CN107463942B - Method for grading quality of juicy peaches based on anti-noise support vector machine with boundary points - Google Patents
Method for grading quality of juicy peaches based on anti-noise support vector machine with boundary points Download PDFInfo
- Publication number
- CN107463942B CN107463942B CN201710541051.9A CN201710541051A CN107463942B CN 107463942 B CN107463942 B CN 107463942B CN 201710541051 A CN201710541051 A CN 201710541051A CN 107463942 B CN107463942 B CN 107463942B
- Authority
- CN
- China
- Prior art keywords
- visible
- near infrared
- infrared spectrum
- juicy
- peaches
- 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
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
- 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
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2323—Non-hierarchical techniques based on graph theory, e.g. minimum spanning trees [MST] or graph cuts
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Discrete Mathematics (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a juicy peach quality grading method based on an anti-noise support vector machine with boundary points, which comprises the following steps: (1) collecting visible/near infrared spectrum data of the juicy peaches with different quality grades, and preprocessing the visible/near infrared spectrum data and extracting PCA (principal component analysis) characteristics; (2) combining the visible/near infrared spectrum sample sets of the juicy peaches with different quality grades pairwise to establish a plurality of groups of visible/near infrared spectrum training sets of the juicy peaches; (3) training the training set by using an anti-noise support vector machine based on boundary points to obtain a plurality of honey peach quality grade classifiers; (4) and detecting the visible/near infrared spectrum of the honey peaches to be classified by using a honey peach quality grade classifier. The method uses the anti-noise support vector machine based on the boundary points to detect the visible/near infrared spectrum of the honey peaches, has the advantages of high detection speed, strong anti-noise capability, high classification accuracy and the like, and can realize the grading of the quality grade of the honey peaches in a noise detection environment.
Description
Technical Field
The invention relates to the field of quality grade identification of juicy peaches, in particular to a juicy peach quality grade grading method based on an anti-noise support vector machine of boundary points.
Background
The peaches are native to China, have good ornamental value and very good edibility, and are rich in protein, fat, sugar, calcium, phosphorus, iron and various vitamins, particularly the iron content of the fruits named as the top cogongrass. In particular to honey peaches which are soft and tender in pulp, sweet and juicy, and strong in fragrance and are reputed to be inside and outside the sea. Meanwhile, from international trade, southeast Asia is taken as one of two fresh peach markets in the world, and the good peach consumption habit and the wide market of the southeast Asia become the power for promoting the development of the honey peach industry in China. However, the production technology level of honey peaches in China is low, a great distance is left between standard production and standard management, and the quality grading operation of honey peaches is basically processed in a manual mode. However, the honey peaches containing crude fiber, little meat, fine and soft texture, big head, enough water and soft skin have the following defects in the manual quality grade identification: 1) the manual method often leaves fingerprints on the honey peaches, and due to oxidation, the fingerprints become dark in color, so that the appearance and quality of the honey peaches are affected; 2) the manual quality grading method has strong subjectivity and low standardization degree; 3) the time for the honey peaches to mature and come into the market is concentrated and is not easy to store, and the manual quality grading method is long in time consumption and not beneficial to the circulation of the honey peaches from the production link to the logistics link. Therefore, the research on the simple, rapid and nondestructive automatic identification method for grading the quality of the honey peaches not only has great economic value, but also has important significance for the healthy and continuous development of the honey peach industry.
The visible/near infrared spectrum analysis technology is qualitative and quantitative analysis technology according to the absorption characteristic of a certain substance component to electromagnetic waves, and has the advantages of high speed, high efficiency, stable result, good test reproducibility, no damage and the like. Pan Lei Qing et al establish a method for detecting sugar content of honey peaches in the shelf life without damage by using a near infrared spectrum technology and a partial least square method in near infrared spectrum detection of sugar content of honey peaches in the shelf life (university of Nanjing agriculture, 4 months in 2013); lixiaoli et al, in New method for discriminating varieties of honey peaches based on principal component analysis and multi-class discriminant analysis (academic newspaper of infrared and millimeter waves, 12 months in 2009), measure spectral curves of honey peaches by using a visible-near infrared spectrometer, perform cluster analysis on samples of different varieties by using a principal component analysis method, and establish a model for discriminating varieties of honey peaches by combining a multi-class discriminant analysis technology.
