CN109668842A - A kind of shellfish heavy metal pollution rapid detection method - Google Patents
A kind of shellfish heavy metal pollution rapid detection method Download PDFInfo
- Publication number
- CN109668842A CN109668842A CN201910129944.1A CN201910129944A CN109668842A CN 109668842 A CN109668842 A CN 109668842A CN 201910129944 A CN201910129944 A CN 201910129944A CN 109668842 A CN109668842 A CN 109668842A
- Authority
- CN
- China
- Prior art keywords
- sample
- heavy metal
- shellfish
- pollution
- classification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (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 present invention relates to heavy metal analysis technical fields, and in particular to a kind of shellfish heavy metal pollution rapid detection method.This method first prepares sample;High spectrum image acquisition, correction and data are carried out to sample again to extract and pre-process;The waveband selection based on neighborhood evidence decision is carried out to data later, extracts characteristic wave bands subset;Then classification and Detection model is established, the classification and Detection model includes quantum nerve network classifier and integrated Study strategies and methods, quantum nerve network classifier using selected band subset to shellfish carry out pollution with pollution-free detection classify, integrated study classifier using selected band subset to shellfish carry out variety classes heavy metal pollution identification classify;Finally obtain the testing result of sample.This method carries out sample data acquisition using hyperion detection technique, waveband selection is carried out by neighborhood evidence decision theory, classification and Detection is carried out at Study strategies and methods using quantum nerve network sum aggregate, it is easy to operate quick, there is preferable test reproducibility, any chemical reagents auxiliary, no pollution to the environment are not needed in analytic process.
Description
Technical field
The present invention relates to heavy metal analysis technical fields, and in particular to a kind of shellfish heavy metal pollution rapid detection method.
Background technique
Shellfish (such as mussel, scallop, clam, oyster, mud blood clam, razor clam of hanging) is grown in bottom sediment, and position migration is small, and one
Denier encounters the more difficult avoidance of water quality, sediment pollution, also can be by the heavy metal contaminants in water body and deposit in feeding process
It accumulates in internal, in-vivo heavy metal content several orders of magnitude higher than content in ambient enviroment have seriously affected edible value,
If long-term consumption, harm can be generated to human health.Therefore, the heavy metal pollution detectability of shellfish is improved, it is ensured that shellfish
Quality and edible safety have become one of underlying issue in the urgent need to address in food safety science.
Existing detection method is mostly traditional heavy metal-polluted dye detecting method, sample need to be done analytical sampling or complexity
Processing, detection be it is destructive, it is complicated for operation, also chemical reagent is needed to assist in analytic process, there are pollutions to environment.
Summary of the invention
The present invention is to solve above-mentioned technological deficiency, provides a kind of shellfish heavy metal pollution based on hyperspectral technique
Rapid detection method, this method using hyperion detection technique carry out sample data acquisition, without to sample do analytical sampling or
Complicated processing, detection be it is nondestructive and it is easy to operate quickly, have preferable test reproducibility, do not needed in analytic process
Any chemical reagents auxiliary, no pollution to the environment.
The technical scheme is that;A kind of shellfish heavy metal pollution rapid detection method, includes the following steps;
S1: sample, including pollution-free sample, Single Pollution sample, combined pollution sample are prepared in laboratory;
S2: carrying out high spectrum image acquisition and correction to sample, carries out spectroscopic data extraction later and is pre-processed;
S3: the waveband selection based on neighborhood evidence decision is carried out to pretreated image data, extraction can be examined accurately
Survey the characteristic wave bands subset of shellfish normal sample and heavy metal pollution sample;
S4: establishing classification and Detection model, which includes quantum nerve network classifier and by multiple quantum
Neural network classifier composition integrated study classifier, quantum nerve network classifier using selected band subset to shellfish into
Row pollution with pollution-free detection classify, integrated study classifier using selected band subset to shellfish progress variety classes heavy metal
Pollution identifies classification;
S5: the testing result of sample is obtained.
Further, the treatment process of the step S2 is;
S21: high spectrum image acquisition is carried out to the sample of preparation;
S22: black and white calibration is carried out to image;
S23: any region image data is extracted;
S24: Pretreated spectra is carried out to the region image data.
