CN106770194B - Cereal crops place of production discrimination method based on wavelet transformation laser induced breakdown spectroscopy - Google Patents
Cereal crops place of production discrimination method based on wavelet transformation laser induced breakdown spectroscopy Download PDFInfo
- Publication number
- CN106770194B CN106770194B CN201710051767.0A CN201710051767A CN106770194B CN 106770194 B CN106770194 B CN 106770194B CN 201710051767 A CN201710051767 A CN 201710051767A CN 106770194 B CN106770194 B CN 106770194B
- Authority
- CN
- China
- Prior art keywords
- cereal crops
- spectral
- production
- sample
- value
- 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.)
- Active
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
- G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
- G01N21/71—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light thermally excited
- G01N21/718—Laser microanalysis, i.e. with formation of sample plasma
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/02—Food
- G01N33/10—Starch-containing substances, e.g. dough
Landscapes
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Plasma & Fusion (AREA)
- Optics & Photonics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention belongs to field of spectral analysis technology, more particularly to a kind of cereal crops place of production discrimination method based on wavelet transformation laser induced breakdown spectroscopy, method includes the following steps: S1 prepares the cereal crops sample of different sources, spectrum data gathering is carried out to cereal crops sample using laser induced breakdown spectroscopy (LIBS)；S2 carries out spectral peak extraction using cereal crops sample spectra of the wavelet transformation to acquisition, obtains the spectral strength value of different sources cereal crops；The spectral strength value that S3 combination wavelet transformation extracts, establishes svm classifier model；S4 is according to obtained spectroscopic data and spectral strength value, in conjunction with svm classifier model, carries out Production area recognition to unknown cereal crops.Method of the invention directly can carry out spectroscopic acquisition to different sources cereal crops, shorten sample sample preparation time and complicated chemical analysis processes, avoid secondary pollution, improve the efficiency of detection, be very suitable for the monitoring on foodstuff traceability ground.
Description
Technical field
The invention belongs to field of spectral analysis technology, and in particular to a kind of based on wavelet transformation laser induced breakdown spectroscopy
Cereal crops place of production discrimination method, this method can cereal crops directly to different sources carry out spectroscopic acquisition, improve
The efficiency of detection.
Background technique
Safety of Food Quality is the whole world widely focus of attention, and concern food safety is exactly concern health.Rice is most
One of main Three major grain crops, sown area account for the 1/5 of grain acreage, about 4.8 hundred million tons of annual output, account for world food
The 1/4 of total output, the population in 1/2 or more the whole world are also one of most important cultivated crop in China using rice as staple food.
In recent years, food-safety problem takes place frequently, such as " cadmium rice ", " rice inferior pretends to be high quality white rice " event, and consumer is caused to compel
A comprehensive understanding can be had to bought foodstuff traceability by cutting expectation.The identification of foodstuff traceability not only contributes to implement local spy
The protection of color product is more advantageous to the source of Time Optimal Control Problems food.
Foodstuff traceability tracer technique is broadly divided into two major classes: traditional electronic information coding techniques and novel comprehensive tracing back
Source technology.Comprehensive tracing technology includes Mineral Elements Analysis technology, stable isotope technology and other compositions analytical technology.Often
Mineral Elements Analysis method mainly has UV-VIS spectrophotometry, atomic spectroscopy, inductivity coupled plasma mass spectrometry
Method and Instrumental Neutron Activation method etc..Chinese patent literature " a kind of rice place of production identification method based on mineral analysis technology with
Using " (notification number 104914156A, the day for announcing are September in 2015 16) disclose and a kind of use inductively coupled plasma body
Mass spectrum (Inductively Coupled Plasma Mass Spectrometry, ICP-MS) measures rice Mineral Elements in Jadeite Shellfish
The method of content progress rice place of production classification.This method obtains higher accuracy of identification in the rice place of production is classified, but due to
It needs to carry out wet digestion to sample in analysis, so that detection process is complicated, analysis speed is slow, and easily causes secondary pollution,
It is not able to satisfy the demand of industrial application.
Laser induced breakdown spectroscopy (Laser-induced Breakdown Spectroscopy, LIBS) technology is a kind of
Based on the elemental analysis method of laser induced plasma emission spectrum analysis, because having, sample preparation is simple, analysis is quick, more
The advantages that element is analyzed simultaneously, is widely used to the fields such as industry, biologic medical, food and soil.LIBS technology is by obtaining
The strength information of plasma emission spectral line is taken to determine in sample at being grouped as, and actual LIBS spectrum is multiple members
Plain characteristic spectral line composition, in the extraction process of spectral line, the accuracy of the excessive very few impact analysis of spectral line, and spectral peak selection is usual
It needs to search spectrum one by one, it is time-consuming, complicated that spectral peak finds process, therefore is badly in need of taking a kind of automation selection characteristic spectrum
The method at peak.
(notification number is Chinese patent literature " a kind of coal characteristic measurement method based on wavelet transformation "
CN103543132A, the day for announcing are on January 29th, 2014) a kind of denoising method based on wavelet transformation is disclosed for deducting
Wavelet coefficient after denoising is substituted into offset minimum binary calibration model, Neng Gouti by the ambient noise and ambient noise of LIBS spectrum
The measurement accuracy of high calibration model.Although wavelet transformation achieves good as a result, being composed using wavelet transformation in quantitative analysis
The qualitative analysis validity that peak extracts is on the knees of the gods.
Since there are drawbacks described above and deficiency, this field needs to make and further improve, a kind of cereal is designed
Crop place of production discrimination method can avoid noise jamming, rapidly and accurately identify to the place of production of cereal crops, so as to
The needs of present food safety management.