The juicy peach identification models of Panyiqing, Liaoli and the like are all established under the condition that spectral data are noiseless, and are not suitable for the noisy juicy peach spectral data. However, the visible/near infrared spectrum belongs to an absolute measurement technology, and is easily affected by noise, and the noise generally has a large influence on the measurement result. The Zhao-Huan-huan-et al in the article "influence of noise on near infrared spectroscopy analysis and corresponding mathematical processing methods" (spectroscopy and spectral analysis, 4 months 2013) take corn kernels as an example, analyze feasibility of near infrared analysis on a near infrared spectrum analyzer, and point out that certain software technology is needed for quantitative analysis of spectral data acquired by a visible/near infrared spectrum analyzer with low signal-to-noise ratio. Furthermore, Rukshan Batuwita et al, FSVM-CIL, Fuzzy Support vectors for ClassImbalance Learning (IEEE TRANSACTIONS ON FUZZY SYSTEMS, 6.2010), noted that the data collected in real world environments was largely more or less noisy.
Aiming at the current situation and a plurality of defects of the honey peach quality grading method and effectively being applicable to grading detection of daily production links, the invention provides the honey peach quality grading method based on the anti-noise support vector machine of the boundary points.
Disclosure of Invention
The invention mainly aims to provide a method for grading the quality of honey peaches based on a boundary point anti-noise support vector machine, which is lossless, easy to operate, high in reliability and high in quality, and mainly relieves the influence of errors caused by noise on the existing visible/near infrared qualitative analysis model, and improves the accuracy and the anti-noise performance of the method for grading the quality of the honey peaches.
The technical scheme of the invention is as follows: in order to achieve the purpose, spectral data of the honey peaches are obtained through a visible/near infrared spectrum, the spectral data are subjected to preprocessing and feature extraction by a PCA method to obtain feature data of the spectral data to form training samples, an anti-noise support vector machine based on boundary points is used for training a training sample set to construct a plurality of honey peach quality grade classifiers, and finally a voting mechanism is adopted to count detection results to realize automatic detection of the honey peach quality grades.
According to the conception, the invention adopts the following technical scheme.
(1) Collecting n visible/near infrared spectrum data of the juicy peaches with different quality grades, and preprocessing the visible/near infrared spectrum data and extracting PCA (principal component analysis) characteristics to obtain visible/near infrared spectrum sample sets of the juicy peaches with different quality grades;
(2) combining the visible/near infrared spectrum sample sets of the juicy peaches with different quality grades pairwise to establishGrouping a visible/near infrared spectrum training set of the juicy peaches;
(3) mixing the aboveInputting the visible/near infrared spectrum training set of the juicy peaches into an anti-noise support vector machine based on boundary points for training to obtainA honey peach quality grade classifier;
(4) detecting a honey peach sample to be graded by using a honey peach quality grade classifier;
the construction steps of the juicy peach quality grade classifier in the step (3) are as follows:
(31) juicy peach visible/near infrared spectrum sample set X with two quality grades in each group of juicy peach visible/near infrared spectrum training set1And X2The sphere center c of the minimum sphere space closed region of the two types of sample sets under the feature space is obtained by respectively using a support vector field description (SVDD) algorithm1And c2And a set of sample points Z distributed over two minimum spherical spatial closed regions1And Z2And is provided with X1And X2Set of boundary points B in feature space1And B2Are each Z1And Z2;
(32) At X1Each sample x is calculated1,iTo the center of the sphere c1Euclidean distance d in feature spacei:
di=||φ(x1,i)-φ(c1)||2, (1)
