used as a soybean meal moisture quantitative model. The preprocessing method corresponding to the model with the smallest error is the minimum and maximum normalization method.
As shown in fig. 5, in some optional embodiments of the first aspect of the present application, the method for constructing the soybean meal quality detection model further includes:
s60: and adopting an independent verification set evaluation qualitative judgment model, a soybean meal moisture quantitative model and a soybean meal crude protein quantitative model.
[ example 1 ] A method for producing a polycarbonate
S10: and (2) sample collection and division, wherein 313 soybean meal samples are collected to form a soybean meal sample set, the soybean meal sample set comprises untreated soybean meal and processed soybean meal, and the soybean meal sample set is divided into a correction set and an independent verification set, wherein all the soybean meal samples in the correction set are untreated soybean meal, and the independent verification set comprises the untreated soybean meal and the processed soybean meal.
In the step S10 of sample collection and division, a Principal Component Analysis (PCA) method using bruker (OPUS) software is used to divide the soybean meal sample set into a preliminary correction set and a preliminary independent verification set according to a ratio of 9. The correction set totally comprises 286 soybean meal samples which are all unprocessed soybean meal, the independent verification set totally comprises 27 soybean meal samples, the independent verification set comprises 7 processed soybean meal, and the rest 20 soybean meal are unprocessed soybean meal.
S20: and acquiring construction model data, wherein the construction model data comprises near infrared spectrum information of each bean pulp sample in the bean pulp sample set, a crude protein national standard method measured value and a water national standard method measured value.
Step S20, obtaining the data of the constructed model, and respectively obtaining the national standard method measured value of crude protein and the national standard method measured value of water of each soybean meal sample in the soybean meal sample set by adopting the national standard 'method for measuring crude protein in feed (GB/T6432-2018) and the' measuring method of water in feed (GB/T6435-2014).
As shown in fig. 6, the near infrared spectrum information of each soybean meal sample in the soybean meal sample set was obtained by a bruker near infrared spectrometer (MPA), and the wave number range of the near infrared spectrum was 11536 -1 ~3952 cm -1 Resolution of 8cm -1 The near infrared spectral information for each meal sample contained 949 data points, each consisting of both (wave number, absorbance).
S30: and constructing a qualitative discrimination model, carrying out centralized processing on the near infrared spectrum information of each bean pulp sample in the correction set to obtain a qualitative discrimination model center, calculating the Mahalanobis distance from the near infrared spectrum information of each bean pulp sample in the correction set to the qualitative discrimination model center, and selecting a Mahalanobis distance threshold value according to a preset confidence level to obtain the qualitative discrimination model for discriminating whether the bean pulp sample is unprocessed bean pulp.
S31: preprocessing the near infrared spectrum information of each bean pulp sample in the correction set by adopting a multivariate scattering correction Method (MSC);
s32: selecting a first wave number range as a wave number range for correcting and concentrating the near infrared spectrum information of each soybean meal sample, wherein the first wave number range is 9400 cm -1 ~5448 cm -1 And 4602 cm -1 ~4246 cm -1 ；
S33: and (3) performing dimensionality reduction treatment on the near infrared spectrum information of each soybean meal sample in the first wave number range in the correction set by adopting Principal Component Analysis (PCA) to obtain a plurality of score data, taking the first 12 score data to enable the variance interpretation rate to be 99%, and obtaining a qualitative judgment model center based on the first 12 score data. And respectively subtracting the average value of the first 12 scoring data from the first 12 scoring data to obtain 12 processed scoring data, and averaging the 12 processed scoring data to obtain a coordinate origin, namely a qualitative judgment model center.
In this example 1, the mahalanobis distance threshold (MD-Limit) was set to 0.074, so that the mahalanobis distance from the near infrared spectrum information of the 99.5% soybean meal sample in the calibration set to the center of the qualitative discrimination model was smaller than the mahalanobis distance threshold.