Summary of the invention
Aiming at the above defects or improvement requirements of the prior art, the present invention provides one kind to be based on wavelet transformation induced with laser
The cereal crops place of production discrimination method of breakdown spectral is a kind of spectral peak extraction method based on continuous wavelet transform, the party
Method includes that the spectral peak of continuous wavelet transform automatically extracts, three processes of svm classifier model and Production area recognition, it is therefore intended that further
LIBS is improved to cereal crops place of production discriminating, to achieve the purpose that quick and precisely to identify.This method avoid for traditional
Cereal crops place of production discriminating detection efficiency is low, the disadvantage of detection process complexity, can directly carry out to different sources cereal crops
Spectroscopic acquisition shortens sample sample preparation time and complicated chemical analysis processes, avoids secondary pollution, improve detection
Efficiency, be very suitable for the monitoring on foodstuff traceability ground.
To achieve the above object, according to one aspect of the present invention, one kind is provided to hit based on wavelet transformation induced with laser
Wear the cereal crops place of production discrimination method of spectrum, which is characterized in that specifically comprise the following steps:
S1. the cereal crops sample for preparing different sources, using laser induced breakdown spectroscopy (LIBS) to cereal crops sample
Product carry out spectrum data gathering；
S2. spectral peak extraction is carried out to the cereal crops sample spectra acquired in step S1 using wavelet transformation, obtains difference
The spectral strength value of place of production cereal crops；
S3. the cereal crops spectral strength value extracted in step S2 is combined, svm classifier model is established:
S4. the spectroscopic data according to obtained in step S1-S2 and spectral strength value, in conjunction with the SVM established in step S3 points
Class model carries out Production area recognition to unknown cereal crops.
It is further preferred that in step sl, when directly being detected using LIBS to different sources cereal crops sample,
LIBS spectrum data gathering is carried out to the cereal crops of n different sources, every kind of place of production cereal crops sample acquires m plasma
Body spectrum obtains N=n × m plasma spectrometry altogether,
Wherein, n is the different cereal crops places of production, and n=1,2 ..., 1000, m be the spectrum width of every kind of sample of LIBS acquisition
Number, m=1,2 ..., 1000.
Preferably, in step s 2, following step is specifically included using the process that Wavelet transformation carries out the extraction of spectral strength value
It is rapid:
S2.1 carries out continuous wavelet decomposition to LIBS spectral signal:
Mexican hat wavelet function series is selected, continuous wavelet decomposition, every kind of wavelet function are carried out to spectral signal respectively
Corresponding Decomposition order is s, a length of t of displacement steps；
Wavelet transformation of the S2.2 by local maximum value mode to s layers selects crestal line, so that the spectral peak of corresponding position is obtained,
And reject duplicate spectrum peak position corresponding to crestal line；
S2.3 setting signal-to-background ratio and signal-to-noise ratio dual threshold further screen the spectral peak after step S2.2 median ridge line options, obtain
To p spectral strength value as algorithm analysis indexes；
S2.4 signal-to-noise ratio is closer with the spectral peak after the selection of signal-to-background ratio threshold value causes still to have spectral peak generation in part Chong Die, picks
Except t spectral strength value being overlapped in 0.1nm wave-length coverage is separated by, final every kind of cereal crops obtain p-t spectral strength value；
P-t spectral strength of the plasma spectrometry for the different sources cereal crops that S2.5 is extracted by wavelet function
Value is normalized, that is, selects a characteristic spectral line intensity in p-t spectral strength value strong to other p-t-1 spectral peak
Angle value makees normalized.
Preferably, in step S2.3, the value of the signal-to-background ratio is that the spectral intensity values of acquisition subtract background signal intensities
Divided by background value after value, the value of the signal-to-noise ratio is that the spectral intensity values of acquisition subtract after background signal intensities value divided by noise
Value.
Preferably, in step s3, the data of linearly inseparable are passed through construction kernel mappings function by svm classifier model
Mode is mapped in higher dimensional space, to linearly be distinguished in higher dimensional space, the kernel mappings function is using radial base letter
Number；By N group cereal crops spectroscopic data, selector is grouped spectral strength value as training set, for establishing support vector machines mould
Type, using remaining set spectral strength value as model measurement collection, for testing established supporting vector machine model accuracy of identification.
Preferably, when carrying out Production area recognition in step s 4, the specific steps are as follows:
S41. spectrum data gathering, the extraction of small echo spectral peak are carried out to cereal crops to be identified according to step S1-S2, if obtaining
Dry group cereal crops spectroscopic data；
S42. using the svm classifier model for the different sources cereal crops established in step S3, to the spectrum in the unknown place of production
Data are classified, to obtain corresponding place of production prediction.
In general, through the invention it is contemplated above technical scheme is compared with the prior art, have following advantages and
The utility model has the advantages that
(1) the method for the present invention does not need complicated chemical pre-treatment compared with traditional cereal crops Production area recognition method,
The present invention directly carries out spectroscopic acquisition to different sources cereal crops by laser induced breakdown spectroscopy, shortens sample
Product sample preparation time and complicated chemical analysis processes, avoid secondary pollution, improve the efficiency of detection.
(2) the present invention is based on the methods that continuous wavelet transform automatically extracts LIBS spectral peak, and combine svm classifier model pair
The cereal crops of different sources are identified, and are had quickly, green, the high advantage of recognition accuracy.The present invention using support to
The advantages of amount machine algorithm, species analysis quick in conjunction with LIBS, carries out quick, Accurate classification to the cereal crops of different sources.This
Inventive method extracts cereal crops characteristic spectral line only with continuous wavelet transform, shortens the time that spectral peak is extracted.