Wherein x1,iSatisfy x1,i∈X1And isPhi () represents the mapping function of the samples from the original space to the kernel space;
at X2Calculate each sample x2,iTo the center of the sphere c2Euclidean distance e in feature spaceiWherein x is2,iSatisfy x2,i∈X2And is
ei=||φ(x2,i)-φ(c2)||2; (2)
(33) According to diValue descending order X1Sample x in (1)1,iWherein x is1,iSatisfy x1,i∈X1And isSequentially mixing the samples x1,iSolving the vector mu by substituting formula (3), and then calculating x1,iUpdating the set of boundary points B together with the obtained μ values for formula (4)1This process continues through traversal X1All of the samples in (1) that satisfy the condition,
wherein | B1I represents B1The number of the middle samples, the threshold value is a normal number;
according to eiValue descending order X2Sample x in (1)2,iWherein x is2,iSatisfy x2,i∈X2And isSequentially mixing the samples x2,iSolving for vector λ by substituting equation (5), and then calculating x2,iSubstituting the obtained lambda value into equation (6) to update the boundary point set B2This process continues through traversal X2All of the samples in (1) that satisfy the condition,
(34) calculating X1Middle set of boundary points B1Weight α of each boundary point ini:
Calculating X2Middle set of boundary points B2Weight β of each boundary point ini:
(35) B is to be1Lumped weight αiRemoving samples with square value smaller than threshold value1Collecting:
b is to be2Lumped weight βiRemoving samples with square value smaller than threshold value2Collecting:
(36) B obtained in the step (35)1And B2Substituting the set with its class label into a median loss function support vector machine:
wherein w is a weight vector, C is a penalty parameter,is a relaxation vector, yi∈ { +1, -1} respectively represents the class labels of two classes of quality grade honey peach visible/near infrared spectrum samples, b is the offset of the classification hyperplane, and tau is one in [0,1 { ]]Real numbers in between;
introducing a Lagrangian function, the above equation can be converted into a quadratic programming form as follows:
wherein γ and ν are lagrangian coefficients;
solving the equations (11) and (12) to obtain the optimal solution w of w and b*,b*And obtaining a juicy peach quality grade classifier of the anti-noise support vector machine based on the boundary points:
f(x)=sign(1-(w*·φ(x)+b*)), (13)
wherein sign () is a sign function;
the method for detecting the juicy peach sample to be graded by using the juicy peach quality grade classifier in the step (4) comprises the following specific steps:
(41) acquiring visible/near infrared spectrum data of the juicy peaches to be detected;
(42) preprocessing the acquired visible/near infrared spectrum data and extracting features by using a PCA method;
(43) inputting the extracted feature data into step (3)Is/are as followsObtaining in a honey peach quality grade classifierDetermining a result;
(44) statistics by voting mechanismAnd (4) judging the result, and taking the grade which accounts for the most in all the results as the quality grade of the juicy peaches to be detected.
The invention has the beneficial effects that:
1) analyzing the appearance quality information of the juicy peaches by using a visible/near infrared spectrum technology, and having the advantages of no use of any chemical reagent, no environmental pollution, no damage to the juicy peaches and no influence on the sale of the juicy peaches;
2) quality grade identification is carried out on the visible/near infrared spectrum data of the juicy peaches by using an anti-noise support vector machine based on boundary points, the identification process is not more than 0.1 second, the identification speed is high, and the method is suitable for the field of large-scale standardized production;
3) the designed anti-noise support vector machine based on the boundary points maps the sample set to a high-dimensional feature space by using a kernel technology, noise samples are effectively removed by identifying the geometric outline of the sample set in the feature space, and a quantile loss function insensitive to noise is used in the selection of the loss function in the support vector machine, so that the accuracy and the anti-noise performance of the honey peach quality grading method are improved.
