CN106934416A - A kind of model matching method based on big data - Google Patents

A kind of model matching method based on big data Download PDF

Info

Publication number
CN106934416A
CN106934416A CN201710102144.1A CN201710102144A CN106934416A CN 106934416 A CN106934416 A CN 106934416A CN 201710102144 A CN201710102144 A CN 201710102144A CN 106934416 A CN106934416 A CN 106934416A
Authority
CN
China
Prior art keywords
model
instrument
classification
spectrum
matching
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
Application number
CN201710102144.1A
Other languages
Chinese (zh)
Other versions
CN106934416B (en
Inventor
刘彤
潘涛
曾永平
肖青青
沈鸿平
凌亚东
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong Zhongtaxun Technology Co.,Ltd.
Original Assignee
Guangzhou Sondon Network Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guangzhou Sondon Network Technology Co Ltd filed Critical Guangzhou Sondon Network Technology Co Ltd
Priority to CN201710102144.1A priority Critical patent/CN106934416B/en
Publication of CN106934416A publication Critical patent/CN106934416A/en
Application granted granted Critical
Publication of CN106934416B publication Critical patent/CN106934416B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/22Matching criteria, e.g. proximity measures
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/23Clustering techniques
    • G06F18/232Non-hierarchical techniques
    • G06F18/2321Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

The invention discloses a kind of model matching method based on big data, the method includes:Obtain instrument parameter, spectrum to be measured and/or the averaged spectrum to standard items of new equipment, then according to the instrument parameter, spectrum to be measured that obtain and/or the averaged spectrum to standard items, the prediction of result of material information is carried out so as to match most precisely applicable analysis model from basic model storehouse and personalized model and is processed;Described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.By using the method for the present invention, difference problem between the platform between big measuring appratus can be solved, model can be transmitted between large batch of equipment, so as to the limitation of near infrared spectroscopy instrument quantity can be broken, suitable for a large amount of production models of infrared spectrum instrument, so as to contribute to promoting the use of for near-infrared spectrum technique.The present invention can be widely applied in Model Matching field as a kind of model matching method based on big data.

