CN110689076B - Pesticide residue detection method based on near infrared spectrum data characteristic extraction - Google Patents
Pesticide residue detection method based on near infrared spectrum data characteristic extraction Download PDFInfo
- Publication number
- CN110689076B CN110689076B CN201910924188.1A CN201910924188A CN110689076B CN 110689076 B CN110689076 B CN 110689076B CN 201910924188 A CN201910924188 A CN 201910924188A CN 110689076 B CN110689076 B CN 110689076B
- Authority
- CN
- China
- Prior art keywords
- near infrared
- sample
- data
- infrared spectrum
- follows
- 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
- 238000002329 infrared spectrum Methods 0.000 title claims abstract description 41
- 239000000447 pesticide residue Substances 0.000 title claims abstract description 38
- 238000000605 extraction Methods 0.000 title claims abstract description 27
- 238000001514 detection method Methods 0.000 title claims abstract description 17
- 238000000034 method Methods 0.000 claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 22
- 239000011159 matrix material Substances 0.000 claims abstract description 18
- 238000013145 classification model Methods 0.000 claims abstract description 16
- 238000012360 testing method Methods 0.000 claims abstract description 13
- 238000009795 derivation Methods 0.000 claims abstract description 12
- 238000013507 mapping Methods 0.000 claims abstract description 8
- 238000001228 spectrum Methods 0.000 claims description 23
- 230000003595 spectral effect Effects 0.000 claims description 17
- 238000005457 optimization Methods 0.000 claims description 10
- 239000013598 vector Substances 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 7
- 238000012706 support-vector machine Methods 0.000 claims description 6
- 239000000126 substance Substances 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 6
- 230000002411 adverse Effects 0.000 abstract description 4
- 238000012545 processing Methods 0.000 abstract description 4
- 238000000513 principal component analysis Methods 0.000 description 16
- 238000009826 distribution Methods 0.000 description 9
- 239000000575 pesticide Substances 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 125000003636 chemical group Chemical group 0.000 description 5
- 230000000052 comparative effect Effects 0.000 description 4
- 238000000862 absorption spectrum Methods 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Chemical compound O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 239000000284 extract Substances 0.000 description 2
- 235000012055 fruits and vegetables Nutrition 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 239000005947 Dimethoate Substances 0.000 description 1
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 description 1
- 238000004497 NIR spectroscopy Methods 0.000 description 1
- CDBYLPFSWZWCQE-UHFFFAOYSA-L Sodium Carbonate Chemical compound [Na+].[Na+].[O-]C([O-])=O CDBYLPFSWZWCQE-UHFFFAOYSA-L 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 238000012271 agricultural production Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- MCWXGJITAZMZEV-UHFFFAOYSA-N dimethoate Chemical group CNC(=O)CSP(=S)(OC)OC MCWXGJITAZMZEV-UHFFFAOYSA-N 0.000 description 1
- 239000012153 distilled water Substances 0.000 description 1
- 239000002552 dosage form Substances 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 229910052739 hydrogen Inorganic materials 0.000 description 1
- 239000001257 hydrogen Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000002791 soaking Methods 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 238000005507 spraying Methods 0.000 description 1
- 235000013311 vegetables Nutrition 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
- G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Spectroscopy & Molecular Physics (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 invention discloses a pesticide residue detection method based on near infrared spectrum data characteristic extraction, which comprises the following steps: collecting a near infrared spectrum of a training sample and performing second-order derivation; step two, feature extraction; step three, constructing a classification model; step four, acquiring a near infrared spectrum of a test sample, performing second-order derivation, and mapping data after second-order derivation processing to a feature space by using a feature matrix w; and (V) utilizing a two-classification model to judge the category of the data obtained in the step (IV). The method amplifies the characteristic information of the near infrared spectrum data, reduces the adverse effect of a noise sample in the acquired near infrared spectrum data on the classification of the original data, and improves the separability of a sample with pesticide residue and a sample without pesticide residue in the near infrared spectrum data.