Further, the Pretreated spectra in the step S24 includes;Baseline correction, variable standardization, polynary scattering school
Just handled with derivation.
Further, the attribute reduction thought of neighborhood rough set is applied in EO-1 hyperion dimensionality reduction in the step S3, is utilized
Rough set theory and evidence theory complementarity between the two, reject redundant data, and extraction can accurately detect shellfish normal sample
The characteristic wave bands subset of this and heavy metal pollution sample, detailed process is as follows:
Using each band image data as conditional attribute, heavy metal pollution situation forms decision attribute as decision attribute
Table, then best features band subset is obtained by property set Algorithm for Reduction.
Further, the property set Algorithm for Reduction includes that neighborhood evidence decision is theoretical, theoretical by neighborhood evidence decision
The process of best features band subset is as follows out:
Given neighborhood decision system NDT=<U, N, D>, D={ ω1,ω2,…,ωc, it is assumed that it is point that feature is divided into c class
Class problem, then ωqFor q-th of classification (1≤q≤c), in neighborhood δ (x), arbitrary sample xiIt is provided which an evidence miTo reflect
The membership class of sample x, by evidence miIt is merged, and calculates the BetP of final evidence mmProbability metrics, then neighborhood evidence is determined
Plan (NeighborhoodEvidential Decision, NED) is defined as:
Wherein, evidence
In formula, BetPm({ωq) it is to belong to ωqThe probability metrics of the sample of class, d (x, xi) phase between different samples
Like the measurement of degree, ωqFor sample xiAffiliated class, α are structural parameters, 0 < α < 1, φqFor the function of distance;
With BetPmThe relevant mistake classification 0-1 loss function of probability metrics are as follows:
In formula, ω (x) is the affiliated class of sample x;
Defined function and 0-1 loss function based on NED, neighborhood evidence decision error rate (Neighborhood
Evidential DecisionError, NEDE) it is defined as follows:
In formula, n is sample number, λ (ω (xi)|NED(xi)) it is by NED (xi) it is mistaken for ω (xi) when loss;
NEDE is a kind of estimation of Bayes's classification error, by selecting different property sets that NEDE is made to reach minimum,
To realize the attribute reduction based on NEDE.
Further, in the step S4 quantum nerve network classifier assorting process are as follows:
The data set of known class is established, which includes training dataset and test data set, from training dataset
The middle suitable quantum nerve network parameter of selection is iterated trained quantum nerve network, to construct quantum nerve network classification
Test data set is inputted in quantum nerve network classifier the performance for testing the classifier by device, can if reaching user and requiring
Classify for the data to unknown classification, otherwise reselects suitable quantum nerve network parameter repetitive exercise amount again
Sub-neural network, until obtaining the quantum nerve network classifier for meeting user's requirement.
Further, during being iterated trained quantum nerve network neuron use multiple excitation functions superposition,
The GradeNDivision to feature space is formed, wherein the excitation function of quantum neuron indicates are as follows:
Wherein:WTFor the transposition of weight vector;X is input vector;β is slope;θsFor amount
Son interval;Ns is quantum layer sum.
Further, the assorting process of the integrated study classifier are as follows: by multiple quantum nerve network classifiers and advance
The most ballots of row are integrated, constitute integrated study classifiers, by data set in need of test input integrated Study strategies and methods into
Row classification output.
Beneficial effects of the present invention:
The present invention is using the hyper-spectral image technique for having both image processing techniques and spectral analysis technique advantage, to realize shellfish
Class heavy metal Fast nondestructive evaluation provides technical feasibility, compared with traditional heavy metal-polluted dye detecting method, this method benefit
It is not necessarily to make sample the processing of analytical sampling or complexity with hyperion detection technique, detection is nondestructive and easy to operate fast
Speed has preferable test reproducibility, does not need any chemical reagents auxiliary, no pollution to the environment in analytic process;Simultaneously originally
Invention utilizes the waveband selection based on neighborhood evidence decision, and extraction can accurately detect shellfish normal sample and heavy metal pollution sample
This characteristic wave bands subset recycles and carries out classification and Detection in this feature band subset input classification and Detection model, and inspection of classifying
Prediction or Accurate classification, the classification and Detection model can be carried out based on quantum nerve network classifier to unknown sample by surveying model
Using the method for integrated study, nicety of grading with higher can screen well pollution metal species.