Description

A kind of model matching method based on big data
Technical field
The present invention relates to model matching technologies, more particularly to a kind of model matching method based on big data digging technology.
Background technology
Technology word is explained:
Near infrared light:Near infrared light (NIR) be between ultraviolet-visible light (UV-Vis) and in electricity between infrared (MIR) Magnetic wave, its wave-length coverage is 700~2500nm;Near infrared spectrum can reflect hydric group X-H (such as C-H, N-H, O-H) The frequency multiplication and sum of fundamental frequencies of vibration absorb, and different groups (such as methyl, methylene, phenyl ring) or same group are in different chemical environments Near infrared absorption wavelength and intensity have significant difference, therefore, near infrared spectrum is highly suitable for the thing of hydrogeneous organic substance Change parameter measurement.
Difference problem between platform:Because manufacturing process is (trickle due to manufacturing process with a collection of instrument, or the instrument of different batches It is poor between platform caused by difference), environment (instrument is influenceed by current environment, such as temperature, humidity, cause to same sample obtain Different results), instrument loss (due to the service wear in itself of service life and instrument, cause there is platform between different instruments Between it is poor) the problems such as so that with a batch of product, to being had differences with the data measured by a sample, so as to cause an instrument The analysis model that device is set up cannot be used directly in other instruments.
Model Transfer side:The instrument of analysis model will be established as main frame, it would be desirable to use the instrument of this analysis model Device scans sample of the same race or standard items and sets up calibration model using slave respectively as slave, and slave can be by straightening die After type correction detection spectrum school is carried out using the analysis model of main frame or directly after the analysis model using main frame to predicting the outcome Just.
Based on Modern Chemometrics method, near infrared spectrum both can be used for quantitative analysis, it is also possible to for qualitative point Analysis.And for the chemometrics method in quantitative analysis and qualitative analysis, it mainly includes the following aspects:1st, spectrum is pre- Treatment and variables choice;2nd, the analysis model for predicting unknown sample property or composition is set up;3rd, mode identification method and mould Type out-of-bounds point detecting method;4th, Model Transfer method.At present, due to the tissue using the technology or personal mostly using only separate unit Or a small amount of near infrared spectrometer is analyzed the foundation of model, therefore difference problem would generally be using in Modern Chemometrics between platform Model Transfer method solve.
Currently used for the Model Transfer method of difference problem between solution platform, it mainly includes two kinds, and one is to improve straightening die The robustness of type, is second the adaptability for strengthening calibration model.Former approach is mainly by the screening of variable, differential, orthogonal The preprocess methods such as signal correction, and expand calibration model under varying environment measuring condition and use the modes such as robustness regression Noise information in filtering spectrum, the multiple partial models of fusion, improves the antimierophonic ability of model, higher to reach calibration model Reliability, the purpose of robustness.Later approach is then by mathematical method to set up slave and main frame institute light-metering spectrum, model are joined Functional relation between counting or predicting the outcome, is achieved in Model Transfer.Because former mode is some common data With the mode for the treatment of, it is impossible to reach accuracy higher, therefore be usually predominantly to use later approach, such as classical Shenk ' S algorithms to solve platform between difference problem.But, for Model Transfer method conventional at present, it still suffers from many shortcomings, For example:1st, correction amount of calculation is excessive, it is impossible to realize the transfer of a large amount of models;2nd, substantial amounts of correcting sample is needed to carry out bolster model biography Pass;3rd, dynamic change is lacked, once after instrumental correction, model has been secured, and the consumption that instrument then can be over time is caused Model is no longer accurate, needs dynamic to update;4th, user's participation is low, and user is only limitted to buyer-seller relationship with the relation of businessman.Therefore by This is visible, and when instrument amount increases severely, the method for Model Transfer is difficult to realize, and easily generation amount of calculation excessively cannot load, institute The quantity of the correcting sample for needing is excessive, the instrument to be modeled to every is required for gathering substantial amounts of spectroscopic data so as to cause work Person works' amount is big, dynamic difference the problems such as, difference problem between its platform that cannot solve substantially between big measuring appratus so then causes Near-infrared spectrum technique cannot be promoted the use of in high volume.
The content of the invention
In order to solve the above-mentioned technical problem, it is an object of the invention to provide a kind of model matching method based on big data, The limitation of near infrared spectroscopy instrument quantity can be broken, it is adaptable to a large amount of production models of infrared spectrum instrument, it is near so as to contribute to Infrared spectrum technology is promoted the use of.
First technical scheme of the present invention is:A kind of model matching method based on big data, the method includes:
Obtain the instrument parameter of new equipment;
The instrument parameter of acquisition is compared with the first category feature corresponding to each instrument classification, so as to find out with The immediate instrument classification of the new equipment;
After the new equipment obtains the corresponding basic model of the instrument classification found out with this from basic model storehouse, using this The basic model of acquisition carries out the prediction of result treatment of material information;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
Used as the preferred embodiment of the first technical scheme, the establishment step in the basic model storehouse includes:
All same instrument type equipment dispatched from the factory are carried out with random sampling, and equipment to extracting out carries out instrument parameter Collection;
Instrument parameter to collecting carries out cluster analysis, so as to obtain multiple aggregates of data, wherein, an aggregate of data table An instrument classification is shown as, the cluster centre of the aggregate of data is expressed as the first category feature corresponding to instrument classification;
For each instrument classification sets up corresponding basic model, the basic model formation base model that all foundation are obtained Storehouse.
It is described that corresponding basic model is set up for instrument classification as the preferred embodiment of the first technical scheme, its Specific steps include:
The equipment that selection belongs under same instrument classification carries out the spectra collection of known substance information sample, then by changing The modeling algorithm in meterological is learned, so as to create the analysis corresponding with the instrument classification according to the spectroscopic data for collecting Model.
Second technical scheme of the present invention is:A kind of model matching method based on big data, the method includes:
Obtain the instrument parameter of new equipment;
The instrument parameter of acquisition is compared with the first category feature corresponding to each instrument classification, so as to find out with The immediate instrument classification of the new equipment;
According to the instrument classification found out, judge whether there be the individual character corresponding with the instrument classification in personalized model storehouse Change model, if so, then finding out the personalized model corresponding with the instrument classification from personalized model storehouse, then, will obtain Instrument parameter carry out matching with the instrument feature stored in the personalized model found out and compare, so as to find out and the new equipment The personalized model for most matching, then, the prediction of result for carrying out material information using the personalized model found out is processed;Conversely, After the corresponding basic model of the instrument classification found out with this is then obtained from basic model storehouse, using the basic model of the acquisition Carry out the prediction of result treatment of material information;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
Used as the preferred embodiment of the second technical scheme, the model rule that is stored with the personalized model and instrument are special Levy.
3rd technical scheme of the present invention is:A kind of model matching method based on big data, the method includes:
Obtain the instrument parameter and spectrum to be measured of new equipment;