Description
Technical Field
The invention belongs to the technical field of pesticide residue detection, and particularly relates to a pesticide residue detection method based on near infrared spectrum data characteristic extraction.
Background
The near infrared spectrum is consistent with the frequency combination of hydrogen-containing group vibration in organic molecules and the absorption region (780-2500 nm) of each level of frequency doubling, and the chemical group information in the detected organic substance can be obtained by utilizing the absorption spectrum peak information of the near infrared spectrum so as to analyze or detect the chemical composition and the content of the detected organic substance. In order to meet the requirements of rapid, convenient and intelligent detection of pesticide residues of agricultural products in the prior agriculture, students use near infrared spectroscopy to detect the pesticide residues.
However, the absorption spectrum peak signals of the near infrared spectrum are weak, and the absorption spectrum peaks of the chemical groups are overlapped seriously, so that the difference of near infrared spectrum data of two different samples, namely pesticide residue and pesticide residue-free samples in spectral analysis is small, the pesticide residue detection is difficult to classify clearly, and the final result of pesticide residue detection is influenced.
Disclosure of Invention
The invention aims to provide a pesticide residue detection method based on near infrared spectrum data characteristic extraction, and the accuracy of pesticide residue detection is improved.
The technical scheme provided by the invention is as follows: a pesticide residue detection method based on near infrared spectrum data feature extraction comprises the following steps:
collecting near infrared spectrum data of a training sample set and performing second-order derivation;
mapping the near infrared spectrum data after the second-order derivation to a feature space, and extracting a weight coefficient mapped to each feature vector of the feature space, wherein the extracted formula is as follows:
y i ＝w T x i
wherein w represents a feature matrix, and an optimization equation for solving w is as follows:
wherein the content of the first and second substances,
j (w) represents the cost function of the optimization equation,
η represents a regularization parameter; x is the number of i Represents the second derivative of the near infrared spectral vector of the ith sample, and x i ＝[x i (λ 1 ) … x i (λ p )] T The subscript p denotes the dimension of the sample spectrum; y is i Is x i The weight coefficients projected onto the eigenvectors that make up the feature matrix w determine the location of the sample i in the feature space; continuous variable u i ∈[0,1]Represents x i A probability of belonging to a non-noise sample; (1-u) i ) Represents x i A probability of belonging to a noise sample; the fuzzy variable m ∈ [1, ∞) ] represents the weighting index；e(x i ) Representing the original spectrum x i Extracting its features and reconstructing its spectrum
Is greater than or equal to>
Step three, establishing a pesticide residue and residue classification model by using a support vector machine based on a Gaussian radial basis function as a classifier and combining a weight coefficient;
collecting a near infrared spectrum of a sample to be detected as test sample data, and mapping corresponding data to a feature space by using a feature matrix w after second-order derivation; and classifying and judging the mapped data by using a binary classification model to finish detection.
As a further description of the above technical solution:
the training sample set comprises a plurality of samples with pesticide residues and without pesticide residues.