The present invention also has following realistic meaning;
(1) to the safety for guaranteeing consumer edible shellfish, ensure that people's health has particularly important realistic meaning.
(2) to carrying out using shellfish as the research of Cultivated water environmental pollution monitoring instruction biology, shellfishery is ensured
Health and sustainable development have a very important significance.
(3) using this method as Research foundation, heavy metal contaminants can be carried out in the intracorporal enrichment discipline research of shellfish, grasped
Marine pollution situation, and then the control pollution that takes measures, the development of protection and coastal economy to marine environment have important meaning
Justice.
Detailed description of the invention
Fig. 1 is flow diagram of the invention.
Fig. 2 is the classification chart of sample preparation.
Fig. 3 is hyperspectral image data acquisition and Preprocessing flow chart.
Fig. 4 is the waveband selection flow chart based on neighborhood rough set.
Fig. 5 is quantum nerve network detection of classifier flow chart.
Fig. 6 is integrated study detection of classifier flow chart.
Fig. 7 is detector general structure diagram of hardware.
Specific embodiment
The attached figures are only used for illustrative purposes and cannot be understood as limitating the patent;In order to better illustrate this embodiment, attached
Scheme certain components to have omission, zoom in or out, does not represent the size of actual product;To those skilled in the art,
The omitting of some known structures and their instructions in the attached drawings are understandable.Being given for example only property of positional relationship is described in attached drawing
Illustrate, should not be understood as the limitation to this patent.
Embodiment 1:
The present invention is with South China Sea relatively common 3 kinds of cultivated shellfish Ruditapes philippinarums, Crassostrea rivularis and Perna viridis
Research object pollutes and compound heavy using four kinds of copper, zinc, cadmium, lead representative heavy metal ions as analysis indexes for single heavy metal
Two kinds of situations of metallic pollution carry out shellfish heavy metal pollution Fast nondestructive evaluation and heavy metal ion using hyper-spectral image technique
Assay research.
As shown in Figure 1, a kind of shellfish heavy metal pollution rapid detection method, includes the following steps;
S1: sample, including pollution-free sample, Single Pollution sample, combined pollution sample are prepared in laboratory;
S2: carrying out high spectrum image acquisition and correction to sample, carries out spectroscopic data extraction later and is pre-processed;
S3: the waveband selection based on neighborhood evidence decision is carried out to pretreated image data, extraction can be examined accurately
Survey the characteristic wave bands subset of shellfish normal sample and heavy metal pollution sample;
S4: establishing classification and Detection model, which includes quantum nerve network classifier and by multiple quantum
Neural network classifier composition integrated study classifier, quantum nerve network classifier using selected band subset to shellfish into
Row pollution with pollution-free detection classify, integrated study classifier using selected band subset to shellfish progress variety classes heavy metal
Pollution identifies classification;
S5: the testing result of sample is obtained.
Sample is constituted as shown in Fig. 2, the preparation of heavy metal pollution sample is carried out in a laboratory environment first, wherein single
Polluting sample includes Cu-W ore deposit, zinc pollution, lead contamination and cadmium pollution, and combined pollution sample includes Cu-W ore deposit, zinc pollution, lead contamination
With cadmium pollution two kinds, three kinds or four kinds combinations therein.Cultured sample is placed in the loading of high spectrum image acquisition system
On platform, the high spectrum image for acquiring each sample carries out subsequent research.
As shown in figure 3, the hyperspectral image data of step S2 obtains and processing analysis process is;
S21: high spectrum image acquisition is carried out to the sample of preparation;
S22: black and white calibration is carried out to image;
S23: any region image data is extracted;
S24: Pretreated spectra is carried out to the region image data.