(3) the method for the present invention flexibility ratio is high, can also be with other intelligent algorithms such as genetic algorithm (Genetic
Algorithm, GA), random forest (Random Forest, RF), neural network (Artificial Neural Network,
The methods of) ANN combine, further increase the accuracy of classification.
(4) the method for the present invention does not limit to the Production area recognition of cereal crops, the identification of other equally applicable substances.And method has
There is the advantages that testing cost is low, and detection efficiency is high, testing result precision is high, is very suitable for the monitoring of food safety.
Detailed description of the invention
Fig. 1 is the structural schematic diagram for the LIBS spectra collection device that present example provides；
The process of the position Fig. 2 cereal crops place of production discrimination method of the invention based on wavelet transformation laser induced breakdown spectroscopy
Figure.
In all the appended drawings, identical appended drawing reference is used to denote the same element or structure, in which:
Wherein, 1- laser；2- optical maser wavelength reflecting mirror；3- condenser lens；4- sample to be tested；5- signal pickup assembly；
6- optical fiber；7- spectrometer；8- firing line；9-ICCD；10- data line；11- displacement platform；12- computer.
Specific embodiment
A specific embodiment of the invention is further illustrated combined with specific embodiments below.It should be noted that
The explanation of these embodiments is used to help to understand the present invention, but and is not constituted a limitation of the invention.In addition, following institute
Involved technical characteristic as long as they do not conflict with each other can be mutual in each embodiment of the present invention of description
Combination.
Due to the difference of cereal crops growing environment, directly result between the cereal crops of the same race that different sources grow out
Component content have differences, therefore the cereal crops of different sources are distinguished according to the material composition in cereal crops
It is feasible.A kind of laser induced breakdown spectroscopy cereal crops place of production discriminating side based on wavelet transformation that present example provides
Method.
The method of the present invention includes that the spectral peak of continuous wavelet transform automatically extracts, three mistakes of svm classifier model and Production area recognition
Journey, specifically include the following steps:
S1. the LIBS spectrum data gathering of sample
It is directly detected using cereal crops sample of the laser induced breakdown spectroscopy to different sources, to n different production
The cereal crops on ground carry out LIBS spectrum data gathering, and every kind of place of production cereal crops sample acquires m plasma spectrometry, obtains altogether
Obtain N=n × m plasma spectrometry, wherein n is the different cereal crops places of production, and n=1,2 ..., 1000, m be LIBS acquisition
The spectrum width number of every kind of sample, m=1,2 ..., 1000；
S2. the automatic spectral peak of continuous wavelet transform is extracted
The extraction of spectral strength value is carried out using spectrum of the continuous wavelet transform to cereal crops sample, specifically includes following mistakes
Journey:
S2.1 carries out wavelet decomposition to LIBS spectral signal:
Mexican hat wavelet function series is selected, wavelet decomposition is carried out to spectral signal respectively, every kind of wavelet function is corresponding
Decomposition order be s, a length of t of displacement steps, specifically abstract formula is as follows:
Wherein s is Decomposition order, and s=1,2 ..., t are to be displaced step-length, t=1,2 ...；ψ is Mexican hat wavelet function；
Transformation of the S2.2 by local maximum value mode to s layers selects crestal line (each Spectra peak recognition is in a crestal line), from
And the spectral peak of corresponding position is obtained, and reject duplicate spectrum peak position；It is embodied as follows: continuous wavelet is carried out to spectral signal
Transformation obtains 2-d wavelet coefficient matrix；In each scale, the local maximum of points for finding out matrix of wavelet coefficients is corresponding
Position；A crestal line is formed by the corresponding local maximum value of left and right 2 pixels of offset, all scales is traversed, will meet
The local maximum of points of condition is successively added in crestal line, and rejecting only includes two local maximum value crestal lines below, then
Every crestal line maximizing intensity spectral peak the most in the pixel of its left and right 5 will be selected；
By signal-to-background ratio threshold value, (it is to subtract background signal by the spectral intensity values of acquisition that signal-to-background ratio specifically calculates to S2.3
Intensity value is divided by background value) and snr threshold (it is to subtract background by the spectral intensity values of acquisition that signal-to-noise ratio, which specifically calculates,
Signal strength indication is divided by noise figure) spectral peak is screened, specific behaviour is as follows: the obtained spectral peak of step S2.2, in its spectral peak
It with the wave-length coverage that 25 pixels be calculating noise and spectral peak background (does not include spectral peak position in the pixel of position or so 500
Set), step-length 5 circuits sequentially, and the letter for obtaining the maximum signal-to-background ratio being calculated and signal-to-noise ratio as the spectral peak spectral intensity is carried on the back
Than and signal-to-noise ratio, and be compared with preset signal-to-background ratio, snr threshold, reject the spectral peak under threshold value；Finally
P spectral strength value being obtained as algorithm analysis indexes, p is spectral strength value number, p=1,2 ..., 10000；
S2.4 rejects the t spectral peak (resolution ratio of spectrometer is 0.05nm or so) for being separated by and being overlapped in 0.1nm wave-length coverage,
Concrete operations: the spectral peak chosen by step S2.3 still has overlapping, by the spectral peak two-by-two in the 0.1nm wave-length coverage into