Fig. 7 is a schematic view of mahalanobis distance under two-dimensional coordinates in embodiment 1 of the first aspect of the present application. In the figure, σ in 1 σ,2 σ and 3 σ represents a standard deviation for determining the ellipse radius, and 1 σ,2 σ and 3 σ represent 1 standard deviation, 2 standard deviation and 3 standard deviation, respectively. In fig. 7, point O is the center of the qualitative discrimination model, the euclidean distances between points a and B and point O of the coordinate center (the center of the qualitative discrimination model) are equal, but if the mahalanobis distances between points a and B and point O are calculated, the mahalanobis distance between point a is greater than the mahalanobis distance between point B.
S40: and (3) constructing a bean pulp moisture quantitative model, taking near infrared spectrum information of each bean pulp sample in the correction set as input of the bean pulp moisture quantitative model, taking a national standard method measured value of moisture of each bean pulp sample in the correction set as output of the bean pulp moisture quantitative model, and constructing the bean pulp moisture quantitative model by using a partial least square method.
S41: and (3) preprocessing the near infrared spectrum information of each soybean meal sample in the correction set by adopting a first derivative and multivariate scattering correction, wherein the number of smoothing points processed by the first derivative is 17.
S42: selecting a second wave number range as a wave number range for correcting and concentrating the near infrared spectrum information of each soybean meal sample, wherein the second wave number range is 9400 cm -1 ~6096 cm -1 And 5454 cm -1 ~4600 cm -1 。
Fig. 8 is a plot of the national standard measured moisture content of the soybean meal and the moisture content scatter plot output by the soybean meal moisture quantitative model of the concentrated corrected soybean meal sample according to example 1 of the first aspect of the present application, as shown in fig. 8, in the soybean meal moisture quantitative model: the number of latent variables in the correction set is 11, the corrected root mean square error RMSEE =0.29, and the coefficient R is determined 2 =0.9528, relative analytical error RPD =4.60 cross-test: RMSECV =0.32, coefficient R is determined 2 =0.94, relative analytical error RPD =4.12. In fig. 8, the reference value is the national standard measurement value of the moisture of the soybean meal of the corrected concentrated soybean meal sample, and the predicted value is the moisture content of the corrected concentrated soybean meal sample output by the soybean meal moisture quantitative model.
S50: the method comprises the steps of constructing a soybean meal crude protein quantitative model, taking near infrared spectrum information of each soybean meal sample in correction set as input of the soybean meal crude protein quantitative model, taking a national standard method measured value of the crude protein of each soybean meal sample in correction set as output of the soybean meal crude protein quantitative model, constructing the soybean meal crude protein quantitative model by using a partial least square method, and constructing the soybean meal quality detection model by using a qualitative judgment model, the soybean meal water quantitative model and the soybean meal crude protein quantitative model.
S51: preprocessing the near infrared spectrum information of each bean pulp sample in the correction set by adopting a minimum and maximum normalization method;
s52: selecting a third wave number range as a wave number range for correcting and concentrating the near infrared spectrum information of each soybean meal sample, wherein the third wave number range is 8448 cm -1 ~7496 cm -1 、6104 cm -1 ~5448 cm -1 And 4600 cm -1 ~4248 cm -1 。
Fig. 9 is a scattergram of the national standard measurement values of the soybean meal crude protein of the corrected and concentrated soybean meal samples and the crude protein content output by the soybean meal crude protein quantification model in example 1 of the first aspect of the present application, as shown in fig. 9, in the soybean meal crude protein quantification model: the number of latent variables is 12, the corrected root mean square error RMSEE =0.64, and the coefficient R is determined 2 =0.79, relative analytical error RPD =2.17 cross-test: RMSECV =0.71, coefficient R is determined 2 =0.74, relative analytical error RPD =1.92. In fig. 9, the reference value is the national standard measurement value of the soybean meal crude protein, and the predicted value is the crude protein content of the soybean meal sample in the correction set output by the soybean meal crude protein quantitative model.
S60: and adopting an independent verification set evaluation qualitative judgment model, a soybean meal moisture quantitative model and a soybean meal crude protein quantitative model.