The instrument parameter of acquisition is compared with the first category feature corresponding to each instrument classification, so as to find out with The immediate instrument classification of the new equipment;
According to the instrument classification found out, found out from basic model storehouse and personalized model corresponding with the instrument classification Analysis model, then carries out spectrum to be measured and the spectral signature corresponding to the analysis model found out matching and compares, and finds out and treats Most like spectral signature is composed in light-metering, so as to obtain the analysis model for most matching;
The prediction of result for carrying out material information using the analysis model for most matching is processed;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
As the preferred embodiment of the 3rd technical scheme, the instrument classification that the basis is found out, from basic model storehouse and The analysis model corresponding with the instrument classification is found out in personalized model, then by spectrum to be measured and the analysis model institute for finding out Corresponding spectral signature carries out matching comparing, finds out the spectral signature most like with spectrum to be measured, thus obtain most match point The step for analysis model, it is specially:
According to the instrument classification found out, found out from basic model storehouse and personalized model corresponding with the instrument classification Analysis model, then carries out spectrum to be measured and the spectral signature corresponding to the analysis model found out matching and compares, and finds out and treats Most like spectral signature is composed in light-metering, so as to the corresponding analysis model of the most like spectral signature for obtaining with find out, then, The instrument parameter of acquisition is carried out into matching with the instrument feature corresponding to the analysis model for obtaining to compare, so as to find out and the instrument The instrument feature that parameter is most matched, so as to obtain the analysis model for most matching.
Used as the preferred embodiment of the 3rd technical scheme, the spectral signature corresponding to the analysis model, it extracts step Suddenly include:
The spectroscopic data that modeling is concentrated is analyzed, so that spectral signature is extracted, wherein, the modeling collection is for building The modeling data collection of vertical analysis model.
4th technical scheme of the present invention is:A kind of model matching method based on big data, the method includes:
Obtain instrument parameter, spectrum to be measured and the averaged spectrum to standard items of new equipment;
The instrument parameter of acquisition is compared with the first category feature corresponding to each instrument classification, so as to find out with The immediate instrument classification of the new equipment;
The second category feature corresponding to the averaged spectrum of standard items and the instrument classification is analyzed to what is obtained, from And obtain correction coefficient;
Light-metering spectrum is treated using the correction coefficient obtained to be corrected, so as to the spectrum to be measured after being corrected;
According to the instrument classification found out, found out from basic model storehouse and personalized model corresponding with the instrument classification Analysis model, then by correction after spectrum to be measured carry out matching with the spectral signature corresponding to the analysis model found out and compare, The spectral signature most like with spectrum to be measured after correction is found out, so as to obtain the analysis model for most matching;
The prediction of result for carrying out material information using the analysis model for most matching is processed;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
Used as the preferred embodiment of the 4th technical scheme, the second category feature corresponding to the instrument classification, it is carried Taking step includes:
The spectra collection that the equipment under same instrument classification carries out standard items will be belonged to;
The spectroscopic data to standard items to collecting carries out the calculating of averaged spectrum, the averaged spectrum conduct for calculating Second category feature corresponding to the instrument classification.
The beneficial effects of the invention are as follows:By using model matching method of the invention, can for new equipment match it is optimal Analysis model, but also without substantial amounts of sample, collection and the amount of calculation of a large amount of spectroscopic datas need not be carried out, therefore, this hair Difference problem between the platform that bright method can solve between big measuring appratus, model can be transmitted between large batch of equipment, so as to can beat The limitation of broken near infrared spectroscopy instrument quantity, it is adaptable to a large amount of production models of infrared spectrum instrument, so as to contribute to near-infrared Spectral technique is promoted the use of, but also has the advantages that accuracy is high, flexibility is high, operation ease is high.
Brief description of the drawings
Fig. 1 is the Establishing process schematic diagram in basic model storehouse;
Fig. 2 is a kind of Model Matching schematic flow sheet based on basic model storehouse;
Fig. 3 is the Establishing process schematic diagram of personalized model;
Fig. 4 is a kind of Model Matching schematic flow sheet based on basic model storehouse and personalized model storehouse;
Fig. 5 is a kind of Model Matching schematic flow sheet based on basic model storehouse, personalized model storehouse and spectral signature;
Fig. 6 is a kind of Model Matching flow based on basic model storehouse, personalized model storehouse, spectral signature and instrument feature Schematic diagram;
Fig. 7 is the extraction step schematic flow sheet to the second category feature corresponding to instrument classification;
Fig. 8 is a kind of Model Matching flow based on basic model storehouse, personalized model storehouse, spectral signature and averaged spectrum Schematic diagram.
Specific embodiment
Equipment in the present embodiment refers near infrared spectrometer.Analysis model in basic model storehouse is referred to as basic mould Type, the analysis model in personalized model storehouse is referred to as personalized model.
Embodiment 1, the foundation in basic model storehouse
For described basic model storehouse, its establishment step includes:
S101, all same instrument type equipment dispatched from the factory are carried out with random sampling, and equipment to extracting out carries out instrument The collection of parameter, wherein, the instrument parameter of required collection includes optical source wavelength, light source luminescent power, light source drive current, white Plate reflectivity, detection dark current etc.;
S102, cluster analysis is carried out to the instrument parameter for collecting using kmeans algorithms, so as to obtain multiple data Cluster, wherein, an aggregate of data is expressed as an instrument classification, and the cluster centre of the aggregate of data is expressed as corresponding to instrument classification First category feature;
For the step S102, it is specifically included:
S1021, initialization, set the k coordinate of initial cluster center;
S1022, (a data point correspondence represents an instrument object, and an instrument object sets including this to calculate each data point Standby instrument parameter) the distance between with each cluster centre, according to the distance for calculating, by each data point accordingly It is divided in the aggregate of data belonging to the cluster centre closest with its own, so as to obtain multiple aggregates of data, i.e. k evidence Cluster;
Whether S1023, the current aggregate of data for calculating of judgement meet cluster and terminate requirement, as judged currently to calculate Aggregate of data cluster centre it is whether equal with the cluster centre of the preceding aggregate of data for once calculating or difference be less than certain threshold Value, if so, then terminating, multiple aggregates of data that currently available multiple aggregates of data are calculated for needed for;Conversely, then recalculating The coordinate of the cluster centre of each aggregate of data, is then back to perform previous step S1022;
Wherein, the coordinate of the cluster centre described in step S1023, its computing formula is as follows:
Above-mentioned CijThe coordinate of ith cluster center jth dimension is expressed as,It is expressed as belonging to i-th k-th number of aggregate of data The coordinate that strong point is tieed up in jth;
S103, the instrument population clustering technique realized by above-mentioned steps S101 and S102, just can calculate each The cluster centre of aggregate of data, i.e. each instrument class another characteristic, now, for each instrument classification, it is corresponded to respectively again The instrument classification is belonged to comprising a series of equipment close with the other fisrt feature of instrument class, i.e. this series of equipment, Now, then can set up corresponding basic model for each instrument classification, and all set up the basic model formation base for obtaining Model library;
Wherein, described that corresponding basic model is set up for instrument classification, its specific steps includes:
Selection belongs to the armamentarium under same instrument classification, or equipment component carries out the spectrum of known substance information sample Collection, then by the modeling algorithm in Chemical Measurement, so as to be created according to the spectroscopic data for collecting and the instrument The corresponding analysis model of classification, the basic model that the analysis model of the establishment is set up needed for being.Then, then by instrument class The information such as instrument feature, modeling collection corresponding to not corresponding first category feature, analysis model, analysis model are beyond the clouds Stored.
In the present embodiment, for the foundation in above-mentioned basic model storehouse, its idiographic flow as shown in Figure 1 (sample with dregs of beans, As a example by corn, index is by taking moisture, protein content as an example):
First, to equipment 1, equipment 2, equipment 3 ..., equipment N carry out the collection of instrument parameter;Then kmeans is used Algorithm carries out cluster analysis to the instrument parameter for collecting, so as to obtain multiple different instrument classifications, and instrument classification The feature a of corresponding first category feature, such as classification A, the feature b of classification B ..., the feature k of classification K;Then, choose Belong to the spectra collection that the equipment under same instrument classification carries out known substance information sample;And then, by Chemical Measurement Modeling algorithm, so as to create the analysis model corresponding with the instrument classification, such as class according to the spectroscopic data for collecting The feature a of other A, the analysis model corresponding to it includes the analysis model for corn moisture, the analysis for zein Model and the analysis model for dregs of beans albumen.
For the establishment step in above-mentioned basic model storehouse, it is suitable for all examples below.
Embodiment 2, a kind of model matching method based on basic model storehouse
In detection-phase, for the basic model storehouse that embodiment 1 is established, when new equipment carries out Model Matching, its side Method step is included:
S201, the instrument parameter for obtaining new equipment;
S202, the first category feature corresponding to the instrument parameter of acquisition and each instrument classification is compared, so that Find out and the immediate instrument classification of the new equipment, such as classification B;
After S203, the new equipment obtain the corresponding basic models of the instrument classification B found out with this from basic model storehouse, The prediction of result for carrying out material information using the basic model of the acquisition is processed;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
For above-mentioned model matching method, its idiographic flow as shown in Figure 2 (sample by taking dregs of beans, corn as an example, index with As a example by moisture, protein content):
First, the instrument parameter of new equipment is obtained, then, corresponding to the instrument parameter that will be obtained and each instrument classification First category feature is compared, so as to find out instrument classification B immediate with the new equipment;Then, the new equipment is to corn When being detected with dregs of beans, spectra collection is carried out to corn and dregs of beans respectively, and then, obtained from basic model storehouse and the instrument After analysis model corresponding to classification B, using the analysis model of the acquisition, the spectral information to collecting carries out material information Prediction of result treatment.
Embodiment 3, a kind of model matching method based on basic model storehouse and personalized model storehouse
During user uses equipment, it will be constantly analyzed the establishment of model and safeguard, and at present for this A little newly-built analysis models, it cannot realize sharing, therefore, in order to realize the shared of these newly-built analysis models, break money The situation of source isolated island, makes other new equipments to get more accurately analysis model, and the present invention combines the concept of internet, by this A little newly-built analysis models and related data are uploaded to high in the clouds, for example, after often having user's establishment or replacement analysis model, system meeting It is automatic that user is ready that shared new analysis model and related data are uploaded to high in the clouds, and be stored in the user and made With in the corresponding instrument classification of equipment, e.g., the equipment that the user is used belongs to instrument classification A, then then created the equipment Analysis model build or renewal and related data are uploaded to high in the clouds, and are stored in instrument classification A.Now, for These analysis models for creating or updating using the process of equipment in user, it is stored beyond the clouds as personalized model, and And all personalized models are built into personalized model storehouse.In addition, in personalized model, its model rule that are not only stored with Then, be also stored with a series of instrument feature, and be accordingly stored with spectral information, algorithm information etc..
In the present embodiment, for the foundation in above-mentioned personalized model storehouse, (sample is with beans as shown in Figure 3 for its idiographic flow As a example by the dregs of rice, corn, index is by taking moisture, protein content, fiber content as an example):
Newly-built analysis model, instrument feature and sample spectra information are uploaded to high in the clouds by equipment 1 and equipment 2 in the lump, And store into corresponding instrument classification them, existing as personalized model, equipment 1 is newly-built for zein Analysis model is respectively 002 and 004, and equipment 2 is 003 for the newly-built analysis model of dregs of beans fiber, and is storing each During personalized model, corresponding storage information include model rule (wherein, the model rule include initialization system Number, constant term, product coefficient, averaged spectrum, Preprocessing Algorithm, regression algorithm), and instrument feature (wherein, the instrument spy Levy and include wavelength, driving current, light-source temperature, environment temperature, ambient humidity, detector temperature, reflectivity) and sample Spectral information etc..
In detection-phase, for the basic model storehouse that the above-mentioned personalized model for establishing and embodiment 1 are established, When new equipment carries out Model Matching, its method and step is included:
S301, the instrument parameter for obtaining new equipment;
S302, the first category feature corresponding to the instrument parameter of acquisition and each instrument classification is compared, so that Find out and the immediate instrument classification of the new equipment, such as classification B;
S303, according to the instrument classification B that finds out, judge whether to have in personalized model storehouse corresponding with instrument classification B Personalized model, if so, then find out the personalized model corresponding with instrument classification B from personalized model storehouse, then, The instrument parameter of new equipment is carried out into matching with the instrument feature stored in the personalized model found out to compare, so as to find out with The personalized model that the new equipment is most matched, then, using the personalized model found out, calls the model of the personalized model to advise Then, processed so as to carry out the prediction of result of material information;Conversely, then obtaining the instrument classification found out with this from basic model storehouse After B corresponding basic model, the prediction of result for carrying out material information using the basic model of the acquisition is processed;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
For above-mentioned model matching method, (by taking corn as an example, index is with albumen for sample as shown in Figure 4 for its idiographic flow As a example by content):
First, the instrument parameter of new equipment is obtained, then, corresponding to the instrument parameter that will be obtained and each instrument classification First category feature is compared, so as to find out instrument classification B immediate with the new equipment;Then, according to the instrument found out Whether classification B, judges there be the personalized model corresponding with instrument classification B in personalized model storehouse, if so, then from personalization The personalized model corresponding with instrument classification B is found out in model library, then, the instrument parameter of new equipment and the individual character found out Changing the instrument feature stored in model carries out matching comparing, so as to find out the personalized model most matched with the new equipment, In the present embodiment, the personalized model found out is the personalized model 003 for zein, then, using the individual character found out Change model, call the anticipation function corresponding to the model rule of the personalized model, so as to the spectrum letter to the sample for collecting Breath, such as spectral information of corn, carry out the prediction of result treatment of material information.
Embodiment 4, a kind of model matching method based on basic model storehouse, personalized model storehouse and spectral signature