As a further description of the above technical solution:
continuous variable u in the step (two) i The formula of (2) is as follows:
calculating the obtained u according to the formula of the constraint condition i The expression is as follows:
according to the calculated u i The formula of the simplified optimization equation J (w) is:
as a further description of the above technical solution:
in the step (two), the calculation steps of the feature matrix w and the regularization parameter η are as follows:
step (1), setting the iteration times T as 1, the limit value of the iteration times as T, and learning coefficient alpha 0 ∈(0，1]；
Step (2), when T is smaller than T, executing step (3) to step (9);
step (3) of setting i =1, σ =0, and learning rate α t The expression of (c) is as follows:
α t ＝α 0 (1-1/T)
step (4), when i is less than n, executing step (5) to step (8);
step (5), calculating y = w T x i ,h＝yw,ν＝w T h
And (6) calculating a weight coefficient, wherein an updated iterative formula is as follows:
step (7), the iterative formula of sigma update is as follows:
σ new ＝σ old +e(x i )
step (8), updating the value of i, and enabling i = i +1;
step (9), updating the value of t, and enabling t = t +1; calculating a regularization parameter η, which is as follows:
η＝(σ/n)。
the invention has the beneficial effects that: the characteristic information of the near infrared spectrum data is amplified, the adverse effect of a noise sample in the acquired near infrared spectrum data on the classification of the original data is reduced, and the separability of a sample with pesticide residue and a sample without pesticide residue in the near infrared spectrum data is improved, and the method specifically comprises the following steps:
(1) The acquired near infrared spectrum data are preprocessed by a derivation method, signals of the original spectrum data can be amplified to a certain extent by processing the spectrum data through a second derivative, background information contained in the acquired spectrum data is eliminated, chemical groups in the original spectrum data are more obvious in spectrum peaks, and the characteristic extractability in the spectrum data is effectively improved;
(2) The feature extraction method of the invention adopts fuzzy variable and continuous variable u i Whether the acquired near infrared spectrum data are noise samples or not is judged, and a threshold eta is automatically obtained by using an iterative method, so that the complexity of manually setting the threshold is avoided, and the adverse effect of the noise samples on the classification of the original data is effectively reduced;
(3) In the pesticide residue detection method, in the characteristic extraction process, the characteristic matrix w of the characteristic space of the infrared spectrum data is obtained by utilizing the iterative method, and the near infrared spectrum data is mapped into the characteristic space through the characteristic matrix, so that the separability of the data among samples in the characteristic space is obviously improved, and the classification accuracy and the generalization capability of a classification model established in the later stage are improved.
(4) Compared with the existing Principal Component Analysis (PCA), the method for detecting pesticide residues has the advantages that the method for extracting the characteristics is unique, and has robustness on a noise sample (namely the noise sample has far less influence on correct extraction noise than the PCA) during characteristic extraction, so that the extracted characteristics are correct, and the characteristics of the PCA which is usually extracted under the condition of the noise sample are incorrect; in addition, PCA is a linear extraction method, and the feature extraction used in the method is a nonlinear extraction method, so that two types of samples can be more separable in a feature space, and the classification of the two types of samples is facilitated.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
FIG. 1 is a spectrum of a near infrared spectrum of a training sample of example 2;
FIG. 2 is a sample distribution diagram in the first two feature space after the method of the present invention extracts the features of the noise-free training sample in embodiment 2;
FIG. 3 is a sample distribution diagram in the first-second feature space after the method of the present invention extracts the features of noisy training samples in example 2;
FIG. 4 is a sample distribution diagram in a first and second feature space after feature extraction of a noise-free training sample in a comparative example by a PCA method;
fig. 5 is a sample distribution diagram in the first two feature space after the features of the noisy training sample are extracted by the PCA method in the comparative example.
Detailed Description
The structural features of the present invention will now be described in detail with reference to the accompanying drawings.
Example 1
The embodiment provides a pesticide residue detection method based on near infrared spectrum data characteristic extraction, which comprises the following steps:
preparing a training sample set, wherein the training sample set comprises a plurality of fruits and vegetables with pesticide residues and without pesticide residues, and acquiring a near infrared spectrum of a training sample by using an IAS-2000 near infrared spectrometer;
performing second-order derivation on the near infrared spectrum data of the sample in the sample data set acquired in the step (first), and processing the spectrum data through the second-order derivation to amplify signals of the original spectrum data to a certain extent and eliminate background information contained in the acquired spectrum data, so that chemical groups in the original spectrum data are more obvious in spectrum peaks, and the characteristic extractability in the spectrum data is effectively improved; the second derivative formula is as follows:
wherein r (λ) is near infrared spectrum, λ ∈ [780,2500] represents wavelength, and Δ λ represents wavelength step.