S61: respectively inputting the near infrared spectrum information of 27 independent verification concentrated bean pulp samples into a qualitative judgment model, calculating the mahalanobis distance from the near infrared spectrum information of each bean pulp sample in the independent verification concentrated to the center of the qualitative judgment model, obtaining the result that the mahalanobis distance of 8 bean pulp samples exceeds the mahalanobis distance threshold value, and confirming that 7 bean pulp samples in the 8 bean pulp samples are the bean pulp processed by the process, and the other 1 bean pulp sample is the bean pulp not processed by the process;
s62: inputting the near infrared spectrum information of the remaining 19 soybean meal samples excluding the 8 soybean meal samples into a soybean meal moisture quantitative model to obtain the moisture contents of the 19 soybean meal samples output by the soybean meal moisture quantitative model, namely, to obtain predicted moisture values corresponding to the 19 soybean meal samples output by the soybean meal moisture quantitative model, and obtaining a predicted root mean square error RMSEP =0.333;
s63: and inputting the near infrared spectrum information of the remaining 19 soybean meal samples excluding the 8 soybean meal samples into a soybean meal crude protein quantitative model to obtain the crude protein contents of the 19 soybean meal samples output by the soybean meal crude protein quantitative model, namely to obtain crude protein predicted values corresponding to the 19 soybean meal samples output by the soybean meal crude protein quantitative model, and obtaining a predicted root mean square error RMSEP =0.65.
FIG. 10 is a water content scattergram of national standard method measurement values of water and output of a soybean meal water quantitative model of the remaining 19 soybean meal samples in the independent test set in example 1 of the first aspect of the present application, wherein the predicted Root Mean Square Error (RMSEP) is 0.333 and the determination coefficient (R) is 0.333 2 ) It was 0.76 and the relative analytical error (RPD) was 1.39. In fig. 10, the reference value is the measured value of the moisture national standard method of the remaining 19 soybean meal samples in the independent inspection set, and the predicted value is the moisture content of the remaining 19 soybean meal samples in the independent inspection set output by the soybean meal moisture quantitative model.
FIG. 11 is a scattergram of the crude protein content output from the national Standard measurement of crude protein and the quantitative model of crude protein from the soybean meal for the remaining 19 soybean meal samples in the independent test set in example 1 of the first aspect of the present application, wherein the Root Mean Square Error (RMSEP) is 0.65 and the coefficient of determination (R) is 0.65 2 ) 0.74 and a relative analytical error (RPD) of 2.14. Reference values in FIG. 11The predicted value is the crude protein content of the remaining 19 soybean meal samples in the independent testing set output by the soybean meal crude protein quantitative model.
Further, the prediction Root Mean Square Error (RMSEP) shown in fig. 10 and 11 is used to determine the coefficient (R) 2 ) The model is evaluated against the analytical error (RPD).
As shown in fig. 12, a second aspect of the present application provides a method for detecting quality of soybean meal, including:
obtaining near infrared spectrum information of a soybean meal sample to be detected;
calculating a first Mahalanobis distance from the near infrared spectrum information of the bean pulp sample to be detected to the center of a qualitative judgment model of the qualitative judgment model, judging the bean pulp sample to be detected to be the processed bean pulp when the first Mahalanobis distance is greater than or equal to a Mahalanobis distance threshold value, and judging the bean pulp sample to be detected to be the unprocessed bean pulp when the first Mahalanobis distance is less than the Mahalanobis distance threshold value;
inputting the near infrared spectrum information of the bean pulp sample to be detected, which is judged to be unprocessed bean pulp, into a bean pulp moisture quantitative model, obtaining the moisture content of the bean pulp sample to be detected, which is output by the bean pulp moisture quantitative model, judging that the bean pulp sample to be detected is a sample with unqualified moisture content when the moisture of the bean pulp sample to be detected is greater than or equal to a preset moisture threshold value, and judging that the bean pulp sample to be detected is a sample with qualified moisture content when the moisture of the bean pulp sample to be detected is less than the preset moisture threshold value;
inputting near infrared spectrum information of the to-be-detected soybean meal sample of the sample with qualified water content into the soybean meal crude protein quantitative model, obtaining the crude protein content of the to-be-detected soybean meal sample output by the soybean meal water quantitative model, and evaluating the quality grade of the to-be-detected soybean meal sample according to a preset crude protein content standard, wherein the qualitative judgment model, the soybean meal water quantitative model and the soybean meal crude protein quantitative model are pre-constructed by the construction method in the first aspect of the application.