Wherein, the Pretreated spectra in step S24 includes;It baseline correction, variable standardization, multiplicative scatter correction and asks
Lead processing.
In the present embodiment, big for high-spectral data amount, wave band correlation is strong, data transmission and processing difficulty are big etc. asks
The attribute reduction thought of neighborhood rough set is applied in EO-1 hyperion dimensionality reduction in step s3, is managed using rough set by topic, this method
By the complementarity between the two with evidence theory, redundant data is rejected, extraction can accurately detect shellfish normal sample and a huge sum of money
Belong to the characteristic wave bands subset of pollution sample, detailed process is as shown in Figure 4:
Waveband selection is carried out based on neighborhood rough set: using each band image data as conditional attribute, heavy metal pollution feelings
Condition forms decision attribute table as decision attribute, then obtains best features band subset by property set Algorithm for Reduction.
Wherein, the property set Algorithm for Reduction includes that neighborhood evidence decision is theoretical, is obtained by neighborhood evidence decision theory
The process of best features band subset is as follows:
Given neighborhood decision system NDT=< U, N, D >, D={ ω1,ω2,…,ωc, it is assumed that it is that feature is divided into c class
Classification problem, then ωqFor q-th of classification (1≤q≤c), in neighborhood δ (x), arbitrary sample xiIt is provided which an evidence miCome anti-
The membership class for reflecting sample x, by evidence miIt is merged, and calculates the BetP of final evidence mmProbability metrics, then neighborhood evidence
Decision (NeighborhoodEvidential Decision, NED) is defined as:
Wherein, evidence
In formula, BetPm({ωq) it is to belong to ωqThe probability metrics of the sample of class, d (x, xi) phase between different samples
Like the measurement of degree, ωqFor sample xiAffiliated class, α are structural parameters, 0 < α < 1, φqFor the function of distance;
With BetPmThe relevant mistake classification 0-1 loss function of probability metrics are as follows:
In formula, ω (x) is the affiliated class of sample x;
Defined function and 0-1 loss function based on NED, neighborhood evidence decision error rate (Neighborhood
EvidentialDecisionError, NEDE) it is defined as follows:
In formula, n is sample number, λ (ω (xi)|NED(xi)) it is by NED (xi) it is mistaken for ω (xi) when loss;
NEDE is a kind of estimation of Bayes's classification error, by selecting different property sets that NEDE is made to reach minimum,
To realize the attribute reduction based on NEDE.
Substance of the same race (tested sample) curve of spectrum more it is similar be difficult to distinguish aiming at the problem that, create heavy metal pollution sample
The Fast Classification model of this and pollution-free sample analyzes influence of the parameter to model, determines optimized parameter.Focus on how needle
Classifier is constructed to particular problem, the training learning algorithm on training dataset makes it grasp rule so as to unknown sample
Carry out prediction or Accurate classification.This method plans quantum shift concept and introduces feedforward neural network, and neuron is swashed using multiple
The superposition of function is encouraged, the GradeNDivision to feature space is formed.Quantum nerve network classifier carries out pollution and uncontaminated sample
The process of detection is as shown in Figure 5;
The data set of known class is established, which includes training dataset and test data set, from training dataset
The middle suitable quantum nerve network parameter of selection is iterated trained quantum nerve network, to construct quantum nerve network classification
Test data set is inputted in quantum nerve network classifier the performance for testing the classifier by device, can if reaching user and requiring
Classify for the data to unknown classification, otherwise reselects suitable quantum nerve network parameter repetitive exercise amount again
Sub-neural network, until obtaining the quantum nerve network classifier for meeting user's requirement.
Wherein, superposition of the neuron using multiple excitation functions, shape during being iterated trained quantum nerve network
The GradeNDivision of pairs of feature space, wherein the excitation function of quantum neuron indicates are as follows:
Wherein:WTFor the transposition of weight vector;X is input vector;β is slope;θsFor amount
Son interval;Ns is quantum layer sum.