Drawings
FIG. 1 is a general flow chart of the classification of honey peach quality of an anti-noise support vector machine based on boundary points of the present invention;
FIG. 2 is a graph of the visible/near infrared spectra of honey peaches in the examples;
fig. 3 is a flow chart illustrating the construction of the quality grade classifier for peaches in step (3) of fig. 1.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings 1 in conjunction with the specific examples.
In this embodiment, Yanghu honey peaches produced in Changzhou city of Jiangsu province are selected as research objects, and 490 Yanghu honey peaches of different quality grades are collected according to 8 quality indexes of single fruit weight, fruit shape index, skin color, meat quality, sugar content, soluble solid content, titratable acid and edibility, wherein 80 super-grade peaches, 150 first-grade peaches, 160 second-grade peaches and 100 third-grade peaches are collected.
(1) Collecting the visible/near infrared spectrum data of n juicy peaches with different quality grades, and performing feature extraction on the visible/near infrared spectrum data by adopting a PCA method to obtain visible/near infrared spectrum sample sets of the juicy peaches with different quality grades, wherein the method comprises the following specific steps:
(11) collecting visible/near infrared spectrum data of 490 juicy peaches in total from 4 different quality grades: in this embodiment, a Handheld field spec spectrometer from american application testing device (asd) is used, the Spectral measurement range is 325-1075 nm, the sampling interval is 1.5nm, the resolution is 3.5nm, the probe field angle is 20 degrees, the light source adopts a 14.5V halogen lamp, the Spectral data is sampled after the Spectral scanning is stable, each honey peach is scanned 50 times along the equatorial track, the visible/near infrared Spectral data of the honey peach and the quality grade information thereof are stored, and the average value thereof is taken as the final experimental data, wherein the visible/near infrared Spectral curves of the honey peach with 4 different quality grades are shown in fig. 2.
(12) Preprocessing the acquired visible/near infrared spectrum data of the juicy peaches: in order to eliminate or reduce the interference of high frequency random noise, baseline drift, light scattering and the like on the spectral data, in the embodiment, the spectral data of the 350-1050nm waveband is intercepted and analyzed, and the ASD View SpecPro, UnscambleBlerV 9.6 and DPS (data processing system for reactive stability) are adopted as analysis software. Firstly, adopting a Savitzky-Golay convolution smoothing method, selecting a smoothing point number as 3, carrying out SNV (standard normal variance) processing, and then using a first-order Savitzky-Golay derivative method for processing, wherein a derivative interval is 3. And finally, processing the spectral data by using vector normalization.
(13) Under the condition that the accumulated credibility of the main components is more than or equal to 99.95%, the dimensionality reduction treatment is carried out on 490 juicy peach visible/near infrared spectrum samples by adopting the PCA method, the number of the obtained main factors is 8, and therefore the dimensionality of the juicy peach visible/near infrared spectrum data is reduced to 8 dimensions.