Row is rejected, that is, finds spectral peak of the corresponding maximum spectral intensity as distance for 0.1nm wave-length coverage within the scope of 0.1nm, and t is to pick
The spectral peak number removed, t=1,2 ..., 5000.Finally obtain p-t spectral strength value number；
P-t spectral strength of the plasma spectrometry for the different sources cereal crops that S2.5 is extracted by wavelet function
Value is normalized, particularly as being to select a characteristic spectral line intensity in p-t spectral strength value a to other p-t-1
Spectral strength value makees normalized, and (general biological species sample is normalized using the intensity of C I 247.85nm, is led to
Often spectral peak Jing Guo abovementioned steps selects, C I 247.85nm within the scope of selected spectral line, if not if individually to C I
247.85nm being extracted accordingly)；
S3. svm classifier model is established:
The data of linearly inseparable are mapped to higher dimensional space by way of constructing kernel mappings function by svm classifier model
In, to linearly be distinguished in higher dimensional space, the kernel mappings function uses radial basis function；
The mathematical procedure of training svm classifier model is as follows: setting in training set each example as (Xi, yi), i=1 ..., n,
Xi, yiI-th of sample data label value corresponding with its is respectively indicated, wherein Xi(xi1, xi2..., xir) ∈ Rr, Rr expression training
Sample data set, r indicate the number of attribute value；yi∈ { 1,2,3 ... }, 1,2,3 ... is the label value of every group of attribute value.Algorithm
Important feature is to be mapped to the data of linearly inseparable in higher dimensional space by way of constructing kernel function, thus in higher-dimension
It is linearly distinguished in space.The present invention selects radial basis function (Radial Basis Function, RBF) to be used as kernel mappings
Function:
Thus obtained non-linear SVM classifier equation are as follows:
Constraint condition:
Wherein,Indicate the sample data of forecast set；Indicate the mapping function of forecast set sample；It indicates
Two norm distances；The number of l expression training set；αiFor Lagrange multiplier；B is the bias factor in equation；C is punishment ginseng
Number, g are mapping kernel parameter.
C can be seen that by equation, g parameter has a major impact model foundation.The range of choice of general setting c, g: c=2e, g
=2w, e, w ∈ { -10, -9 ..., 9,10 }.
Present example carries out data processing using support vector machines software, and support vector machines software selects Lin Zhiren et al.
The support vector machines software tool box (A Library for Support Vector Machines-LIBSVM) of exploitation.By N
Group cereal crops spectroscopic data selects part of group spectrum as training set, for establishing supporting vector machine model, remaining set
Spectrum is as model measurement collection, for testing established supporting vector machine model accuracy of identification.
S4. Production area recognition
S4.1 carries out spectrum data gathering to cereal crops to be identified according to step S1 to S2, small echo spectral peak is extracted, and obtains
Several groups cereal crops spectroscopic data；
S4.2 utilizes the svm classifier model for the different sources cereal crops established in step S3, to the spectrum in the unknown place of production
Data are classified, to obtain corresponding place of production prediction.
Preferably to explain the present invention, several specific embodiments are given below:
Embodiment 1
1. sample preparation
The implementation case chooses the rice sample (shown in table 1) of 10 kinds of different sources, and sample specific name and the place of production are as follows:
Golden globefish THAI Fragrant rice (Foshan), Hubei Guan Miaoshan specialty organic rice (Hubei Zhijiang), Bama of Guangxi orthodox school glutinous rice (Guangxi
Bar horse), disk mansion Feng Jin rice (Liaoning Panjin), 5 constant virtues rice fragrance of a flower rice (Heilungkiang 5 constant virtues), (Siping is double for northeast brown rice
The Liao Dynasty), cross people Yu Gong (Anqing), Hunan pond rice (Fauna of Taoyuan, Nw Hunan), Wan Niangong meters (Jiangxi Shangrao) and Chongming Island rice (Jiangsu
Taizhou).Place of production classification is carried out to them using the method for the present invention.Process for simplifying the analysis divides different sources rice
It does not number, label value is successively are as follows: 1,2,3,4,5,6,7,8,9,10；The corresponding place of production is with place name abbreviation Guangdong (GD), Hubei
(HB), Guangxi (GX), Liaoning (LN), Heilungkiang (HLJ), Jilin (JL), Anhui (AH), Hunan (HN), Jiangxi (JX), Jiangsu
(JS).Each place of production rice weighs 15g, is poured into directly as in the aluminium box of 40mm；Then letter is carried out to sample using blade
It is singly bulldozed as sample, every kind of place of production rice suppresses 4.
Table 1 is different province different sources rice inventories
2. the LIBS spectrum data gathering of rice sample
Spectra collection carries out under air environment, and experimental provision is as shown in Figure 1.Using Q-switch Nd:YAG pulse laser
Device 1 (Quantel Brilliant B, wavelength 532nm, pulse width 8ns, maximum repetition rate 10Hz) is used as excitation light source,
The plasma resonance light inspired collected by collection head 5 and be transmitted to spectrometer 7 (Andor Technology,
Mechelle5000,200~975nm of wave-length coverage, resolution lambda/λ=5000 △) it is divided, 9 (Andor of ICCD
Technology, iStar DH-334T, 1024 × 1024 pixels) the spectral signal progress photoelectric conversion come is transmitted through to spectrometer.
Sample surfaces, which are controlled, using electricity driving displacement platform 11 during spectra collection does the movement of " bow " shape with Y-direction in X direction.To prevent
Air breakdown, laser focusing lens 3 (focal length 10cm) focus are located at sample surfaces or less 1.25mm.To obtain optimal spectrum
Intensity and spectrum signal-to-background ratio, pulsed laser energy are set as 40mJ, and ICCD delay and gate-width are set to 1.5 μ s and 3 μ s.Above
Under technological parameter, the spectrum of the plasma of 10 kinds of rice is acquired, every kind of sample acquires 100 spectrum, therefore 10 kinds of rice are total
Acquire 1000 spectrum.