Heavy metals (copper, zinc, cadmium, lead) different for four kinds of shellfish are polluted and the examination problem of heavy metal free pollution, belong to
More classification problems, the precision of classifier are influenced very big by domain knowledge and training data and its distribution.It is efficient in order to reach
Solve the purpose of more classification problems, the method for this method proposed adoption integrated study, specific steps are as shown in Figure 6;
The assorting process of integrated study classifier are as follows: most ballots, which are formed, by multiple quantum nerve network classifiers integrates,
Integrated study classifier is constituted, data set in need of test is inputted into integrated Study strategies and methods and carries out classification output.
Embodiment 2:
In conjunction with best features band subset acquired in above-described embodiment 1, the portable shellfish weight based on ARM9 is designed
Metal detection instrument, using the design method of structuring, son of the hardware body of device respectively by several with standalone feature is tied
Structure composition, including spectra collection module, control module, display module, power supply module and some set up circuit etc..The device valence
Lattice are cheap, easy to carry, detection accuracy is relatively high, can be realized quickly and easily shellfish heavy metal analysis.
The processes of research & development of portable shellfish heavy metal pollution fast detector based on hyper-spectral image technique includes:
(1) demand analysis, specificity analysis and research and development feasibility analysis including test object;
(2) preliminary experiment early period and overall plan determine;
(3) instrument hardware platform is built, including spectra collection module design, the type selecting of Embedded Hardware Platform and other
The structure of attachment designs;
(4) exploitation of instrument software system, the customization transplanting including built-in Linux (network operating system), drives journey
Sequence is write, and (Qt is a kind of support several operation systems, and a variety of architecture hardware are flat for the design of the interface application based on Qt
The cross-platform program frame of platform);
(5) instrument whole design, function debugging and parameter optimization.
Instrument hardware is mainly made of spectra collection module and embedded control system two parts based on ARM9, overall to tie
Structure is as shown in Figure 7.The embedded control system of ARM9 series has good data-handling capacity, and more complicated data is supported to calculate
Method supports touch screen human-computer interaction, is suitble to use in this project design.The collected spectrum of Photoelectric Detection in optical system
Signal passes through signal amplification circuit and A/D D/A converting circuit, and serial input carries out at data into S3C2440 microprocessor
Reason, is calibrated and is analyzed to collected spectroscopic data using microprocessor, while being realized to spectra collection module and people
The control of machine interactive interface.
Wherein, Nand Flash is one kind of flash storage (also known as flash memory), internal to use non-linear macroelement mould
Formula provides cheap effective solution scheme for the realization of solid-state large-capacity memory.It is a kind of nonvolatile memory, is used for
Store spectroscopic data;
SDRAM (Synchronous Dynamic RandomAccess Memory) is synchronous DRAM,
For synchronizing storing data;
JTAG (Joint TestAction Group) joint test working group, is a kind of international standard test protocol
(IEEE 1149.1 is compatible) is mainly used for chip interior test.The jtag interface for referring here to standard, for accessing outside
Equipment;
LCD is used to show the classification results of data;
Touch screen is for loading human-computer interaction interface, and input operational order operates the instrument;
UART (UniversalAsynchronous Receiver/Transmitter) refers to that universal asynchronous receiving-transmitting transmits
Device.The data that it will be transmitted is converted between serial communication and parallel communications, for being communicated with external equipment.
Form of the optical path using light emitting diode in conjunction with optical filter in spectra collection module, mainly by light source
(near-infrared LED light source), beam splitting system and a few part compositions of detector (OPT101 photoelectric detector).Photoelectric detector collects
Photosignal be it is very faint, generally all in mV rank, in order to improve system signal noise ratio, need that signal amplification circuit is added
Photosignal is adjusted.A/D (analog quantity/digital quantity) conversion circuit is converted to the analog signal that photoelectric detector exports
Digital signal, serial input carry out data processing into S3C2440 microprocessor, using microprocessor to collected spectrum number
According to being analyzed.
The microprocessor is to be programmed design based on the method for research conducted in 1 based on the above embodiment, main
If based on formula and the principle for being used to analyze high-spectral data, the function of the Programming microprocessor.