In some optional embodiments of the second aspect of the present application, the preset moisture threshold is 12.5% by mass of moisture in the soybean meal. When the mass percentage of the moisture in the bean pulp is more than 12.5%, the bean pulp sample is a sample with unqualified moisture content; and when the mass percentage of the moisture in the bean pulp is less than or equal to 12.5%, the bean pulp sample is a qualified sample with qualified moisture content.
The preset crude protein content standard is as follows: the mass percent of the crude protein in the soybean meal is more than or equal to 48 percent, and the soybean meal sample is a special grade product; if the mass percentage of the crude protein in the soybean meal is more than or equal to 46 percent, the soybean meal sample is a first-grade product; the mass percent of the crude protein in the soybean meal is more than or equal to 43 percent, and the soybean meal sample is a second-grade product; and (3) the mass percent of the crude protein in the soybean meal is more than or equal to 41 percent, so that the soybean meal sample is a third-grade product.
Because the mass percentage of the water in the bean pulp is required to be less than or equal to 12.5 percent in the preset crude protein content standard. Therefore, in the process of detecting the quality of the soybean meal, the moisture of the soybean meal is detected firstly, and then the content of the crude protein in the soybean meal is detected.
The soybean meal quality detection method provided by the second aspect of the application realizes efficient, rapid, accurate and nondestructive detection on the quality of the soybean meal, and is low in detection cost.
The third aspect of the present application provides an apparatus for constructing a soybean meal quality detection model, the apparatus comprising:
the device comprises a first acquisition unit, a second acquisition unit and a third acquisition unit, wherein the first acquisition unit is used for acquiring construction model data, the construction model data comprise near infrared spectrum information, crude protein national standard method measurement values and water national standard method measurement values of all bean pulp samples in a bean pulp sample set, and the bean pulp sample set comprises untreated bean pulp and process-treated bean pulp;
the model building unit is used for dividing the bean pulp sample set into a correction set and an independent verification set, all bean pulp samples in the correction set are unprocessed bean pulp, the independent verification set comprises unprocessed bean pulp and processed bean pulp, the model building unit is used for building a qualitative discrimination model, near infrared spectrum information of all the bean pulp samples in the correction set is processed in a centralized mode to obtain a qualitative discrimination model center, the Mahalanobis distance from the near infrared spectrum information of all the bean pulp samples in the correction set to the qualitative discrimination model center is calculated, and the Mahalanobis distance threshold value is selected according to a preset confidence level to obtain the qualitative discrimination model used for discriminating whether the bean pulp samples are unprocessed bean pulp;
the model construction unit is used for constructing a bean pulp moisture quantitative model, taking near infrared spectrum information of each bean pulp sample in the correction set as input of the bean pulp moisture quantitative model, taking a moisture national standard method measured value of each bean pulp sample in the correction set as output of the bean pulp moisture quantitative model, and constructing the bean pulp moisture quantitative model by using a partial least square method;
the model construction unit is also used for constructing a soybean meal crude protein quantitative model, near infrared spectrum information of each soybean meal sample in correction set is used as input of the soybean meal crude protein quantitative model, a crude protein national standard method measured value of each soybean meal sample in correction set is used as output of the soybean meal crude protein quantitative model, the soybean meal crude protein quantitative model is constructed by a partial least square method, and the qualitative judgment model, the soybean meal moisture quantitative model and the soybean meal crude protein quantitative model jointly form the soybean meal quality detection model.