3. the spectral peak of wavelet transformation is extracted
Mexican hat wavelet function series is selected, continuous wavelet decomposition, every kind of wavelet function are carried out to spectral signal respectively
Corresponding Decomposition order is 6, displacement steps a length of 1, and the transformation by local maximum value mode to 6 layers selects crestal line, to obtain
The spectral peak of corresponding position, and reject duplicate spectrum peak position；And it is carried out pair with preset signal-to-background ratio 10,20 threshold value of signal-to-noise ratio
Than rejecting the spectral peak under threshold value, obtaining 93 spectral strength values as algorithm analysis indexes；Reject the left side for being separated by 0.1nm
10 spectral peaks that right avertence is moved, according to the available 83 spectral strength values of snr value；To reduce spectral intensity fluctuation to classification
As a result influence, each characteristic spectral line intensity are normalized divided by C I 247.86nm the intensity of spectral line, final each production
Ground rice obtains 82 spectral strength values.
4.SVM disaggregated model is established
Mathematical modeling, the kernel of use are carried out using the tool box LIBSVM under Matlab2010b editions software environments
Function is radial basis function, is suitble to the classification of nonlinear data, there is higher stability.G parameter in kernel radial basis function and
Penalty factor c parameter on slack variable is to influence the principal element of SVM algorithm performance, and the central principle of algorithm is also to find out most
Good mapping kernel parameter g and penalty factor c.In order to which training result has more convincingness, modeling overfitting is prevented, is tested using interaction
Demonstration optimizes impact factor, and c, g parameter is respectively 36.7583 and 0.0272.
5. Production area recognition
Spectroscopic data obtained in as procedure described above carries out the spectroscopic data in the unknown place of production using svm classifier model
Classification, to obtain corresponding place of production prediction.It is as shown in table 2 to specific different sources recognition result.Finally with SVM training
It is 99% that model, which obtains training set discrimination, and forecast set discrimination is 99.4%, therefore passes through LIBS combination continuous wavelet transform
Auxiliary support vector machines can effectively identify the rice place of production, and new technological means is provided for Production area recognition.
Table 2 is the result of different province different sources rice svm classifiers
Embodiment 2
1. sample preparation
The implementation case chooses the rice sample (such as table 3) of 6 kinds of same province different sources, sample specific product title and
The place of production is as follows: seven riverheads (Heilungkiang Suihua), 5 constant virtues rice fragrance of a flower rice (Heilungkiang 5 constant virtues), baby's diatery supplement (Heilungkiang Ning'an lotus
Hua Cun), extract woods glutinous rice (Heilungkiang Qiqihar), northeast Xiangshui County rice (Heilungkiang village, Ning'an Xiangshui County) and vegetarian diet cat Old Taylor rice
(Heilungkiang Qiqihar).Place of production classification is carried out to them using the method for the present invention.Process for simplifying the analysis, to different sources
Rice is numbered respectively, and label value is successively are as follows: 1,2,3,4,5,6；The corresponding place of production is with place name abbreviation Suihua (SH), 5 constant virtues
(WC), lotus flower (LH), Qiqihar 1 (QQHE1), Xiangshui County (XS), Qiqihar 2 (QQHE2).Same each place of production rice weighs
15g is poured into directly as in the aluminium box of 40mm；Then blade is used simply to be bulldozed as sample, often sample progress
Kind place of production rice suppresses 4.
Table 3 is same province different sources rice inventory
2. the LIBS spectrum data gathering of rice sample
Under identical experiment condition, the spectrum of the plasma of 6 kinds of rice is acquired, every kind of sample acquires 100 spectrum,
Therefore 6 kinds of rice acquire 600 spectrum altogether.
3. the spectral peak of wavelet transformation is extracted
Mexican hat wavelet function series is selected, continuous wavelet decomposition, every kind of wavelet function are carried out to spectral signal respectively
Corresponding Decomposition order is 6, displacement steps a length of 1, and the transformation by local maximum value mode to 6 layers selects crestal line, to obtain
The spectral peak of corresponding position, and reject duplicate spectrum peak position；And it is carried out pair with preset signal-to-background ratio 10,20 threshold value of signal-to-noise ratio
Than rejecting the spectral peak under threshold value, obtaining 101 spectral strength values as algorithm analysis indexes；Reject the left side for being separated by 0.1nm
4 spectral peaks that right avertence is moved, according to the available 97 spectral strength values of snr value；Classification is tied to reduce spectral intensity fluctuation
The influence of fruit, each characteristic spectral line intensity are normalized divided by C I 247.86nm the intensity of spectral line, final each place of production
Rice obtains 96 spectral strength values.
4.SVM disaggregated model is established
Mathematical modeling is carried out using the tool box LIBSVM under Matlab2010b editions software environments.C, g parameter is distinguished
For 36.7583 and 0.009.
5. Production area recognition
Spectroscopic data obtained in as procedure described above carries out the spectroscopic data in the unknown place of production using svm classifier model
Classification, to obtain corresponding place of production prediction.It is as shown in table 4 to specific different sources recognition result.Finally with SVM training
It is 99.33% that model, which obtains training set discrimination, and forecast set discrimination is 98.33%, obtains higher discrimination.