Under the promotion of the new and high technologies such as computer technology and photoelectric technology, using this method as Research foundation, it can develop fast
Speed, high-precision, not damaged, intelligentized shellfish heavy metal analysis instrument, and the reality for being put into quality shellfish detection is answered
In, the edible safety of shellfish can be improved, really safe and pollution-free shellfish is provided for consumer, meets the excellent of consumer
Matter demand meets the requirement in market.
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair
The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description
To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this
Made any modifications, equivalent replacements, and improvements etc., should be included in the claims in the present invention within the spirit and principle of invention
Protection scope within.
Claims (8)
1. a kind of shellfish heavy metal pollution rapid detection method, which is characterized in that include the following steps;
S1: sample, including pollution-free sample, Single Pollution sample, combined pollution sample are prepared in laboratory;
S2: carrying out high spectrum image acquisition and correction to sample, carries out spectroscopic data extraction later and is pre-processed;
S3: the waveband selection based on neighborhood evidence decision is carried out to pretreated image data, extraction can accurately detect shellfish
The characteristic wave bands subset of class normal sample and heavy metal pollution sample;
S4: establishing classification and Detection model, which includes quantum nerve network classifier and by multiple quantum neurals
The integrated study classifier of network classifier composition, quantum nerve network classifier are dirty using having selected band subset to carry out shellfish
Dye with pollution-free detection classify, integrated study classifier using selected band subset to shellfish progress variety classes heavy metal pollution
Identify classification;
S5: the testing result of sample is obtained.
2. a kind of shellfish heavy metal pollution rapid detection method according to claim 1, which is characterized in that the step S2
Treatment process be;
S21: high spectrum image acquisition is carried out to the sample of preparation;
S22: black and white correction is carried out to image;
S23: any region image data is extracted;
S24: Pretreated spectra is carried out to the region image data.
3. a kind of shellfish heavy metal pollution rapid detection method according to claim 2, which is characterized in that the step
Pretreated spectra in S24 includes;Baseline correction, variable standardization, multiplicative scatter correction and derivation processing.
4. a kind of shellfish heavy metal pollution rapid detection method according to claim 1, which is characterized in that the step S3
It is middle that the attribute reduction thought of neighborhood rough set is applied in EO-1 hyperion dimensionality reduction, using both rough set theory and evidence theory it
Between complementarity, reject redundant data, the characteristic wave of shellfish normal sample Yu heavy metal pollution sample can accurately be detected by extracting
Cross-talk collection, detailed process is as follows: using each band image data as conditional attribute, heavy metal pollution situation as decision attribute,
Decision attribute table is formed, then best features band subset is obtained by property set Algorithm for Reduction.
5. a kind of shellfish heavy metal pollution rapid detection method according to claim 4, which is characterized in that the property set
Algorithm for Reduction includes that neighborhood evidence decision is theoretical, obtains the process of best features band subset such as by neighborhood evidence decision theory
Under:
Given neighborhood decision system NDT=<U, N, D>, D={ ω1,ω2,…,ωc, it is assumed that it is that feature is divided into the classification of c class and asks
It inscribes, then ωqFor q-th of classification (1≤q≤c), in neighborhood δ (x), arbitrary sample xiIt is provided which an evidence miCarry out reflected sample
The membership class of x, by evidence miIt is merged, and calculates the BetP of final evidence mmProbability metrics, then neighborhood evidence decision
NED is defined as:
Wherein, evidence
In formula, BetPm({ωq) it is to belong to ωqThe probability metrics of the sample of class, d (x, xi) similar journey between different samples
The measurement of degree, ωqFor sample xiAffiliated class, α are structural parameters, 0 < α < 1, φqFor the function of distance;
With BetPmThe relevant mistake classification 0-1 loss function of probability metrics are as follows:
In formula, ω (x) is the affiliated class of sample x;
Defined function and 0-1 loss function based on NED, neighborhood evidence decision error rate NEDE are defined as follows:
In formula, n is sample number, λ (ω (xi)NED(xi)) it is by NED (xi) it is mistaken for ω (xi) when loss;
NEDE is a kind of estimation of Bayes's classification error, by selecting different property sets that NEDE is made to reach minimum, thus
Realize the attribute reduction based on NEDE.