The second derivative processing of the spectral data can amplify the signals of the original spectral data to a certain extent and eliminate background information contained in the acquired spectral data, so that chemical groups in the original spectral data are more remarkable in spectral peaks;
step (three),
After the data obtained in the step (two) is processed, mapping the data to a weight coefficient on each feature vector of a feature space, wherein the formula is as follows:
y i ＝w T x i formula (2)
Wherein, w represents a feature matrix formed by feature vectors, and an optimization equation for solving the feature matrix w is as follows:
wherein J (w) represents the cost function of the optimization equation, and the formula is shown as follows,
where η represents a regularization parameter, x i Represents the second derivative of the near infrared spectral vector of the ith sample, and x i ＝[x i (λ 1 ) … x i (λ p )] T The subscript p denotes the dimension of the sample spectrum; y is i Is x i The respective weight coefficients projected onto the eigenvectors (obtained by the method of the invention) constituting the eigenmatrix w determine the position of this i sample in the eigenspace; continuous variable u i ∈[0,1]Represents x i Probability of belonging to a non-noise sample, (1-u) i ) Represents x i Probability of belonging to a noise sample; the fuzzy variable m ∈ [1, ∞) ] represents a weighting index; e (x) i ) Representing the original spectrum x i Extracting its features and reconstructing its spectrum
Error between, e (x) i ) The formula is expressed as follows:
related in equation (3)Continuous variable u i The constraints of (2) are as follows:
u is calculated according to the formula (5) i The expression is as follows:
according to the calculated u i The formula of the simplified optimization equation J (w) is:
by using
Solving for w in accordance with the optimization equation of equation (6)>
Is represented as follows:
and step four, aiming at the weight coefficient data of the near infrared spectrum obtained in the step three, mapped in the feature space, and establishing a pesticide residue and non-residue two-classification model by using a Support Vector Machine (SVM) based on a Gaussian Radial Basis Function (RBF) as a classifier.
Preparing a plurality of fruits and vegetables to be classified as test samples, collecting the near infrared spectrum of the test samples by using an IAS-2000 near infrared spectrometer, and taking the obtained near infrared spectrum as a test sample data set;
step six, performing second derivative data on the test sample data obtained in the step five, and mapping the data to a feature space by using the feature matrix w in the step three;
and (seventhly), judging the category of the data obtained in the step (six) by using the two classification models obtained in the step (four), and realizing the detection of whether the pesticide residue of the corresponding sample exists.
Further, in the step (three), the feature matrix w and the regularization parameter η in the formula (6) are obtained by calculation, and the specific steps are as follows:
step (1), firstly, setting the iteration number T as 1, the limit value of the iteration number as T, and learning coefficient alpha 0 ∈(0，1]；
Step (2), when T is smaller than T, executing the step (3) to the step (9);
step (3) of setting i =1, σ =0, and learning rate α t The expression of (a) is as follows:
α t ＝α 0 (1-1/T)
step (4), when i is smaller than n, executing step (5) to step (8);
step (5), calculating y = w T x i ,h＝yw,ν＝w T h
And (6) calculating a characteristic weight coefficient, wherein an updated iterative formula is as follows:
step (7), the iterative formula of sigma update is as follows:
σ new ＝σ old +e(x i ) Formula (10)
Step (8), updating the value of i, and enabling i = i +1;
step (9), updating the value t, and enabling t = t +1; calculating a regularization parameter η, which is as follows:
eta = (sigma/n) formula (11)
In this embodiment, fuzzy variables and continuous variables u are used i Whether the acquired near infrared spectrum data is a noise sample or not is judged, and the threshold eta is automatically acquired by using an iterative method, so that the complexity of manually setting the threshold is avoided, and the adverse effect of the noise sample on the classification of the original data is effectively reduced.
In the feature extraction method in this embodiment, the feature matrix w of the near infrared spectrum data is extracted by using an iterative method, and the spectrum data is mapped to the feature space through the feature matrix, so that the separability of the sample in the feature space is significantly improved, and the classification accuracy and generalization capability of a classification model established at a later stage are improved.
