KR20180051239A - A intelligent classification method for preprocessing based of a black plastic sorting by raman spectroscopy - Google Patents

A intelligent classification method for preprocessing based of a black plastic sorting by raman spectroscopy Download PDF

Info

Publication number
KR20180051239A
KR20180051239A KR1020160148293A KR20160148293A KR20180051239A KR 20180051239 A KR20180051239 A KR 20180051239A KR 1020160148293 A KR1020160148293 A KR 1020160148293A KR 20160148293 A KR20160148293 A KR 20160148293A KR 20180051239 A KR20180051239 A KR 20180051239A
Authority
KR
South Korea
Prior art keywords
black plastic
data
black
raman spectroscopy
intelligent classification
Prior art date
Application number
KR1020160148293A
Other languages
Korean (ko)
Inventor
오성권
최우진
Original Assignee
수원대학교산학협력단
(주)이오니아이엔티
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 수원대학교산학협력단, (주)이오니아이엔티 filed Critical 수원대학교산학협력단
Priority to KR1020160148293A priority Critical patent/KR20180051239A/en
Publication of KR20180051239A publication Critical patent/KR20180051239A/en

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
    • B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
    • B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
    • B07C5/02—Measures preceding sorting, e.g. arranging articles in a stream orientating
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
    • B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
    • B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
    • B07C5/34—Sorting according to other particular properties
    • B07C5/342—Sorting according to other particular properties according to optical properties, e.g. colour
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B29—WORKING OF PLASTICS; WORKING OF SUBSTANCES IN A PLASTIC STATE IN GENERAL
    • B29B—PREPARATION OR PRETREATMENT OF THE MATERIAL TO BE SHAPED; MAKING GRANULES OR PREFORMS; RECOVERY OF PLASTICS OR OTHER CONSTITUENTS OF WASTE MATERIAL CONTAINING PLASTICS
    • B29B17/00—Recovery of plastics or other constituents of waste material containing plastics
    • B29B17/02—Separating plastics from other materials
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B29—WORKING OF PLASTICS; WORKING OF SUBSTANCES IN A PLASTIC STATE IN GENERAL
    • B29B—PREPARATION OR PRETREATMENT OF THE MATERIAL TO BE SHAPED; MAKING GRANULES OR PREFORMS; RECOVERY OF PLASTICS OR OTHER CONSTITUENTS OF WASTE MATERIAL CONTAINING PLASTICS
    • B29B17/00—Recovery of plastics or other constituents of waste material containing plastics
    • B29B17/02—Separating plastics from other materials
    • B29B2017/0213—Specific separating techniques
    • B29B2017/0282—Specific separating techniques using information associated with the materials, e.g. labels on products

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Environmental & Geological Engineering (AREA)
  • Mechanical Engineering (AREA)
  • Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)

Abstract

The present invention relates to a pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy, and more particularly, to a pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy, which comprises (1) Obtaining black plastic material-specific (PET, PP, PS) spectral data; (2) preprocessing spectral data of the black plastic material obtained through the step (1) through a principal component analysis (PCA) algorithm; (3) inputting the preprocessed data through the step (2) into a support vector machine (SVM) pattern classifier, and learning the chemical properties of the black plastic material by the SVM pattern classifier; And (4) the step of classifying the input black plastics based on the chemical characteristics of each material by using the SVM pattern classifier learned through the steps (1) to (3) And includes the constitutional features thereof.
According to the pretreatment-based intelligent classification method for black plastic sorting by the Raman spectroscopic method proposed in the present invention, spectral data using the chemical characteristics of each material of black plastic is constructed and analyzed using Raman spectroscopic equipment, The high-dimensional data of the extracted characteristic peaks are preliminarily processed at low dimensions through the principal component analysis algorithm and then used as the input variables of the SVM classifier to determine the chemical characteristics It is possible to classify them automatically according to the chemical characteristics of the PET, PP, and PS of the black plastic by configuring the classifying learning through the optimal SVM classifier considered.
According to the present invention, the data considering the chemical characteristics of each material of black plastic using the Raman spectroscopic equipment are constructed, the dimension reduction using the principal component analysis algorithm, and the learning considering the chemical characteristics of each material of the black plastic using the SVM classifier , It is possible to classify the pretreatment-based intelligent classification of PET, PP, and PS of black plastic through black plastic, and to improve the reliability of classifier performance by classifying accurate and quick black plastic materials.

Description

TECHNICAL FIELD [0001] The present invention relates to a pre-processing based intelligent classification method for black plastic sorting by Raman spectroscopy,

The present invention relates to a method of classifying black plastics, and more particularly, to a pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy.

In recent years, numerous plastics have been used extensively in the industrial field, and a large amount of plastic waste is being generated. Research on the recycling of waste plastics has emerged as an important issue to prevent the abandonment of limited useful resources as well as environmental pollution. As such, the recycling of waste plastics is drawing attention from a reuse point of view.

Currently, the recycling center is building and operating plastic sorting system using NIR sensor to classify plastic materials. However, there is a problem that the black plastic still can not be classified. That is, in the case of the black plastic containing carbon black, since it absorbs the light emitted from the NIR equipment due to the characteristic of the black color, it has been limited to classify the plastic into materials by using only near infrared spectroscopy. With respect to such a plastic classification, Korean Patent Laid-Open Publication No. 10-2013-0019818 is disclosed in the prior art document.