6. a kind of shellfish heavy metal pollution rapid detection method according to claim 1, which is characterized in that the step S4
The assorting process of middle quantum nerve network classifier are as follows:
The data set of known class is established, which includes training dataset and test data set, concentrates and selects from training data
It selects suitable quantum nerve network parameter and is iterated trained quantum nerve network, so that quantum nerve network classifier is constructed,
Test data set is inputted in quantum nerve network classifier to the performance for testing the classifier, can be used for if reaching user and requiring
Classify to the data of unknown classification, otherwise reselects suitable quantum nerve network parameter repetitive exercise quantum mind again
Through network, until obtaining the quantum nerve network classifier for meeting user's requirement.
7. a kind of shellfish heavy metal pollution rapid detection method according to claim 6, which is characterized in that be iterated
Neuron uses the superposition of multiple excitation functions during training quantum nerve network, forms the GradeNDivision to feature space,
Wherein the excitation function of quantum neuron indicates are as follows:
Wherein:WTFor the transposition of weight vector;X is input vector;β is slope;θsBetween quantum
Every;Ns is quantum layer sum.
8. a kind of shellfish heavy metal pollution rapid detection method according to claim 6, which is characterized in that integrated
Practise the assorting process of classifier are as follows: carry out multi-thread training parallel by multiple quantum nerve network classifiers and form most ballot collection
At data set in need of test is inputted integrated Study strategies and methods and carries out classification output by composition integrated study classifier.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910129944.1A CN109668842A (en) | 2019-02-21 | 2019-02-21 | A kind of shellfish heavy metal pollution rapid detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910129944.1A CN109668842A (en) | 2019-02-21 | 2019-02-21 | A kind of shellfish heavy metal pollution rapid detection method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109668842A true CN109668842A (en) | 2019-04-23 |
Family
ID=66152353
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910129944.1A Pending CN109668842A (en) | 2019-02-21 | 2019-02-21 | A kind of shellfish heavy metal pollution rapid detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109668842A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110596558A (en) * | 2019-10-24 | 2019-12-20 | 福州大学 | Transformer oil-paper insulation state comprehensive evaluation method combining neighborhood rough set and evidence theory |
CN112816422A (en) * | 2021-02-05 | 2021-05-18 | 岭南师范学院 | Shellfish heavy metal dynamic detection structure and method based on spectral analysis |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101539629A (en) * | 2009-04-17 | 2009-09-23 | 南京师范大学 | Remote sensing image change detection method based on multi-feature evidence integration and structure similarity |
CN103278464A (en) * | 2013-04-18 | 2013-09-04 | 北京工商大学 | Method and device for fish flesh detection |
CN103439472A (en) * | 2013-06-14 | 2013-12-11 | 北京工商大学 | Lake-reservoir cyanobacteria water bloom recognition method based on remote sensing monitoring and evidence fusion technology improvement |
CN106862284A (en) * | 2017-03-24 | 2017-06-20 | 燕山大学 | A kind of cold rolled sheet signal mode knows method for distinguishing |
KR101756960B1 (en) * | 2016-05-31 | 2017-07-13 | (주)아세아항측 | Monitoring method of heavy metals in soil using the aerial hyperspectral images |
CN109060771A (en) * | 2018-07-26 | 2018-12-21 | 温州大学 | A kind of common recognition model building method based on spectrum different characteristic collection |
-
2019
- 2019-02-21 CN CN201910129944.1A patent/CN109668842A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101539629A (en) * | 2009-04-17 | 2009-09-23 | 南京师范大学 | Remote sensing image change detection method based on multi-feature evidence integration and structure similarity |
CN103278464A (en) * | 2013-04-18 | 2013-09-04 | 北京工商大学 | Method and device for fish flesh detection |