In this embodiment, the feature matrix w and the classification model are obtained through training data, then the test data are directly projected by using w to obtain a weight coefficient in the feature space (the coefficient also determines the distribution in the feature space), and then the obtained classification model is used to determine which type of the test data is distributed correspondingly to determine whether pesticide residue exists.
Example 2
On the basis of embodiment 1, embodiment 2 is different in that: the preparation method of the training sample and the test sample comprises the following specific steps:
selecting Shanghai green purchased in a vegetable market as a sample, and randomly selecting 80 leaves with uniform color, moderate leaf size, no damage to the leaves and no yellow edge;
in the step (2), the pesticide used in the embodiment is dimethoate (C5H 12NO3PS 2) widely applied in agricultural production, the dosage form is missible oil, the mass percentage content is 40%, and the pesticide is produced in Jiangsu Tenglong biological pharmaceutical industry Co. Preparing 30 solutions of pesticides at intervals of 1 to 1 and 1500 by using distilled water, and placing the solutions in 30 volumetric flasks of 250 ml;
step (3), firstly, placing 80 leaves in soda water for soaking for about 5 minutes, then repeatedly washing the leaves by using clear water, and placing the leaves in a ventilated place for natural airing; then dividing 60 leaves into a group A and a group B, wherein each group comprises 30 leaves, spraying pesticides with corresponding proportion on the front and back surfaces of each group of leaves by using a glass sprayer which is uniformly sprayed, placing the sprayed pesticides at a vent for natural airing, and preparing the pesticides in the step (2);
and (4) averagely dividing 20 leaves which are not sprayed with the pesticide into a group A and a group B, wherein the group A is a training sample, the group B is a testing sample, and the two groups respectively have 40 samples which respectively comprise 30 samples with pesticide residues and 10 samples without pesticide residues.
In this embodiment, an obtained near infrared spectrum of a training sample is shown in fig. 1, and on the basis of second derivative preprocessing, after extracting features of spectral data of a noise-free training sample by using a feature extraction method, a sample distribution diagram in which the data are mapped to a first feature space and a second feature space is shown in fig. 2; fig. 3 shows a sample distribution diagram obtained by extracting the features of the spectral data of the noisy training sample by using a feature extraction method and mapping the data into a first and second feature space.
Comparative example
On the basis of example 2, comparative example 1 differs in that:
(1) The feature extraction method in the step (III) adopts a Principal Component Analysis (PCA) which is commonly used in the prior art;
(2) And in the step (IV), a Gaussian Radial Basis Function (RBF) Support Vector Machine (SVM) is used as a classifier, the obtained data of the characteristic space are trained, a two-classification model for judging whether pesticide residues exist is obtained, and the accuracy of the two-classification model in the classification judgment of the near infrared spectrum in the test set is verified.
As can be seen from fig. 4, after the PCA processes the spectral data containing noise, the sample classes are seriously interleaved and overlapped, the overall sample distribution is more aggregated, and the classification is difficult; compared with PCA, the method has robustness to noise samples, and the separability of the two types of samples is good after the spectral data containing noise is processed. As can be seen from fig. 5, for the spectral data of the noisy samples, PCA is susceptible to the noise during the sample classification process, and causes more serious interleaving between sample classes; as can be seen from fig. 3, the noisy samples have less influence on the data distribution processed by the method of the present invention. Further, the results of comparing the results of the two feature extraction methods of PCA and example 2 are shown in the following table:
TABLE 1 comparison of recognition accuracy rates after modeling by two feature extraction methods
As can be seen from Table 1, after PCA treatment, the recognition accuracy of the obtained two-classification model to the training set is 92.5%, and the recognition rate of the prediction set is 87.5%. The recognition rate of the obtained binary model to the training set is 97.5% and the recognition accuracy to the prediction set is 97.5% by the method of the embodiment 2. The recognition accuracy of the bipartite model processed by the method for the training set and the test set containing the noise sample is higher than that of PCA. Therefore, the method has better characteristic extraction effect on the spectral data than PCA, and improves the classification accuracy and generalization capability of the classification model established at the later stage.