The fourth aspect of this application provides a device of short-term test dregs of beans quality, and the device includes:
the second acquisition unit is used for acquiring the near infrared spectrum information of the soybean meal sample to be detected;
the rapid measuring unit of the bean pulp quality is used for inputting the near infrared spectrum information of the bean pulp sample to be measured, which is determined as unprocessed bean pulp, into the bean pulp moisture quantitative model to obtain the moisture content of the bean pulp sample to be measured, which is output by the bean pulp moisture quantitative model, when the moisture of the bean pulp sample to be measured is greater than or equal to a preset moisture threshold value, the bean pulp sample to be measured is determined as a sample with unqualified moisture content, and when the moisture of the bean pulp sample to be measured is less than the preset moisture threshold value, the bean pulp sample to be measured is determined as a sample with qualified moisture content;
the rapid determination unit for the quality of the soybean meal is further used for inputting near infrared spectrum information of the soybean meal sample to be determined as the qualified sample of the moisture content into the soybean meal crude protein quantitative model, acquiring the crude protein content of the soybean meal sample to be determined output by the soybean meal moisture quantitative model, and dividing the quality grade of the soybean meal sample to be determined according to a preset crude protein content standard, wherein the qualitative determination model, the soybean meal moisture quantitative model and the soybean meal crude protein quantitative model are obtained by pre-construction of the device for constructing the soybean meal quality detection model in the third aspect of the application.
The above-described embodiments should not be construed as limiting the scope of the present application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims (10)
1. A construction method of a soybean meal quality detection model is characterized by comprising the following steps:
the method comprises the steps of sample collection and division, wherein a plurality of bean pulp samples are collected to form a bean pulp sample set, the bean pulp sample set comprises unprocessed bean pulp and processed bean pulp, the bean pulp sample set is divided into a correction set and an independent verification set, all bean pulp samples in the correction set are unprocessed bean pulp, and the independent verification set comprises the unprocessed bean pulp and the processed bean pulp;
acquiring construction model data, wherein the construction model data comprises near infrared spectrum information, crude protein national standard method measurement values and water national standard method measurement values of all the soybean meal samples in the soybean meal sample set;
establishing a qualitative discrimination model, performing centralized processing on the near infrared spectrum information of each bean pulp sample in the correction set to obtain a qualitative discrimination model center, calculating the mahalanobis distance from the near infrared spectrum information of each bean pulp sample in the correction set to the qualitative discrimination model center, and selecting a mahalanobis distance threshold value according to a preset confidence level to obtain the qualitative discrimination model for discriminating whether the bean pulp sample is unprocessed bean pulp;
constructing a bean pulp moisture quantitative model, taking near infrared spectrum information of each bean pulp sample in the correction set as the input of the bean pulp moisture quantitative model, taking the national standard method measured value of the moisture of each bean pulp sample in the correction set as the output of the bean pulp moisture quantitative model, and constructing the bean pulp moisture quantitative model by using a partial least square method;
constructing a quantitative model of the crude protein of the soybean meal, taking the near infrared spectrum information of each soybean meal sample in the correction set as the input of the quantitative model of the crude protein of the soybean meal, taking the national standard method measured value of the crude protein of each soybean meal sample in the correction set as the output of the quantitative model of the crude protein of the soybean meal, constructing the quantitative model of the crude protein of the soybean meal by using a partial least square method,
the qualitative discrimination model, the soybean meal moisture quantitative model and the soybean meal crude protein quantitative model jointly form a soybean meal quality detection model.
2. The construction method according to claim 1, wherein in the step of collecting and dividing the samples, the soybean meal sample set is pre-divided into a preliminary correction set and a preliminary independent verification set by a principal component analysis method;
rejecting the process-treated soybean meal in the preliminary correction set to form the correction set;
merging the process-processed soybean meal removed from the preliminary correction set with the preliminary independent validation set into the independent validation set.