Set up spectral signature and carry out implementation model matching, this can be in the analysis model in basic model storehouse and personalized model storehouse During increasing number, can preferably match and search out optimum analysis model.
Due to needing that Model Matching is carried out using spectral signature, therefore, for basic model storehouse and personalized model storehouse In analysis model, during its foundation, Spectra feature extraction step is had additional, specifically, corresponding to the analysis model Spectral signature, its extraction step includes:The spectroscopic data that modeling is concentrated is analyzed, so as to extract corresponding Spectral Properties Levy, and the spectral signature for extracting, it is stored beyond the clouds correspondingly with the analysis model for establishing.For example, using setting When standby 1, analysis model O is set up, and the data for being now used to set up used in this analysis model O constitute a modeling data collection, Referred to as modeling collection, then, is analyzed to the spectroscopic data that this modeling is concentrated, and so as to extract corresponding spectral signature o, connects , then by this analysis model O and its corresponding spectral signature o be uploaded to high in the clouds storage, and store to corresponding to equipment 1 In instrument classification.
In detection-phase, the personalized model established for the spectral signature corresponding to analysis model, embodiment 3 and The basic model storehouse that embodiment 1 is established, when new equipment carries out Model Matching, its method and step is included:
S401, the instrument parameter and spectrum to be measured that obtain new equipment;
S402, the first category feature corresponding to the instrument parameter of acquisition and each instrument classification is compared, so that Find out and the immediate instrument classification of the new equipment, such as classification B;
The instrument classification B that S403, basis are found out, finds out and instrument classification B from basic model storehouse and personalized model Corresponding analysis model, then carries out spectrum to be measured and the spectral signature corresponding to the analysis model found out matching and compares, The spectral signature most like with spectrum to be measured is found out, so that the analysis model for most matching is obtained, it is now, most like with what is found out The corresponding analysis model of spectral signature is the analysis model for most matching;
S404, the prediction of result treatment that material information is carried out using the analysis model for most matching;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
In addition, the step for for finding out the spectral signature most like with spectrum to be measured, the amendment cosine phase that it is used It is as follows like the formula of degree algorithm:
Above-mentioned U is expressed as feature set, and u is expressed as the information representated by single feature dimension in U, and i represents that new equipment is detected When testing sample spectral signature set, i.e., spectrum to be measured, j is expressed as the Spectral Properties corresponding to each analysis model to be contrasted Levy, i.e., the characteristic set of the modeling spectra collection in analysis model, sim (i, j) is expressed as spectrum to be measured and to be contrasted divides with each Comparison result between the corresponding spectral signature of analysis model.
For above-mentioned model matching method, (by taking corn as an example, index is with albumen for sample as shown in Figure 5 for its idiographic flow As a example by content):
First, the instrument parameter and spectrum to be measured of new equipment are obtained;Then, the instrument parameter that will be obtained and each instrument First category feature corresponding to classification is compared, so as to find out instrument classification B immediate with the new equipment;Then, root According to the instrument classification B for finding out, the analysis mould corresponding with instrument classification B is found out from basic model storehouse and personalized model Type, then carries out spectrum to be measured and the spectral signature corresponding to the analysis model found out matching and compares, and finds out and spectrum to be measured Most like spectral signature, so as to obtain the analysis model for most matching;Now, it is corresponding with the most like spectral signature found out Analysis model be the analysis model for most matching, in the present embodiment, the analysis model for most matching found out is instrument classification B institutes Corresponding personalized model 007;Finally, using the analysis model for most matching, i.e. personalized model 007 calls the personalized mould Anticipation function corresponding to the model rule of type 007, so as to carry out the result of material information to unknown spectrum X.
Embodiment 5, a kind of Model Matching based on basic model storehouse, personalized model storehouse, spectral signature and instrument feature Method
Above-mentioned instrument population clustering technique, the personalized model technology of the model matching method integrated use of the present embodiment And spectral characteristic matching technology, and also it is further pre- so as to realize carrying out multi objective to unknown sample with reference to instrument feature Survey.
In detection-phase, for the individual character that the spectral signature corresponding to analysis model, instrument feature, embodiment 3 are established Change the basic model storehouse that model and embodiment 1 are established, when new equipment carries out Model Matching, its method and step is included:
S501, the instrument parameter and spectrum to be measured that obtain new equipment;
S502, the first category feature corresponding to the instrument parameter of acquisition and each instrument classification is compared, so that Find out and the immediate instrument classification of the new equipment, such as classification B;
The instrument classification B that S503, basis are found out, finds out and instrument classification B from basic model storehouse and personalized model Corresponding analysis model, then carries out spectrum to be measured and the spectral signature corresponding to the analysis model found out matching and compares, The spectral signature most like with spectrum to be measured is found out, so as to the corresponding analysis mould of the most like spectral signature for obtaining with find out Type, it is then, the instrument parameter of new equipment is (corresponding i.e. with the most like spectral signature found out with the analysis model for obtaining Analysis model) corresponding to instrument feature carry out matching comparing, so as to find out the instrument most matched with the instrument parameter of new equipment Feature, so as to obtain the analysis model for most matching, now, the analysis model corresponding with the instrument feature for most matching found out is The analysis model for most matching;
S504, the prediction of result treatment that material information is carried out using the analysis model for most matching;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
For above-mentioned model matching method, (by taking corn as an example, index is with moisture for sample as shown in Figure 6 for its idiographic flow As a example by content, protein content):
First, the instrument parameter and spectrum to be measured (unknown spectrum) of new equipment (unknown device) are obtained;Then, will obtain Instrument parameter be compared with the first category feature corresponding to each instrument classification, it is closest with the new equipment so as to find out Instrument classification B;Then, according to the instrument classification B for finding out, found out from basic model storehouse and personalized model and the instrument class Other B corresponding analysis model, spectrum to be measured and the spectral signature corresponding to the analysis model found out then carried out matching ratio Compared with, find out the spectral signature most like with spectrum to be measured, so as to the most like spectral signature for obtaining with find out it is corresponding point Analysis model, the analysis model that now these are found out is the analysis model of available model matching, and in the present embodiment, that finds out is available The analysis model of Model Matching be for the analysis model 005 and 009 of zein, the analysis model 001 for corn moisture, 005 and 010;And then, the instrument feature corresponding to the analysis model for the instrument parameter of new equipment being obtained with these is matched Compare, so as to find out the instrument feature most matched with the instrument parameter of new equipment, so that the analysis model for most matching is obtained, at this In embodiment, the analysis model for most matching found out is respectively for the analysis model 009 of zein and for corn water The analysis model 005 divided;Finally, using the two analysis models, call the model rule institute of the two analysis models right respectively The anticipation function answered, so as to carry out the result of material information to unknown spectrum X.