In conclusion, the method has robustness for the noise-containing sample, and has more remarkable effect on improving the separability of the sample data.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that changes may be made in the embodiments and/or equivalents thereof without departing from the spirit and scope of the invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (4)
1. A pesticide residue detection method based on near infrared spectrum data characteristic extraction is characterized by comprising the following steps:
collecting near infrared spectrum data of a training sample set and performing second-order derivation;
mapping the near infrared spectrum data subjected to second-order derivation to a feature space, and extracting a weight coefficient mapped to each feature vector of the feature space, wherein the extracted formula is as follows:
y i ＝w T x i
wherein w represents a feature matrix, and an optimization equation for solving w is as follows:
wherein the content of the first and second substances,
j (w) represents the cost function of the optimization equation,
η represents a regularization parameter; x is the number of i Represents the second derivative of the near infrared spectral vector of the ith sample, and x i ＝[x i (λ 1 ) … x i (λ p )] T The subscript p denotes the dimension of the sample spectrum; y is i Is x i The weight coefficients are projected onto the feature vectors forming the feature matrix w, and the weight coefficients determine the position of the sample i in the feature space; continuous variable u i ∈[0,1]Represents x i A probability of belonging to a non-noise sample; (1-u) i ) Represents x i A probability of belonging to a noise sample; the fuzzy variable m belongs to [1, ∞) represents a weighting index; e (x) i ) Representing the original spectrum x i Extracting its features and reconstructing its spectrum
Is greater than or equal to>
Step three, establishing a pesticide residue and residue classification model by using a support vector machine based on a Gaussian radial basis function as a classifier and combining a weight coefficient;
acquiring a near infrared spectrum of a sample to be detected as test sample data, and mapping the corresponding data to a feature space by using a feature matrix w after second-order derivation; and classifying and judging the mapped data by using a binary classification model to finish detection.
2. The method for detecting pesticide residue based on near infrared spectral data feature extraction of claim 1,the method is characterized in that: continuous variable u in the step (two) i The formula of (c) is as follows:
calculating the available u according to the formula of the constraint condition i The expression is as follows:
according to the calculated u i The formula of the simplified optimization equation J (w) is:
3. the method for detecting pesticide residue based on near infrared spectral data feature extraction of claim 1, characterized by comprising the following steps: the training sample set comprises a plurality of samples with pesticide residues and without pesticide residues.
4. The method for detecting pesticide residue based on near infrared spectral data feature extraction of claim 1, characterized by comprising the following steps: in the step (two), the calculation steps of the feature matrix w and the regularization parameter η are as follows:
step (1), setting the iteration times T as 1, the limit value of the iteration times as T, and learning coefficient alpha 0 ∈(0，1]；
Step (2), when T is smaller than T, executing step (3) to step (9);
step (3) of setting i =1, σ =0, and learning rate α t The expression of (a) is as follows:
α t ＝α 0 (1-1/T)
step (4), when i is less than n, executing step (5) to step (8);
step (5), calculating y = w T x i ,h＝yw,ν＝w T h
And (6) calculating a weight coefficient, wherein an updated iterative formula is as follows:
step (7), the iterative formula of sigma update is as follows:
σ new ＝σ old +e(x i )
step (8), updating the value of i, and enabling i = i +1;
step (9), updating the value of t, and enabling t = t +1; calculating a regularization parameter η, which is as follows:
η＝(σ/n)。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910924188.1A CN110689076B (en) | 2019-09-27 | 2019-09-27 | Pesticide residue detection method based on near infrared spectrum data characteristic extraction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910924188.1A CN110689076B (en) | 2019-09-27 | 2019-09-27 | Pesticide residue detection method based on near infrared spectrum data characteristic extraction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110689076A CN110689076A (en) | 2020-01-14 |