3. The construction method according to claim 1, wherein in the step of obtaining the construction model data, the spectral number range in the near infrared spectrum information of each soybean meal sample is 11536 cm -1 ~3952 cm -1 。
4. The method according to claim 1, wherein in the step of constructing a qualitative judgment model, the step of centering the near infrared spectrum information of each soybean meal sample in the correction set comprises:
preprocessing the near infrared spectrum information of each soybean meal sample in the correction set by adopting a multivariate scattering correction method,
selecting a first wave number range as the wave number range of the near infrared spectrum information of each soybean meal sample in the correction set, wherein the first wave number range is 9400 cm -1 ~5448 cm -1 And 4602 cm -1 ~4246 cm -1 ；
And performing dimensionality reduction processing on near infrared spectrum information of each soybean meal sample in the first wave number range in the correction set by adopting principal component analysis to obtain a plurality of score data, taking the first 12 score data to enable the variance interpretation rate to be 99%, and obtaining the qualitative judgment model center based on the first 12 score data.
5. The construction method according to claim 1, wherein the step of constructing the quantitative model of the moisture of the soybean meal further comprises the steps of:
preprocessing the near infrared spectrum information of each soybean meal sample in the correction set by adopting a first derivative and multivariate scattering correction;
selecting a second wave number range as the wave number range of the near infrared spectrum information of each soybean meal sample in the correction set, wherein the second wave number range is 9400 cm -1 ~6096 cm -1 And 5454 cm -1 ~4600 cm -1 。
6. The construction method according to claim 1, wherein the step of constructing the quantitative model of the crude protein of the soybean meal further comprises:
preprocessing the near infrared spectrum information of each soybean meal sample in the correction set by adopting a minimum and maximum normalization method;
selecting a third wave number range as the wave number range of the near infrared spectrum information of each soybean meal sample in the correction set, wherein the third wave number range is 8448 cm -1 ~7496 cm -1 、6104 cm -1 ~5448 cm -1 And 4600 cm -1 ~4248 cm -1 。
7. The construction method according to claim 1, characterized in that the construction method of the soybean meal quality detection model further comprises:
and evaluating the qualitative judgment model, the soybean meal water quantitative model and the soybean meal crude protein quantitative model by adopting the independent verification set.
8. A soybean meal quality detection method is characterized by comprising the following steps:
obtaining near infrared spectrum information of the soybean meal sample to be detected;
calculating a first Mahalanobis distance from the near infrared spectrum information of the to-be-detected soybean meal sample to a center of a qualitative judgment model of the qualitative judgment model, when the first Mahalanobis distance is greater than or equal to the Mahalanobis distance threshold, judging the to-be-detected soybean meal sample as processed soybean meal, and when the first Mahalanobis distance is smaller than the Mahalanobis distance threshold, judging the to-be-detected soybean meal sample as unprocessed soybean meal;
inputting the near infrared spectrum information of the to-be-detected soybean meal sample which is determined to be unprocessed soybean meal into a soybean meal moisture quantitative model, acquiring the moisture content of the to-be-detected soybean meal sample output by the soybean meal moisture quantitative model, determining that the to-be-detected soybean meal sample is a sample with unqualified moisture content when the moisture of the to-be-detected soybean meal sample is greater than or equal to a preset moisture threshold value, and determining that the to-be-detected soybean meal sample is a sample with qualified moisture content when the moisture of the to-be-detected soybean meal sample is less than the preset moisture threshold value;
inputting the near infrared spectrum information of the soybean meal sample to be detected, which is judged to be a qualified sample of the moisture content, into a soybean meal crude protein quantitative model, obtaining the crude protein content of the soybean meal sample to be detected, which is output by the soybean meal moisture quantitative model, evaluating the quality grade of the soybean meal sample to be detected according to a preset crude protein content standard,
wherein the qualitative discrimination model, the quantitative soybean meal moisture model and the quantitative soybean meal crude protein model are constructed in advance according to the construction method of any one of claims 1 to 7.