Table 4 is the result of same province different sources rice svm classifier
Embodiment 3
1. sample preparation
The implementation case chooses the rice sample (such as table 5) of 10 different 20 kinds of provinces different sources, sample specific product
Title and the place of production are as follows: golden globefish THAI Fragrant rice (Foshan), Hubei Guan Miaoshan specialty organic rice (Hubei Zhijiang), seven riverheads
(Heilungkiang Suihua), the beautiful soft rice of shrimp Wang Xiang (Dongguan, Guangdong), world grain people (Chaoyang), Bama of Guangxi orthodox school glutinous rice (Guangxi
Bar horse), disk mansion Feng Jin rice (Liaoning Panjin), the small sago of five cereals (Guangzhou Guangdong), rice fragrance of a flower rice (Heilungkiang five, 5 constant virtues
Often), (Siping is double for baby's diatery supplement (Heilungkiang Ning'an lotus flower village), extraction woods glutinous rice (Heilungkiang Qiqihar), northeast brown rice
The Liao Dynasty), cross people Yu Gong (Anqing), northeast Xiangshui County rice (Heilungkiang village, Ning'an Xiangshui County) Hunan pond rice (Fauna of Taoyuan, Nw Hunan), Wan Nian
Tribute rice (Jiangxi Shangrao), Chongming Island rice (Jiangsu Taizhou), a river autumn selenium-rich rice (Jiangxi Ji'an), Zhuxi tribute rice (Hubei ten
Weir Zhuxi) and vegetarian diet cat Old Taylor rice (Heilungkiang Qiqihar).Place of production classification is carried out to them using the method for the present invention.In order to
Simplified analysis process numbers different sources rice respectively, and label value is successively are as follows: and 1,2,3,4,5,6,7,8,9,10,11,
12,13,14,15,16,17,18,19,20；The corresponding place of production is with place name abbreviation Foshan (FS), Zhijiang (ZJ), Suihua (SH), east
Tabernaemontanus bulrush (DG), southern exposure (CY), rivers and ponds (HC), Panjin (PJ), Guangzhou (GZ), 5 constant virtues (WC), lotus flower village (LHC), Qiqihar
(QQHE), Shuangliao (SL), Anqing (AQ), village, Xiangshui County (XSC), the Land of Peach Blossoms (TY), Shangrao (SR), Taizhou (TZ), Ji'an (JA), Zhuxi
(ZX) and Heilungkiang (HLJ).Same each place of production rice weighs 15g, is poured into directly as in the aluminium box of 40mm；Then it uses
Blade be simply bulldozed as sample to sample, and every kind of place of production rice suppresses 4.
Table 5 is different provinces and same province different sources rice inventory
2. the LIBS spectrum data gathering of rice sample
Under identical experiment condition, the spectrum of the plasma of 20 kinds of rice is acquired, every kind of sample acquires 100 light
Spectrum, therefore 20 kinds of rice acquire 2000 spectrum altogether.
3. the spectral peak of wavelet transformation is extracted
Mexican hat wavelet function series is selected, continuous wavelet decomposition, every kind of wavelet function are carried out to spectral signal respectively
Corresponding Decomposition order is 6, displacement steps a length of 1, and the transformation by local maximum value mode to 6 layers selects crestal line, to obtain
The spectral peak of corresponding position, and reject duplicate spectrum peak position；And it is carried out pair with preset signal-to-background ratio 10,20 threshold value of signal-to-noise ratio
Than rejecting the spectral peak under threshold value, obtaining 93 spectral strength values as algorithm analysis indexes；Reject the left side for being separated by 0.1nm
10 spectral peaks that right avertence is moved, according to the available 83 spectral strength values of snr value；To reduce spectral intensity fluctuation to classification
As a result influence, each characteristic spectral line intensity are normalized divided by C I 247.86nm the intensity of spectral line, final each production
Ground rice obtains 82 spectral strength values.
4.SVM disaggregated model is established
Mathematical modeling is carried out using the tool box LIBSVM under Matlab2010b editions software environments.C, g parameter is distinguished
For 337.794 and 0.009.
5. Production area recognition
Spectroscopic data obtained in as procedure described above carries out the spectroscopic data in the unknown place of production using svm classifier model
Classification, to obtain corresponding place of production prediction.It is as shown in table 6 to specific different sources recognition result.Finally with SVM training
It is 98.3% that model, which obtains training set discrimination, and forecast set discrimination is 98.6%, obtains higher discrimination.
Table 6 is the result in different provinces and same province different sources rice svm classifier
Basic principles and main features and advantages of the present invention of the invention have been shown and described above.The section of the industry
Personnel are ground it should be appreciated that the present invention is not limited to the above embodiments, the above embodiments and description only describe
The principle of the present invention, without departing from the spirit and scope of the present invention, various changes and improvements may be made to the invention, these
Changes and improvements all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appended claims and
Its equivalent thereof.