(2) Combining the visible/near infrared spectrum sample sets of the juicy peaches with different quality grades pairwise to establishThe method comprises the following steps of:
in this embodiment, there are 4 kinds of quality grades of honey peaches, and the combination of each two honey peaches is 6 combinations: constructing 6 groups of visible/near infrared spectrum training sets of the juicy peaches according to the 6 combinations, wherein each group of training sets consists of corresponding visible/near infrared spectrum samples of the juicy peaches with two quality grades;
(3) mixing the aboveInputting the visible/near infrared spectrum training set of the juicy peaches into an anti-noise support vector machine based on boundary points for training to obtainThe honey peach quality grade classifier comprises the following specific steps:
(31) juicy peach visible/near infrared spectrum sample set X with two quality grades in each juicy peach visible/near infrared spectrum training set1And X2The sphere center c of the minimum sphere space closed region of the two types of sample sets under the feature space is obtained by respectively using a support vector field description (SVDD) algorithm1And c2And a set of sample points Z distributed over two spherical closed-space regions1And Z2And is provided with X1And X2In a feature spaceSet of boundary points B1And B2Is Z as an initial value1And Z2Wherein, the expression of the SVDD algorithm is as follows:
in the formula (1)A certain quality grade honey peach visible/near infrared spectrum sample set, c is the sphere center of the minimum spherical space closed region under the characteristic space, r is the sphere radius of the minimum spherical space, xiIs the visible/near infrared spectrum sample of the ith juicy peach,represents the mapping function of the sample from the original space to the kernel space, the mapping function used by phi () in this embodiment is a gaussian kernel function, and the value range of the kernel parameter is {0.001,0.1,0.1,1,10,100,1000 };
(32) at X1Each sample x is calculated1,iTo the center of the sphere c1Euclidean distance d in feature spacei:
di=||φ(x1,i)-φ(c1)||2, (2)
Wherein x1,iSatisfy x1,i∈X1And isPhi () represents the mapping function of the samples from the original space to the kernel space;
at X2Calculate each sample x2,iTo the center of the sphere c2Euclidean distance e in feature spaceiWherein x is2,iSatisfy x2,i∈X2And is
ei=||φ(x2,i)-φ(c2)||2; (3)
(33) According to diValue descending order X1Sample x in (1)1,iWherein x is1,iSatisfy x1,i∈X1And isSequentially mixing the samples x1,iSolving the vector mu by substituting formula (4), and then calculating x1,iUpdating the set of boundary points B together with the obtained μ values for formula (5)1This process continues through traversal X1All of the samples in (1) that satisfy the condition,
wherein | B1I represents B1The number of the middle samples, the threshold value is a normal number;
according to eiValue descending order X2Sample x in (1)2,iWherein x is2,iSatisfy x2,i∈X2And isSequentially mixing the samples x2,iSolving for vector λ by substituting equation (6), and then calculating xiUpdating the set of boundary points B together with the resulting lambda value substituting equation (7)2This process continues through traversal X2All of the samples in (1) that satisfy the condition,
(34) calculating X1Middle set of boundary points B1Weight α of each boundary point ini:
Calculating X2Middle set of boundary points B2Weight β of each boundary point ini:
(35) B is to be1Lumped weight αiRemoving samples with square value smaller than threshold value1Collecting:
b is to be2Lumped weight βiRemoving samples with square value smaller than threshold value2Collecting:
(36) B obtained in the step (35)1And B2The pooled samples are substituted into a median loss function support vector machine along with their class labels:
wherein xiIs B1And B2Union of (A) and (B)1∪B2W is a weight vector, the penalty parameter C obtains an optimum value by optimizing in a grid {0.001,0.1,0.1,1,10,100,1000},is a relaxation vector, yi∈ { +1, -1} respectively represents the class labels of two classes of quality grade honey peach visible/near infrared spectrum samples, b is the offset of the classification hyperplane, and tau is one in [0,1 { ]]The real number between, in this embodiment, τ is 0.05, Φ (x)i) The middle mapping function uses a Gaussian kernel function, and uses the same kernel parameters as the formula (1);
introducing a Lagrangian function, the above equation can be converted into a quadratic programming form as follows:
wherein γ and ν are lagrangian coefficients;
obtaining w and b optimal solution w by solving equations (12) and (13)*,b*Obtaining a juicy peach quality grade classifier of the anti-noise support vector machine based on the boundary points:
f(x)=sign(1-|w*·φ(x)+b*|), (14)
wherein sign () is a sign function;
(4) as shown in fig. 3, the method for detecting the juicy peach sample to be classified and detected by using the juicy peach quality classification classifier comprises the following specific steps:
(41) in the embodiment, 30 samples of each of the super-grade, first-grade, second-grade and third-grade honey peaches are collected again, and visible/near infrared spectrum data of the honey peaches to be detected in a grading manner are obtained by using a visible/near infrared spectrometer;
(42) preprocessing the acquired visible/near infrared spectrum data, and performing feature extraction on the visible/near infrared spectrum data by using a PCA (principal component analysis) method to obtain 8-dimensional feature data;
(43) respectively inputting the obtained 8-dimensional feature vectors into the 6 juicy peach quality grade classifiers obtained in the step (3) to obtain 6 detection results,
for example: when the feature data of a super juicy peach is input into the super classifier and the first classifier, the detection result is as follows: a special grade; when the test result is input into the super classifier and the second classifier, the test result is as follows: a special grade; when the detection result is input into the super classifier and the third classifier, the detection result is as follows: a special grade; when the signals are input into the first-stage classifier and the second-stage classifier, the detection result is as follows: a first stage; when the signals are input into the first-level classifier and the third-level classifier, the detection results are as follows: a first stage; when the detection result is input into the second-level classifier and the third-level classifier, the detection result is as follows: second-stage;
(44) and counting 6 detection results by adopting a voting mechanism, wherein the highest occupation rate of the special grade in the 6 results is the grade, and the quality grade of the juicy peaches is judged to be the special grade.