9. A device for constructing a soybean meal quality detection model is characterized by comprising:
the device comprises a first acquisition unit, a second acquisition unit and a third acquisition unit, wherein the first acquisition unit is used for acquiring construction model data, the construction model data comprise near infrared spectrum information, crude protein national standard method measurement values and water national standard method measurement values of all bean pulp samples in a bean pulp sample set, and the bean pulp sample set comprises untreated bean pulp and process-treated bean pulp;
the model construction unit is used for dividing the bean pulp sample set into a correction set and an independent verification set, all bean pulp samples in the correction set are unprocessed bean pulp, the independent verification set comprises unprocessed bean pulp and processed bean pulp,
the model building unit is used for building a qualitative discrimination model, performing centralized processing on the near infrared spectrum information of each bean pulp sample in the correction set to obtain a qualitative discrimination model center, calculating the Mahalanobis distance from the near infrared spectrum information of each bean pulp sample in the correction set to the qualitative discrimination model center, and selecting a Mahalanobis distance threshold value according to a preset confidence level to obtain the qualitative discrimination model for discriminating whether the bean pulp sample is unprocessed bean pulp;
the model construction unit is used for constructing a bean pulp moisture quantitative model, near infrared spectrum information of each bean pulp sample in the correction set is used as input of the bean pulp moisture quantitative model, the national standard method measured value of moisture of each bean pulp sample in the correction set is used as output of the bean pulp moisture quantitative model, and the partial least square method is used for constructing the bean pulp moisture quantitative model;
the model construction unit is also used for constructing a quantitative model of the coarse protein of the soybean meal, the near infrared spectrum information of each soybean meal sample in the correction set is used as the input of the quantitative model of the coarse protein of the soybean meal, the national standard method measured value of the coarse protein of each soybean meal sample in the correction set is used as the output of the quantitative model of the coarse protein of the soybean meal, the quantitative model of the coarse protein of the soybean meal is constructed by utilizing a partial least square method,
the qualitative discrimination model, the soybean meal moisture quantitative model and the soybean meal crude protein quantitative model jointly form a soybean meal quality detection model.
10. A device for rapidly detecting the quality of soybean meal is characterized by comprising:
the second acquisition unit is used for acquiring the near infrared spectrum information of the soybean meal sample to be detected;
a rapid determination unit for the quality of the soybean meal, which is used for calculating a first mahalanobis distance from the near infrared spectrum information of the soybean meal sample to be determined to the center of a qualitative judgment model of the qualitative judgment model, determining that the soybean meal sample to be determined is the processed soybean meal when the first mahalanobis distance is greater than or equal to the mahalanobis distance threshold, determining that the soybean meal sample to be determined is the unprocessed soybean meal when the first mahalanobis distance is less than the mahalanobis distance threshold,
the rapid determination unit for the quality of the soybean meal is used for inputting the near infrared spectrum information of the to-be-determined soybean meal sample which is determined as unprocessed soybean meal into a soybean meal moisture quantitative model, acquiring the moisture content of the to-be-determined soybean meal sample output by the soybean meal moisture quantitative model, determining that the to-be-determined soybean meal sample is a sample with unqualified moisture content when the moisture of the to-be-determined soybean meal sample is greater than or equal to a preset moisture threshold value, and determining that the to-be-determined soybean meal sample is a sample with qualified moisture content when the moisture of the to-be-determined soybean meal sample is less than the preset moisture threshold value;
the rapid soybean meal quality determination unit is also used for inputting the near infrared spectrum information of the soybean meal sample to be determined as the qualified sample of the moisture content into a soybean meal crude protein quantitative model, acquiring the crude protein content of the soybean meal sample to be determined output by the soybean meal moisture quantitative model, dividing the quality grade of the soybean meal sample to be determined according to a preset crude protein content standard,