From above-mentioned, by using the model matching method of the present embodiment, it can be entered by many characteristic matchings Testing sample model the most applicable is found out to one step, so as to greatly improve the degree of accuracy of predicted value.By the inventive technique, supply Answering business only needs to set up the part instrument for dispatching from the factory basic model, and supplier and user can be during the uses of instrument constantly Personalized model is filled to high in the clouds, with gradually increasing for personalized model, the matching effect of precision of prediction and model all can be by Constantly improve, the cooperation circulation benign so as to realize both parties.
Embodiment 6, a kind of Model Matching based on basic model storehouse, personalized model storehouse, spectral signature and averaged spectrum Method
In the present embodiment, it is also organic to combine Model Transfer technology in Modern Chemometrics, so that further The degree of accuracy that lifting predicts the outcome.Model Matching is carried out using averaged spectrum, therefore, after instrument category division terminates, i.e., After step S102, the step for have additional standard items (one of instrument annex) spectra collection, so as to draw each instrument classification The corresponding averaged spectrum to standard items, in the present embodiment, the averaged spectrum to standard items corresponding to instrument classification is made Second category feature corresponding to instrument classification.
For the second category feature corresponding to the instrument classification, its extraction step S104 includes:After step S102, Different instrument classifications are marked off, then will belong to the armamentarium under same instrument classification, or part is representational Equipment carries out the spectra collection of standard items, and then, the spectroscopic data to standard items to collecting carries out the calculating of averaged spectrum, The averaged spectrum for calculating as the second category feature corresponding to the instrument classification, finally by corresponding to the instrument classification Averaged spectrum is uploaded to high in the clouds and is stored accordingly;For example, after instrument category division terminates, 10 kinds of instruments are marked off altogether Classification, and 30 equipment are included in the first instrument classification (i.e. classification 1), it is now then representative to this 30 equipment or part Equipment, such as 15 near infrared spectrometers, the spectra collection of standard items is carried out, even this 30 or 15 near infrared spectrometers pair Standard items carry out spectra collection, and then, the spectroscopic data to standard items collected according to these carries out averaged spectrum calculating, this When the averaged spectrum that calculates then as the second category feature corresponding to the first instrument classification (i.e. classification 1).As can be seen here, Above-mentioned steps S104 is set after step s 102.
For above-mentioned step S104, its idiographic flow is as shown in Figure 7:
First, after cluster analysis is carried out according to step S101 and S102, classification 1, classification 2 ... classification N this N are marked off Individual instrument classification, the equipment that then will belong under same instrument classification carries out spectra collection to standard items, then to collecting Spectroscopic data to standard items carries out the calculating of averaged spectrum, is averaged spectrum so as to obtain the averaged spectrum corresponding to classification 1 1, and averaged spectrum 1 is then as the second category feature corresponding to classification 1.
In detection-phase, for the averaged spectrum corresponding to instrument classification, the spectral signature corresponding to analysis model, implement The basic model storehouse that the personalized model and embodiment 1 that example 3 is established are established, when new equipment carries out Model Matching, its Method and step is included:
S601, make new equipment that standard items are taken multiple scan with (at least 3 times), thus obtain new equipment to standard items Averaged spectrum, will new equipment standard items are taken multiple scan obtained by multiple spectroscopic datas carry out mean value calculation, obtain The average value for arriving is then the averaged spectrum to standard items of new equipment;
S602, the instrument parameter and spectrum to be measured that obtain new equipment;
S603, the first category feature corresponding to the instrument parameter of acquisition and each instrument classification is compared, so that Find out and the immediate instrument classification of the new equipment, such as classification B;
S604, to new equipment to the second feature corresponding to the averaged spectrum of standard items and instrument classification B, i.e., should The averaged spectrum to standard items corresponding to instrument classification B, is analyzed, so as to use the method in Modern Chemometrics (such as Shenk ' s algorithms, direct correcting algorithm DS, the direct correcting algorithm PDS of segmentation) obtains correction coefficient;
S605, treated using the correction coefficient obtained light-metering spectrum be corrected, so as to the spectrum to be measured after being corrected;
The classification B that S606, basis are found out, finds out relative with instrument classification B from basic model storehouse and personalized model The analysis model answered, then, by correction after spectrum to be measured matched with the spectral signature corresponding to the analysis model found out Compare, find out the spectral signature most like with spectrum to be measured after correction, so that the analysis model for most matching is obtained, now, institute It is the analysis model for most matching to state the analysis model corresponding to the most like spectral signature found out;
S607, the prediction of result treatment that material information is carried out using the analysis model for most matching;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
For above-mentioned model matching method, its idiographic flow is as shown in Figure 8:
First, the averaged spectrum to standard items, instrument parameter and the spectrum to be measured of new equipment are obtained;Then, by acquisition Instrument parameter is compared with the first category feature corresponding to each instrument classification, immediate with the new equipment so as to find out Instrument classification 2;Then, the averaged spectrum 2 corresponding to the averaged spectrum and instrument classification 2 to standard items of new equipment is carried out Analysis, so as to obtain correction coefficient;And then, treat light-metering using the correction coefficient obtained to compose, respectively including adopting new samples A Unknown spectrum a, the unknown spectrum b gathered to new samples B and the unknown spectrum c gathered to new samples C of collection, carry out school Just, so as to the spectrum to be measured after being corrected, spectrum A, correction spectrum B, correction spectrum C are respectively corrected;Then, according to finding out Classification 2, the analysis model corresponding with the instrument classification 2, then, high-ranking officers are found out from basic model storehouse and personalized model Positive spectrum carries out matching and compares with the spectral signature corresponding to the analysis model found out, and finds out the spectrum most like with correction spectrum Feature, so that the analysis model for most matching is obtained, respectively analysis model 2A, analysis model 2B and analysis model 2C;Finally, profit With the analysis model for most matching, the anticipation function corresponding to its model rule is called accordingly to carry out material information to correction spectrum Prediction of result treatment.
Obtained by above-mentioned, model matching method of the invention have the advantage that including:
1st, the present invention can meet the automation of larger intensity during Model transfer, and implementation model is quick, accurate transfer
In matching process of the invention, can according to instrument, feature and environment set up personalized metastasis model in itself, meet Flexibility during Model transfer, breaks through limited device, the transfer of finite model, realizes that a large amount of models are quickly and accurately migrated;
2nd, the dynamic renewal of model
In matching process of the invention, according to instrument own situation, and field feedback can be combined, set up instrument Dynamic model in its whole life cycle, becomes an ecological model system for constantly improving, and flexibility is high, and The accuracy and applicability of Model Matching can also be improved;
3rd, meet user's participation, customize the demand of personalized model
Break enterprise and the traditional of short duration buyer-seller relationship of user, with instrument as channel, integrate on-line off-line platform, it is established that With the long-term association of user, the demand of user's participation is met as bridge, and the data interacted with instrument by user are mould Type updates, and customization personalized model provides foundation.
Above is preferable implementation of the invention is illustrated, but the invention is not limited to the implementation Example, those of ordinary skill in the art can also make a variety of equivalent variations or replace on the premise of without prejudice to spirit of the invention Change, these equivalent deformations or replacement are all contained in the application claim limited range.