The determination accuracy (%) of this example is shown in table 2, and the results of the method of the present invention were compared with those obtained by using a hinge loss function support vector machine (L1-SVM), a partial least squares method, and a three-layer BP neural network method, and the experimental platform was MATLAB 2009 (a).
Table 2: the method, the support vector machine, the partial least square method and the three-layer BP neural network of the invention have the classification judgment accuracy rate (%)
Quality grade of honey peach | The method of the invention | Support vector machine | Partial least squares method | Three-layer BP neural network |
Special class | 100 | 96.67 | 93.33 | 93.33 |
First stage | 100 | 93.33 | 90.00 | 93.33 |
Second stage | 100 | 90.00 | 86.76 | 93.33 |
Three-stage | 100 | 90.00 | 86.76 | 90.00 |
The above examples are intended to be illustrative of the present invention and are not to be construed as limiting the invention. Those skilled in the art can make various other modifications and alterations without departing from the spirit of the invention in light of the teachings of the present disclosure, and such modifications and alterations are intended to be included within the scope of the invention.
Claims (1)
1. A juicy peach quality grading method based on an anti-noise support vector machine of boundary points is characterized by comprising the following steps:
(1) collecting n visible/near infrared spectrum data of the juicy peaches with different quality grades, and preprocessing the visible/near infrared spectrum data and extracting PCA (principal component analysis) characteristics to obtain visible/near infrared spectrum sample sets of the juicy peaches with different quality grades;
(2) combining the visible/near infrared spectrum sample sets of the juicy peaches with different quality grades pairwise to establishGrouping a visible/near infrared spectrum training set of the juicy peaches;
(3) mixing the aboveInputting the visible/near infrared spectrum training set of the juicy peaches into an anti-noise support vector machine based on boundary points for training to obtainA honey peach quality grade classifier;
(4) detecting a honey peach sample to be graded by using a honey peach quality grade classifier;
the construction steps of the juicy peach quality grade classifier in the step (3) are as follows:
(31) juicy peach visible/near infrared spectrum sample set X with two quality grades in each group of juicy peach visible/near infrared spectrum training set1And X2The sphere center c of the minimum sphere space closed region of the two types of sample sets under the feature space is obtained by respectively using a support vector field description (SVDD) algorithm1And c2And a set of sample points Z distributed over two minimum spherical spatial closed regions1And Z2And is provided with X1And X2Set of boundary points B in feature space1And B2Are each Z1And Z2;
(32) At X1Each sample x is calculated1,iTo the center of the sphere c1Euclidean distance d in feature spacei:
di=||φ(x1,i)-φ(c1)||2, (1)
Wherein x1,iSatisfy x1,i∈X1And isPhi () represents the mapping function of the samples from the original space to the kernel space,
at X2Calculate each sample x2,iTo the center of the sphere c2Euclidean distance e in feature spacei:
ei=||φ(x2,i)-φ(c2)||2, (2)
(33) According to diValue descending order X1Sample x in (1)1,iWherein x is1,iSatisfy x1,i∈X1And isSequentially mixing the samples x1,iSolving the vector mu by substituting formula (3), and then calculating x1,iUpdating the set of boundary points B together with the obtained μ values for formula (4)1This process continues through traversal X1All of the samples in (1) that satisfy the condition,
wherein | B1I represents B1The number of the middle samples, the threshold value is a normal number;