| CN110689076B true CN110689076B (en) | 2023-04-07 |
Family
ID=69110734
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910924188.1A Active CN110689076B (en) | 2019-09-27 | 2019-09-27 | Pesticide residue detection method based on near infrared spectrum data characteristic extraction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110689076B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111337449B (en) * | 2020-02-29 | 2021-03-26 | 同济大学 | Aptamer infrared spectrum sensor for detecting ATZ and detection method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6280381B1 (en) * | 1999-07-22 | 2001-08-28 | Instrumentation Metrics, Inc. | Intelligent system for noninvasive blood analyte prediction |
| CN110118749A (en) * | 2019-06-06 | 2019-08-13 | 南京林业大学 | A kind of garden stuff pesticide residue detection method based near infrared spectrum |
-
2019
- 2019-09-27 CN CN201910924188.1A patent/CN110689076B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6280381B1 (en) * | 1999-07-22 | 2001-08-28 | Instrumentation Metrics, Inc. | Intelligent system for noninvasive blood analyte prediction |
| CN110118749A (en) * | 2019-06-06 | 2019-08-13 | 南京林业大学 | A kind of garden stuff pesticide residue detection method based near infrared spectrum |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110689076A (en) | 2020-01-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ninomiya | High-throughput field crop phenotyping: current status and challenges | |
| CN103048273A (en) | Fruit near infrared spectrum sorting method based on fuzzy clustering | |
| Zhang et al. | Real-time monitoring of optimum timing for harvesting fresh tea leaves based on machine vision | |
| CN111340132B (en) | Machine olfaction mode identification method based on DA-SVM | |
| CN107066959A (en) | A kind of hyperspectral image classification method based on Steerable filter and linear space correlation information | |
| CN107967694A (en) | A kind of EO-1 hyperion object detection method, system, storage medium and processor based on feedback abundance constraint | |
| CN110689076B (en) | Pesticide residue detection method based on near infrared spectrum data characteristic extraction | |
| CN110208211B (en) | Near infrared spectrum noise reduction method for pesticide residue detection | |
| CN109145685B (en) | Fruit and vegetable hyperspectral quality detection method based on ensemble learning | |
| CN113033066A (en) | Method for establishing near infrared spectrum identification model of sargassum fusiforme production area, strain and cultivation mode and identification method | |
| CN107192686B (en) | Method for identifying possible fuzzy clustering tea varieties by fuzzy covariance matrix | |
| Granitto et al. | Boosting classifiers for weed seeds identification | |
| CN107527001A (en) | A kind of hyperspectral image classification method based on Steerable filter and linear space correlation information | |
| Zhou et al. | Green plums surface defect detection based on deep learning methods | |
| CN114112983A (en) | Python data fusion-based Tibetan medicine all-leaf artemisia rupestris L producing area distinguishing method | |
| CN113077452A (en) | Apple tree pest and disease detection method based on DNN network and spot detection algorithm | |
| CN110542659B (en) | Pearl luster detection method based on visible light spectrum | |
| CN117132843A (en) | Wild ginseng, under-forest mountain ginseng and garden ginseng in-situ identification method, system and related equipment | |
| CN111104876A (en) | Infrared spectrum deconvolution method based on neural network | |
| CN115910223A (en) | Irradiation white spirit production process optimization method based on PLS-LSboost gradient lifting tree | |
| CN112881323B (en) | Quality evaluation method of cigarette mainstream smoke | |
| Feng et al. | Research of image recognition of plant diseases and pests based on deep learning | |
| Pusadan et al. | k-Nearest Neighbor and Feature Extraction on Detection of Pest and Diseases of Cocoa | |
| CN111595802A (en) | Construction method and application of Clinacanthus nutans seed source place classification model based on NIR (near infrared spectroscopy) | |
| CN112926450A (en) | Leaf disease feature identification method based on deep belief network |
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 |