Claims (5)
1. a kind of cereal crops place of production discrimination method based on wavelet transformation laser induced breakdown spectroscopy, which is characterized in that specific
Include the following steps:
S1. the cereal crops sample for preparing different sources carries out spectrum data gathering to cereal crops sample using LIBS；
S2. spectral peak extraction is carried out to the cereal crops sample spectra acquired in step S1 using wavelet transformation, obtains different sources
The spectral strength value of cereal crops；Specifically comprise the following steps:
S2.1 carries out continuous wavelet decomposition to LIBS spectral signal:
Mexican hat wavelet function series is selected, continuous wavelet decomposition is carried out to spectral signal respectively, every kind of wavelet function is corresponding
Decomposition order be s, a length of t of displacement steps；
Wavelet transformation of the S2.2 by local maximum value mode to s layers selects crestal line, to obtain the spectral peak of corresponding position, and picks
Except duplicate spectrum peak position corresponding to crestal line；
S2.3 setting signal-to-background ratio and signal-to-noise ratio dual threshold further screen the spectral peak after step S2.2 median ridge line options, obtain p
A spectral strength value is as algorithm analysis indexes；
S2.4 signal-to-noise ratio is closer with the spectral peak after the selection of signal-to-background ratio threshold value causes still to have spectral peak generation in part Chong Die, rejects phase
Every t spectral strength value being overlapped in 0.1nm wave-length coverage, final every kind of cereal crops obtain p-t spectral strength value；
P-t spectral strength value of the plasma spectrometry for the different sources cereal crops that S2.5 is extracted by wavelet function into
Row normalized selects a characteristic spectral line intensity value in p-t spectral strength value to other p-t-1 spectral strength
Value makees normalized；
S3. the cereal crops spectral strength value extracted in step S2 is combined, svm classifier model is established；
S4. the spectroscopic data according to obtained in step S1-S2 and spectral strength value, in conjunction with the svm classifier mould established in step S3
Type carries out Production area recognition to unknown cereal crops.
2. cereal crops place of production discrimination method as described in claim 1, which is characterized in that in step sl, utilize LIBS pairs
When different sources cereal crops sample is directly detected, LIBS spectroscopic data is carried out to the cereal crops of n different sources and is adopted
Collection, every kind of place of production cereal crops sample acquire m plasma spectrometry, obtain N=n × m plasma spectrometry altogether,
Wherein, n is the different cereal crops places of production, and n=1,2 ..., 1000, m be the spectrum width number of every kind of sample of LIBS acquisition, m
=1,2 ..., 1000.
3. cereal crops place of production discrimination method as claimed in claim 2, which is characterized in that in step S2.3, the letter back
The value of ratio is that the spectral intensity values of acquisition subtract after background signal intensities value divided by background value, and the value of the signal-to-noise ratio be to acquire
Spectral intensity values subtract after background signal intensities value divided by noise figure.
4. cereal crops place of production discrimination method as claimed in claim 3, which is characterized in that in step s3, svm classifier model
The data of linearly inseparable are mapped in higher dimensional space by way of constructing kernel mappings function, to be obtained in higher dimensional space
It is distinguished to linear, the kernel mappings function uses radial basis function；By N group cereal crops spectroscopic data, selector grouping spectrum
Peak intensity angle value is as training set, for establishing supporting vector machine model, using remaining set spectral strength value as model measurement collection, uses
In the established supporting vector machine model accuracy of identification of test.
5. cereal crops place of production discrimination method as claimed in claim 4, which is characterized in that carry out Production area recognition in step s 4
When, the specific steps are as follows:
S4. 1 spectrum data gathering, the extraction of small echo spectral peak are carried out to cereal crops to be identified according to step S1-S2, obtained several
Group cereal crops spectroscopic data；
S4. 2 using the different sources cereal crops established in step S3 svm classifier model, to the spectroscopic data in the unknown place of production
Classify, to obtain corresponding place of production prediction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710051767.0A CN106770194B (en) | 2017-01-20 | 2017-01-20 | Cereal crops place of production discrimination method based on wavelet transformation laser induced breakdown spectroscopy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710051767.0A CN106770194B (en) | 2017-01-20 | 2017-01-20 | Cereal crops place of production discrimination method based on wavelet transformation laser induced breakdown spectroscopy |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106770194A CN106770194A (en) | 2017-05-31 |
| CN106770194B true CN106770194B (en) | 2019-07-19 |
Family
ID=58942795
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710051767.0A Active CN106770194B (en) | 2017-01-20 | 2017-01-20 | Cereal crops place of production discrimination method based on wavelet transformation laser induced breakdown spectroscopy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106770194B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107220625B (en) * | 2017-06-01 | 2020-05-01 | 中国海洋大学 | LIBS spectrum automatic identification and classification method for sample |
| CN107632010B (en) * | 2017-09-22 | 2020-09-18 | 段忆翔 | Method for quantifying steel sample by combining laser-induced breakdown spectroscopy |
| CN109115748A (en) * | 2018-10-10 | 2019-01-01 | 浙江农林大学 | The discrimination method of different sources kudzu-vine root powder based on laser induced breakdown spectroscopy |
| CN109297952B (en) * | 2018-11-09 | 2024-02-06 | 南京信息工程大学 | Rice paper quality identification system based on laser-induced breakdown spectroscopy technology |