according to eiValue descending order X2Sample x in (1)2,iWherein x is2,iSatisfy x2,i∈X2And isSequentially mixing the samples x2,iSolving for vector λ by substituting equation (5), and then calculating x2,iSubstituting the obtained lambda value into equation (6) to update the boundary point set B2This process continues through traversal X2All of the samples in (1) that satisfy the condition,
wherein | B2I represents B2The number of the middle samples;
(34) calculating X1Middle set of boundary points B1Weight α of each boundary point ini:
Calculating X2Middle set of boundary points B2Weight β of each boundary point ini:
(35) B is to be1Lumped weight αiRemoving samples with square value smaller than threshold value1Collecting:
b is to be2Lumped weight βiRemoving samples with square value smaller than threshold value2Collecting:
(36) B obtained in the step (35)1And B2Substituting the set with the class label of its sample into a median loss function support vector machine:
wherein w is a weight vector, C is a penalty parameter,is a relaxation vector, yi∈ { +1, -1} respectively represents the class labels of two classes of quality grade honey peach visible/near infrared spectrum samples, b is the offset of the classification hyperplane, and tau is one in [0,1 { ]]Real numbers in between;
introducing a Lagrangian function, the above equation can be converted into a quadratic programming form as follows:
wherein γ and ν are lagrangian coefficients;
solving the equations (11) and (12) to obtain the optimal solution w of w and b*,b*And obtaining a juicy peach quality grade classifier of the anti-noise support vector machine based on the boundary points:
f(x)=sign(1-(w*·φ(x)+b*)), (13)
wherein sign () is a sign function;
the method for detecting the juicy peach sample to be graded by using the juicy peach quality grade classifier in the step (4) comprises the following specific steps:
(41) acquiring visible/near infrared spectrum data of the juicy peaches to be detected;
(42) preprocessing the acquired visible/near infrared spectrum data and extracting features by using a PCA method;
(43) inputting the extracted feature data into step (3)Obtaining in a honey peach quality grade classifierDetermining a result;
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710541051.9A CN107463942B (en) | 2017-07-05 | 2017-07-05 | Method for grading quality of juicy peaches based on anti-noise support vector machine with boundary points |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710541051.9A CN107463942B (en) | 2017-07-05 | 2017-07-05 | Method for grading quality of juicy peaches based on anti-noise support vector machine with boundary points |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107463942A CN107463942A (en) | 2017-12-12 |
CN107463942B true CN107463942B (en) | 2020-09-08 |
Family
ID=60546561
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710541051.9A Active CN107463942B (en) | 2017-07-05 | 2017-07-05 | Method for grading quality of juicy peaches based on anti-noise support vector machine with boundary points |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107463942B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101564328A (en) * | 2009-05-07 | 2009-10-28 | 杭州电子科技大学 | Laptop artificial limb multi-movement-mode identifying method based on support vector data description |
CN101571844A (en) * | 2009-06-10 | 2009-11-04 | 北京工业大学 | Training method of support vector machine for pattern classification |
CN102938060A (en) * | 2012-12-07 | 2013-02-20 | 上海电机学院 | Dynamic gesture recognition system and method |
CN104318260A (en) * | 2014-10-28 | 2015-01-28 | 常州大学 | Fur near-infrared spectral discrimination method based on packet support vector machine |
-
2017
- 2017-07-05 CN CN201710541051.9A patent/CN107463942B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101564328A (en) * | 2009-05-07 | 2009-10-28 | 杭州电子科技大学 | Laptop artificial limb multi-movement-mode identifying method based on support vector data description |