wherein the qualitative discrimination model, the quantitative soybean meal moisture model and the quantitative soybean meal crude protein model are pre-constructed by the apparatus for constructing a soybean meal quality detection model according to claim 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211360269.1A CN115468931B (en) | 2022-11-02 | 2022-11-02 | Construction method of soybean meal quality detection model, and soybean meal quality detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211360269.1A CN115468931B (en) | 2022-11-02 | 2022-11-02 | Construction method of soybean meal quality detection model, and soybean meal quality detection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115468931A true CN115468931A (en) | 2022-12-13 |
| CN115468931B CN115468931B (en) | 2023-04-07 |
Family
ID=84336142
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211360269.1A Active CN115468931B (en) | 2022-11-02 | 2022-11-02 | Construction method of soybean meal quality detection model, and soybean meal quality detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115468931B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110308114A (en) * | 2019-07-31 | 2019-10-08 | 太仓安佑生物科技有限公司 | A kind of near infrared detection method of quick identification dregs of beans degree of raw and cooked |
| CN114518339A (en) * | 2022-02-23 | 2022-05-20 | 新希望六和股份有限公司 | Method for establishing near-infrared prediction model of wet-based fermented soybean meal and application |
-
2022
- 2022-11-02 CN CN202211360269.1A patent/CN115468931B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110308114A (en) * | 2019-07-31 | 2019-10-08 | 太仓安佑生物科技有限公司 | A kind of near infrared detection method of quick identification dregs of beans degree of raw and cooked |
| CN114518339A (en) * | 2022-02-23 | 2022-05-20 | 新希望六和股份有限公司 | Method for establishing near-infrared prediction model of wet-based fermented soybean meal and application |
Non-Patent Citations (3)
| Title |
|---|
| 李辉: "配合饲料中蛋白质原料的溯源技术研究", 《中国优秀硕士学位论文全文数据库 农业科技辑》 * |
| 杨增玲等: "豆粕品质近红外定量分析实验室模型在线应用", 《农业机械学报》 * |
| 王倩等: "豆粕品质近红外光谱分析模型传递方法研究", 《动物生产》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115468931B (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10031118B1 (en) | Method of determining and evaluating quality of peanut raw material suitable for protein processing | |
| CN109085136B (en) | Method for measuring content of oxide components in cement raw material by near-infrared diffuse reflection spectrum | |
| CN107024450A (en) | A kind of method for differentiating different brands and hop count milk powder based on near-infrared spectrum technique | |
| CN102967578A (en) | Method for obtaining near-infrared spectrum of beef sample online and application thereof in evaluating beef quality | |
| CN102353644A (en) | Rapid near infrared spectroscopy method for simultaneously detecting moisture content and protein content of Trichiurus japonicus surimi | |
| CN110749565A (en) | Method for rapidly identifying storage years of Pu' er tea | |
| CN109211829A (en) | A method of moisture content in the near infrared spectroscopy measurement rice based on SiPLS | |
| Cecchini et al. | Application of near‐infrared handheld spectrometers to predict semolina quality | |
| CN104155264A (en) | Method for predicting content of turpentine in loblolly pine gum by using near infrared spectroscopy | |
| CN112098357B (en) | Strawberry sensory quality grade evaluation method based on near infrared spectrum | |
| CN104568815A (en) | Method for quickly and nondestructively detecting content of volatile basic nitrogen in fresh beef | |
| CN106226267B (en) | A kind of near-infrared assay method of dry chili color value | |
| Liu et al. | Digital image analysis method for rapid measurement of rice degree of milling | |
| Turgut et al. | Estimation of the sensory properties of black tea samples using non-destructive near-infrared spectroscopy sensors | |
| CN105675538B (en) | A kind of detection method of oil cake of flax seed nutrient | |
| CN106338488A (en) | Method for fast undamaged determination of transgenic soybean milk powder | |
| CN110210005A (en) | A kind of spectrum wave number selection method of no reference value | |
| CN110231302A (en) | A kind of method of the odd sub- seed crude fat content of quick measurement | |
| CN115468931B (en) | Construction method of soybean meal quality detection model, and soybean meal quality detection method and device | |
| CN113310929A (en) | Soybean powder doped in high-temperature sterilized milk and spectral identification method of doping proportion thereof | |
| CN112945901A (en) | Method for detecting quality of ensiled soybeans based on near infrared spectrum | |
| CN110231306A (en) | A kind of method of lossless, the quick odd sub- seed protein content of measurement | |
| CN110609011A (en) | Near-infrared hyperspectral detection method and system for starch content of single-kernel corn seeds | |
| CN104502307A (en) | Method for quickly detecting content of glycogen and protein of crassostrea gigas | |
| CN108318446A (en) | It is a kind of detection the total content of vitamin E of peanut seed near-infrared spectroscopy and application |
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 |