Claims (10)

1. a kind of model matching method based on big data, it is characterised in that:The method includes:
Obtain the instrument parameter of new equipment;
The instrument parameter of acquisition is compared with the first category feature corresponding to each instrument classification, it is new with this so as to find out The immediate instrument classification of equipment;
After the new equipment obtains the corresponding basic model of the instrument classification found out with this from basic model storehouse, using the acquisition Basic model carry out material information prediction of result treatment;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
2. a kind of model matching method based on big data according to claim 1, it is characterised in that:The basic model storehouse Establishment step include:
All same instrument type equipment dispatched from the factory are carried out with random sampling, and equipment to extracting out carries out adopting for instrument parameter Collection;
Instrument parameter to collecting carries out cluster analysis, so as to obtain multiple aggregates of data, wherein, an aggregate of data is expressed as One instrument classification, the cluster centre of the aggregate of data is expressed as the first category feature corresponding to instrument classification;
For each instrument classification sets up corresponding basic model, the basic model formation base model library that all foundation are obtained.
3. a kind of model matching method based on big data according to claim 2, it is characterised in that:Described is instrument classification Corresponding basic model is set up, its specific steps includes:
The equipment that selection belongs under same instrument classification carries out the spectra collection of known substance information sample, is then counted by chemistry Modeling algorithm in amount, so as to create the analysis mould corresponding with the instrument classification according to the spectroscopic data for collecting Type.
4. a kind of model matching method based on big data, it is characterised in that:The method includes:
Obtain the instrument parameter of new equipment;
The instrument parameter of acquisition is compared with the first category feature corresponding to each instrument classification, it is new with this so as to find out The immediate instrument classification of equipment;
According to the instrument classification found out, judge whether there be the personalized mould corresponding with the instrument classification in personalized model storehouse Type, if so, the personalized model corresponding with the instrument classification is then found out from personalized model storehouse, then, the instrument that will be obtained Device parameter carries out matching and compares with the instrument feature stored in the personalized model found out, so as to find out with the new equipment most The personalized model matched somebody with somebody, then, the prediction of result for carrying out material information using the personalized model found out is processed;Conversely, then from After obtaining the basic model corresponding with the instrument classification that this is found out in basic model storehouse, carried out using the basic model of the acquisition The prediction of result treatment of material information;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
5. a kind of model matching method based on big data according to claim 4, it is characterised in that:The personalized model In be stored with model rule and instrument feature.
6. a kind of model matching method based on big data, it is characterised in that:The method includes:
Obtain the instrument parameter and spectrum to be measured of new equipment;
The instrument parameter of acquisition is compared with the first category feature corresponding to each instrument classification, it is new with this so as to find out The immediate instrument classification of equipment;
According to the instrument classification found out, the analysis corresponding with the instrument classification is found out from basic model storehouse and personalized model Model, then carries out spectrum to be measured and the spectral signature corresponding to the analysis model found out matching and compares, and finds out and treat light-metering The most like spectral signature of spectrum, so as to obtain the analysis model for most matching;
The prediction of result for carrying out material information using the analysis model for most matching is processed;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
7. a kind of model matching method based on big data according to claim 6, it is characterised in that:What the basis was found out Instrument classification, finds out the analysis model corresponding with the instrument classification from basic model storehouse and personalized model, then will treat Light-metering spectrum carries out matching and compares with the spectral signature corresponding to the analysis model found out, and finds out the spectrum most like with spectrum to be measured Feature, so that the step for obtaining the analysis model for most matching, it is specially:
According to the instrument classification found out, the analysis corresponding with the instrument classification is found out from basic model storehouse and personalized model Model, then carries out spectrum to be measured and the spectral signature corresponding to the analysis model found out matching and compares, and finds out and treat light-metering The most like spectral signature of spectrum, so as to the corresponding analysis model of the most like spectral signature for obtaining with find out, then, will obtain The instrument parameter for obtaining carries out matching and compares with the instrument feature corresponding to the analysis model for obtaining, so as to find out and the instrument parameter The instrument feature for most matching, so as to obtain the analysis model for most matching.
8. a kind of model matching method based on big data according to claim 6 or 7, it is characterised in that:The analysis mould Spectral signature corresponding to type, its extraction step includes:
The spectroscopic data that modeling is concentrated is analyzed, so that spectral signature is extracted, wherein, the modeling collection is for foundation point Analyse the modeling data collection of model.
9. a kind of model matching method based on big data, it is characterised in that:The method includes:
Obtain instrument parameter, spectrum to be measured and the averaged spectrum to standard items of new equipment;
The instrument parameter of acquisition is compared with the first category feature corresponding to each instrument classification, it is new with this so as to find out The immediate instrument classification of equipment;
The second category feature corresponding to the averaged spectrum of standard items and the instrument classification is analyzed to what is obtained, so as to ask Go out correction coefficient;
Light-metering spectrum is treated using the correction coefficient obtained to be corrected, so as to the spectrum to be measured after being corrected;
According to the instrument classification found out, the analysis corresponding with the instrument classification is found out from basic model storehouse and personalized model Model, then by correction after spectrum to be measured carry out matching with the spectral signature corresponding to the analysis model found out and compare, find out The spectral signature most like with spectrum to be measured after correction, so as to obtain the analysis model for most matching;
The prediction of result for carrying out material information using the analysis model for most matching is processed;
Wherein, described basic model storehouse is the basic model storehouse for clustering mode based on instrument population and building up.
10. a kind of model matching method based on big data according to claim 9, it is characterised in that:The instrument classification Corresponding first category feature, its extraction step includes:
The spectra collection that the equipment under same instrument classification carries out standard items will be belonged to;
The spectroscopic data to standard items to collecting carries out the calculating of averaged spectrum, and the averaged spectrum for calculating is used as the instrument Second category feature corresponding to device classification.
CN201710102144.1A 2017-02-23 2017-02-23 Big data-based model matching method Active CN106934416B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710102144.1A CN106934416B (en) 2017-02-23 2017-02-23 Big data-based model matching method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710102144.1A CN106934416B (en) 2017-02-23 2017-02-23 Big data-based model matching method