CN101571844A (en) * | 2009-06-10 | 2009-11-04 | 北京工业大学 | Training method of support vector machine for pattern classification |
CN102938060A (en) * | 2012-12-07 | 2013-02-20 | 上海电机学院 | Dynamic gesture recognition system and method |
CN104318260A (en) * | 2014-10-28 | 2015-01-28 | 常州大学 | Fur near-infrared spectral discrimination method based on packet support vector machine |
Non-Patent Citations (3)
Title |
---|
FSVM-CIL: Fuzzy Support Vector Machines for Class Imbalance Learning;Rukshan Batuwita 等;《IEEE TRANSACTIONS ON FUZZY SYSTEMS》;20100630;第18卷(第3期);558-571 * |
基于主成分和多类判别分析的可见-红外光谱水蜜桃品种鉴别新方法;李晓丽 等;《红外与毫米波学报》;20061230;第25卷(第6期);417-420 * |
核最近邻凸包分类算法;周晓飞 等;《中国图象图形学报》;20070715;第12卷(第7期);1209-1213 * |
Also Published As
Publication number | Publication date |
---|---|
CN107463942A (en) | 2017-12-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Xing et al. | Combination of chemometric tools and image processing for bruise detection on apples | |
Theanjumpol et al. | Non-destructive identification and estimation of granulation in ‘Sai Num Pung’tangerine fruit using near infrared spectroscopy and chemometrics | |
CN103048273B (en) | Fruit near infrared spectrum sorting method based on fuzzy clustering | |
Zhang et al. | Detection of apple defects based on the FCM-NPGA and a multivariate image analysis | |
Antonucci et al. | Quantitative method for shape description of almond cultivars (Prunus amygdalus Batsch) | |
Toher et al. | A comparison of model-based and regression classification techniques applied to near infrared spectroscopic data in food authentication studies | |
CN113310936A (en) | Rapid identification method for four high-temperature sterilized commercial milks | |
Wu et al. | Discrimination of apples using near infrared spectroscopy and sorting discriminant analysis | |
Sharma et al. | Rapid ripening stage classification and dry matter prediction of durian pulp using a pushbroom near infrared hyperspectral imaging system | |
CN103528967A (en) | Hyperspectral image based overripe Lonicera edulis fruit identification method | |
CN113310937A (en) | Method for rapidly identifying high-temperature sterilized milk, pasteurized fresh milk of dairy cow and reconstituted milk of milk powder | |
Raihen et al. | Prediction modeling using deep learning for the classification of grape-type dried fruits | |
CN113310934A (en) | Method for quickly identifying milk cow milk mixed in camel milk and mixing proportion thereof | |
Tang et al. | A study on nitrogen concentration detection model of rubber leaf based on spatial-spectral information with NIR hyperspectral data | |
CN107463942B (en) | Method for grading quality of juicy peaches based on anti-noise support vector machine with boundary points | |
Zhou et al. | Hyperspectral imaging technology for detection of moisture content of tomato leaves | |
Fu et al. | Discrimination of pear varieties using three classification methods based on near-infrared spectroscopy | |
CN112801173B (en) | Lettuce near infrared spectrum classification method based on QR fuzzy discriminant analysis | |
CN113310933A (en) | Spectrum identification method for number of days for storing raw buffalo milk | |
Behera et al. | A novel approach for classification of apple using on-tree images based on image processing | |
CN111595805A (en) | Possibility-clustering Chinese cabbage pesticide residue qualitative analysis method | |
ZHANG et al. | Nutrient deficiency diagnosis in whole hydroponic lettuce based on random forest. | |
Wu et al. | Identification of lambda-cyhalothrin residues on Chinese cabbage using fuzzy uncorrelated discriminant vector analysis and MIR spectroscopy | |
CN115060663B (en) | Method for quantitatively testing banana quality based on machine learning | |
Sheu et al. | A study on species identification based on leaf contours of Taiwan Lauraceae and Fagaceae plants |
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 |