| CN110363119A (en) * | 2019-06-28 | 2019-10-22 | 广东中烟工业有限责任公司 | Based on wavelet transformation-random forests algorithm tobacco mildew method for quickly identifying |
| CN110658174B (en) * | 2019-08-27 | 2022-05-20 | 厦门谱识科仪有限公司 | Intelligent identification method and system based on surface enhanced Raman spectrum detection |
| CN110751048B (en) * | 2019-09-20 | 2022-06-14 | 华中科技大学 | Laser probe classification method and device based on image characteristic automatic spectral line selection |
| CN111141808B (en) * | 2020-01-20 | 2022-04-29 | 中国科学院合肥物质科学研究院 | Soil available state analysis method based on electrophoresis analysis technology and soil nutrient grade evaluation method thereof |
| CN111141809B (en) * | 2020-01-20 | 2022-04-29 | 中国科学院合肥物质科学研究院 | Soil nutrient ion content detection method based on non-contact type conductivity signal |
| CN111624193B (en) * | 2020-06-04 | 2023-01-17 | 杭州岚达科技有限公司 | LIBS (laser-induced breakdown Spectroscopy) identification method and system for wild gentiana rigescens |
| CN111982855A (en) * | 2020-08-12 | 2020-11-24 | 广东工业大学 | Method for identifying substance through spectrum signal and application thereof |
| CN112113954A (en) * | 2020-09-16 | 2020-12-22 | 江苏天瑞仪器股份有限公司 | Linear array CMOS data processing method for spectrometer |
| CN114579635B (en) * | 2022-03-04 | 2022-11-04 | 北京三月雨文化传播有限责任公司 | Big data information analysis processing system based on cloud computing |
| DE102022116302A1 (en) * | 2022-06-30 | 2024-01-04 | Clear Sea GmbH | Method and device for categorizing foodstuffs |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101915753A (en) * | 2010-07-30 | 2010-12-15 | 浙江师范大学 | Genetic Neural NetworkQuantitative analysis method for laser induced breakdown spectroscopy based on gGenetic Neural Networkgenetic neural network |
| CN103983617A (en) * | 2014-05-04 | 2014-08-13 | 华中科技大学 | Improved laser probe quantitative analysis method based on wavelet transform |
-
2017
- 2017-01-20 CN CN201710051767.0A patent/CN106770194B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101915753A (en) * | 2010-07-30 | 2010-12-15 | 浙江师范大学 | Genetic Neural NetworkQuantitative analysis method for laser induced breakdown spectroscopy based on gGenetic Neural Networkgenetic neural network |
| CN103983617A (en) * | 2014-05-04 | 2014-08-13 | 华中科技大学 | Improved laser probe quantitative analysis method based on wavelet transform |
Non-Patent Citations (3)
| Title |
|---|
| LIBS技术中光谱数据处理及在中药材分析中的应用;杨锋;《硕士学位论文》;20160603;第1，16，45页 |
| 一种基于连续小波变换的LIBS光谱自动寻峰方法;陈鹏飞等;《光谱学与光谱分析》;20140731;第34卷(第7期);1969-1972 |
| 连续小波变换在FTIR分析中的应用研究;田玉梅;《硕士学位论文》;20090530;第26-32页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106770194A (en) | 2017-05-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106770194B (en) | Cereal crops place of production discrimination method based on wavelet transformation laser induced breakdown spectroscopy | |
| CN106404748B (en) | A kind of multiline combination laser induced breakdown spectroscopy cereal crops Production area recognition method | |
| Shi et al. | Tree species classification using plant functional traits from LiDAR and hyperspectral data | |
| CN104807787B (en) | A kind of Classification of Tea discrimination method based on LIBS | |
| Chen et al. | Feasibility study of ore classification using active hyperspectral LiDAR | |
| JP6708829B1 (en) | Mixing tracking method and system for imported beef producing areas | |
| CN104730041B (en) | Method and apparatus for improving plastic identification precision of laser probe | |
| CN104142311B (en) | A kind of method of near-infrared spectrum technique prediction torch pine pine resin yield | |
| CN103884661A (en) | Soil total nitrogen real-time detection method based on soil visible-near infrared spectrum library | |
| CN104374752A (en) | Rapid detection method for nutrient elements of crops based on collinear laser-induced breakdown spectroscopy | |
| CN109916838A (en) | A kind of detection method of the rice seed germination ability based on high light spectrum image-forming and artificial neural network | |
| CN109858477A (en) | The Raman spectrum analysis method of object is identified in complex environment with depth forest | |
| CN108596085A (en) | The method for building up of soil heavy metal content detection model based on convolutional neural networks | |
| Wang et al. | Classification of Chinese tea leaves using laser-induced breakdown spectroscopy combined with the discriminant analysis method | |
| CN108830249B (en) | Winter wheat powdery mildew remote sensing monitoring method based on ASD hyperspectral data | |
| CN106018332A (en) | Near-infrared-spectrum citrus yellow shoot disease field detection method | |
| Liang et al. | Rapid discrimination of Salvia miltiorrhiza according to their geographical regions by laser induced breakdown spectroscopy (LIBS) and particle swarm optimization-kernel extreme learning machine (PSO-KELM) | |
| CN102507453A (en) | Hyperspectral monitoring method of pine wood nematode disease of black pine | |
| CN102519886A (en) | Method for detecting contents of chlorophyll a and carotinoid in crop laminas | |
| CN108181262A (en) | A kind of method using Near Infrared Spectroscopy for Rapid Sargassum horneri content of cellulose | |
| CN108573105A (en) | The method for building up of soil heavy metal content detection model based on depth confidence network | |
| CN108956584B (en) | The quick and precisely detection method of heavy metal element chromium in a kind of mulberry fruit | |
| CN115905881B (en) | Yellow pearl classification method and device, electronic equipment and storage medium | |
| CN109115748A (en) | The discrimination method of different sources kudzu-vine root powder based on laser induced breakdown spectroscopy | |
| Liu et al. | Combined laser-induced breakdown spectroscopy and hyperspectral imaging with machine learning for the classification and identification of rice geographical origin |
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 | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Li Xiangyou Inventor after: Yang Ping Inventor after: Zhu Yining Inventor after: Zhu Zhihao Inventor after: Yang Xinyan Inventor after: Li Jiaming Inventor after: Guo Lianbo Inventor after: Zeng Xiaoyan Inventor before: Li Xiangyou Inventor before: Yang Ping Inventor before: Zhu Yining Inventor before: Zhu Zhihao Inventor before: Yang Xinyan Inventor before: Li Jiaming Inventor before: Guo Lianbo Inventor before: Zeng Xiaoyan Inventor before: Lu Yongfeng |