Publications (2)

Publication Number Publication Date
CN106934416A true CN106934416A (en) 2017-07-07
CN106934416B CN106934416B (en) 2021-03-30

Family

ID=59424576

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710102144.1A Active CN106934416B (en) 2017-02-23 2017-02-23 Big data-based model matching method

Country Status (1)

Country Link
CN (1) CN106934416B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108228533A (en) * 2018-01-04 2018-06-29 中煤航测遥感集团有限公司 Materials analysis methods and device
CN108777019A (en) * 2018-04-28 2018-11-09 深圳市芭田生态工程股份有限公司 A kind of near-infrared spectroscopy transition strategy preferred method and device
CN110893100A (en) * 2019-12-16 2020-03-20 广东轻工职业技术学院 Device and method for monitoring posture change based on plantar pressure sensor
CN111239054A (en) * 2018-11-28 2020-06-05 中移物联网有限公司 Spectral analysis model application method and device
CN112629659A (en) * 2019-10-08 2021-04-09 中强光电股份有限公司 Automated model training apparatus and automated model training method for training pipelines for different spectrometers
CN112633307A (en) * 2019-10-08 2021-04-09 中强光电股份有限公司 Automatic model training device and automatic model training method for spectrometer

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101832922A (en) * 2010-05-19 2010-09-15 中国农业大学 Method for transferring near infrared model of organic fertilizer product
CN104915375A (en) * 2012-02-28 2015-09-16 艾康生物技术(杭州)有限公司 Method for automatically identifying identity information of biosensor
CN105181650A (en) * 2015-10-08 2015-12-23 滁州职业技术学院 Method for quickly identifying tea varieties through near-infrared spectroscopy technology
US20160275375A1 (en) * 2015-03-20 2016-09-22 Netra, Inc. Object detection and classification
CN106096563A (en) * 2016-06-17 2016-11-09 深圳市易特科信息技术有限公司 Plant automatic recognition system and method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101832922A (en) * 2010-05-19 2010-09-15 中国农业大学 Method for transferring near infrared model of organic fertilizer product
CN104915375A (en) * 2012-02-28 2015-09-16 艾康生物技术(杭州)有限公司 Method for automatically identifying identity information of biosensor
US20160275375A1 (en) * 2015-03-20 2016-09-22 Netra, Inc. Object detection and classification
CN105181650A (en) * 2015-10-08 2015-12-23 滁州职业技术学院 Method for quickly identifying tea varieties through near-infrared spectroscopy technology
CN106096563A (en) * 2016-06-17 2016-11-09 深圳市易特科信息技术有限公司 Plant automatic recognition system and method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
刘言等: "近红外光谱分析方法研究:从传统数据到大数据", 《中国科学》 *
鲁雄等: "近红外光谱分析技术在饲料行业中的应用", 《安徽农业科学》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108228533A (en) * 2018-01-04 2018-06-29 中煤航测遥感集团有限公司 Materials analysis methods and device
CN108777019A (en) * 2018-04-28 2018-11-09 深圳市芭田生态工程股份有限公司 A kind of near-infrared spectroscopy transition strategy preferred method and device
CN111239054A (en) * 2018-11-28 2020-06-05 中移物联网有限公司 Spectral analysis model application method and device
CN112629659A (en) * 2019-10-08 2021-04-09 中强光电股份有限公司 Automated model training apparatus and automated model training method for training pipelines for different spectrometers
CN112633307A (en) * 2019-10-08 2021-04-09 中强光电股份有限公司 Automatic model training device and automatic model training method for spectrometer
CN110893100A (en) * 2019-12-16 2020-03-20 广东轻工职业技术学院 Device and method for monitoring posture change based on plantar pressure sensor

Also Published As

Publication number Publication date
CN106934416B (en) 2021-03-30

Similar Documents

Publication Publication Date Title
CN106934416A (en) A kind of model matching method based on big data
Qaiser et al. Her 2 challenge contest: a detailed assessment of automated her 2 scoring algorithms in whole slide images of breast cancer tissues
Nawar et al. Optimal sample selection for measurement of soil organic carbon using on-line vis-NIR spectroscopy
CN105630743B (en) A kind of system of selection of spectrum wave number
CN106815643B (en) Infrared spectroscopy Model Transfer method based on random forest transfer learning
EP2161555B1 (en) Coating color database creating method, search method using the database, their system, program, and recording medium
CN101713731B (en) Method for distinguishing coating quality of medicine preparation
US10481081B2 (en) Texture analysis of a coated surface using pivot-normalization
CN108663339A (en) Corn online test method of going mouldy based on spectrum and image information fusion
CN112819054B (en) Method and device for configuring slicing template
KR20140084680A (en) Integrated Rice Distribution and Inventory Management System
Chen et al. A hybrid optimization method for sample partitioning in near-infrared analysis
CN106018331B (en) The method for estimating stability and pretreatment optimization method of multi-channel spectral system
CN115422788B (en) Power distribution network line loss analysis management method, device, storage medium and system
CN108548794A (en) A kind of biological products method for transferring near infrared model
CN107290297B (en) A kind of IR spectrum quantitative analysis method and system based on from step study
Sun et al. Research of moldy tea identification based on RF-RFE-Softmax model and hyperspectra
Vicente‐Munuera et al. The topology of vitronectin: A complementary feature for neuroblastoma risk classification based on computer‐aided detection
Li et al. scDEA: differential expression analysis in single-cell RNA-sequencing data via ensemble learning
CN113959974A (en) Universal near-infrared spectrum model transfer method
CN108827909A (en) Soil rapid classification method based on visible and near infrared spectrum and multiple targets fusion
US20140050385A1 (en) Identifying location biomarkers
CN108663334A (en) The method for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination
CN112651173A (en) Agricultural product quality nondestructive testing method based on cross-domain spectral information and generalizable system
Pessoa et al. Development of ant colony optimization (aco) algorithms based on statistical analysis and hypothesis testing for variable selection

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
TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20230710

Address after: Room A95, No. 66, Honghuagang West Street, Cencun Village, Tianhe District, Guangzhou, Guangdong 510000

Patentee after: Guangdong Zhongtaxun Technology Co.,Ltd.

Address before: 516000 room 806-812, B building, 89 Zhongshan Avenue West, Tianhe District, Guangzhou, Guangdong.

Patentee before: GUANGZHOU SONDON NETWORK TECHNOLOGY Co.,Ltd.