CN103439472A (en) * | 2013-06-14 | 2013-12-11 | 北京工商大学 | Lake-reservoir cyanobacteria water bloom recognition method based on remote sensing monitoring and evidence fusion technology improvement |
KR101756960B1 (en) * | 2016-05-31 | 2017-07-13 | (주)아세아항측 | Monitoring method of heavy metals in soil using the aerial hyperspectral images |
CN106862284A (en) * | 2017-03-24 | 2017-06-20 | 燕山大学 | A kind of cold rolled sheet signal mode knows method for distinguishing |
CN109060771A (en) * | 2018-07-26 | 2018-12-21 | 温州大学 | A kind of common recognition model building method based on spectrum different characteristic collection |
Non-Patent Citations (5)
Title |
---|
刘瑶: "基于邻域粗糙集的高光谱波段选择算法研究", 《中国博士学位论文全文数据库》 * |
张培林: "《机械故障信号的量子计算分析及智能诊断》", 31 July 2015 * |
张朝龙: "一种基于量子神经网络的模拟电路故障诊断方法", 《辽宁石油化工大学学报》 * |
李俊华: "《2007中国控制与决策学术年会论文集》", 31 December 2015 * |
王培红: "《热工过程特殊参量的认知建模方法 证据理论的拓展与应用》", 31 August 2012 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110596558A (en) * | 2019-10-24 | 2019-12-20 | 福州大学 | Transformer oil-paper insulation state comprehensive evaluation method combining neighborhood rough set and evidence theory |
CN112816422A (en) * | 2021-02-05 | 2021-05-18 | 岭南师范学院 | Shellfish heavy metal dynamic detection structure and method based on spectral analysis |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Irisson et al. | Machine learning for the study of plankton and marine snow from images | |
Lombard et al. | Globally consistent quantitative observations of planktonic ecosystems | |
Brewin et al. | An intercomparison of bio-optical techniques for detecting dominant phytoplankton size class from satellite remote sensing | |
CN108875913B (en) | Tricholoma matsutake rapid nondestructive testing system and method based on convolutional neural network | |
Martsenyuk et al. | Multispectral control of water bodies for biological diversity with the index of phytoplankton | |
Leow et al. | Automated identification of copepods using digital image processing and artificial neural network | |
CN103955938A (en) | Wheat growing status diagnosing method based on mobile internet mode and leaf color analysis | |
CN107045637A (en) | A kind of blood species identification instrument and recognition methods based on spectrum | |
CN107219221A (en) | Drugs Checkout Gear and detection method | |
CN105486655A (en) | Rapid detection method for organic matters in soil based on infrared spectroscopic intelligent identification model | |
Helser et al. | A transformative approach to ageing fish otoliths using Fourier transform near infrared spectroscopy: a case study of eastern Bering Sea walleye pollock (Gadus chalcogrammus) | |
Cai et al. | Rapid identification of ore minerals using multi-scale dilated convolutional attention network associated with portable Raman spectroscopy | |
Pan et al. | Identification of complex mixtures for Raman spectroscopy using a novel scheme based on a new multi-label deep neural network | |
Di Mauro et al. | Using ZooImage automated system for the estimation of biovolume of copepods from the northern Argentine Sea | |
CN109668842A (en) | A kind of shellfish heavy metal pollution rapid detection method | |
CN108802002A (en) | A kind of quick nondestructive differentiates the silkworm seed Raman spectrum model building method of termination of diapause | |
CN115905881B (en) | Yellow pearl classification method and device, electronic equipment and storage medium | |
CN114739919A (en) | Water quality detection method based on spectrum inversion analysis | |
CN109060771A (en) | A kind of common recognition model building method based on spectrum different characteristic collection | |
Sun et al. | Hyperspectral remote sensing of the pigment c-phycocyanin in turbid inland waters, based on optical classification | |
CN101614663A (en) | As seen and near infrared spectrum bee pollen variety discriminating | |
Barman et al. | Prediction of soil pH using smartphone based digital image processing and prediction algorithm | |
US10235495B2 (en) | Method for analysis and interpretation of flow cytometry data | |
Zhu et al. | Smart polarization and spectroscopic holography for real-time microplastics identification | |
CN208476785U (en) | A kind of laser induced breakdown spectroscopy detection device suitable for aquatic products heavy metal analysis |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190423 |
|
RJ01 | Rejection of invention patent application after publication |