The present invention has been proposed in order to solve the above-mentioned problems of the previously proposed methods. Spectral data utilizing the chemical properties of black plastic materials are constructed and analyzed using Raman spectroscopic equipment, The characteristic peaks of chemical characteristics are taken into consideration and the high dimensional data of the extracted characteristic peaks are preliminarily processed at low dimensions through the principal component analysis algorithm and then used as the input variables of the SVM classifier, Processing for black plastic sorting by Raman spectroscopy, which enables classifying automatically according to the chemical characteristics of the materials of PET, PP, PS of black plastic by configuring the classification learning through the optimal SVM classifier Based intelligent classification method.

In addition, the present invention is based on the construction of data considering the chemical characteristics of each material of black plastic using Raman spectroscopy equipment, the learning of chemical characteristics of each material of black plastic using dimension reduction and SVM classifier using principal component analysis algorithm For black plastic sorting by Raman spectroscopy, which makes it possible to classify the pretreatment-based intelligence of PET, PP, and PS of black plastic and to improve the reliability of the classifier performance by sorting the black plastic materials by accurate and fast Another object of the present invention is to provide a pre-processing-based intelligent classification method.

According to an aspect of the present invention, there is provided a pre-processing-based intelligent classification method for black plastic sorting by Raman spectroscopy,

A pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy,

(1) obtaining spectral data for each of black plastic materials (PET, PP, PS) using Raman spectroscopy equipment;

(2) preprocessing spectral data of the black plastic material obtained through the step (1) through a principal component analysis (PCA) algorithm;

(3) inputting the preprocessed data through the step (2) into a support vector machine (SVM) pattern classifier, and learning the chemical properties of the black plastic material by the SVM pattern classifier; And

(4) For the black plastic arbitrarily inputted, the input black plastics are classified based on the chemical characteristics of each material by using the SVM pattern classifier learned through the above-mentioned steps (1) to (3) As a feature of the configuration.

Preferably, in the step (1)

Three characteristic peaks in which the peak point regions are not overlapped in consideration of the chemical properties of the black plastic material (PET, PP, PS) are selected and extracted from spectral data obtained by the black plastic materials (PET, PP, PS) Process.

More preferably, in the step (1)

Dimensional input parameter considering up to intensity values of five wavelength regions on both sides of the characteristic peak of each of the black plastic materials (PET, PP, PS).

Preferably, in the step (2)

Through the principal component analysis algorithm, in step (1), a pre-processing process may be performed in which the chemical characteristics of the black plastic material are taken into consideration and the selected high-dimensional input variables are converted into low-dimensional input variables.

More preferably, the step (2) is a preprocessing process using an algorithm of the principal component analysis method,

(2-1) constructing a set of learning data for recognition;

(2-2) calculating an average of the data set in the set of learning data;

(2-3) calculating a difference between an average of the learning data and the learning data;

(2-4) calculating a covariance matrix from learning data for recognition;

(2-5) calculating an eigenvalue matrix and an eigenvector matrix of a covariance matrix by analyzing eigenvalues and selecting a specific number having the largest eigenvalue; And

(2-6) extracting the reduced feature data by the linear transformation in each of the data and the transformation matrix having a specific number of eigenvectors.

More preferably, in the step (3)

5-fold cross validation for 5-fold cross validation to improve the reliability of the classifier performance of the SVM pattern classifier can be used.

More preferably, the step (4)

(4-1) inputting any plastic;

(4-2) obtaining spectral data of the black plastic inputted through the step (4-1) through the Raman spectroscopic equipment;

(4-3) a step of pre-processing the dimension reduction of spectral data of the black plastic obtained through the step (4-2); And

(4-4) In the step (4-3), the preprocessed data is input to the SVM pattern classifier learned through the steps (1) to (3), and the arbitrarily inputted black plastic is classified into materials . ≪ / RTI >

According to the pretreatment-based intelligent classification method for black plastic sorting by the Raman spectroscopic method proposed in the present invention, spectral data using the chemical characteristics of each material of black plastic is constructed and analyzed using Raman spectroscopic equipment, The high-dimensional data of the extracted characteristic peaks are preliminarily processed by the principal component analysis algorithm and used as the input variables of the SVM classifier to determine the chemical characteristics It is possible to classify them automatically according to the chemical characteristics of the PET, PP, and PS of the black plastic by configuring the classifying learning through the optimal SVM classifier considered.

According to the present invention, the data considering the chemical characteristics of each material of black plastic using the Raman spectroscopic equipment are constructed, the dimension reduction using the principal component analysis algorithm, and the learning considering the chemical characteristics of each material of the black plastic using the SVM classifier , It is possible to classify the pretreatment-based intelligent classification of PET, PP, and PS of black plastic through black plastic, and to improve the reliability of classifier performance by classifying accurate and quick black plastic materials.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a flowchart illustrating a pre-processing-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention. FIG.
FIG. 2 is a diagram illustrating a process of extracting characteristic peaks and generating input parameters in a pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention. FIG.
FIG. 3 is a diagram illustrating a preprocessing process in a pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention. FIG.
FIG. 4 is a diagram illustrating a classification process of arbitrary black plastic in a pre-processing based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention. FIG.
FIG. 5 is a diagram showing Raman scattering, Raman scattering, and chemical structural formulas for black plastic applied to a pretreatment-based intelligent classification method for sorting black plastic by Raman spectroscopy according to an embodiment of the present invention. FIG.
6 is a graph showing spectrum results and graphs of PP among black plastics applied to a pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention.
FIG. 7 is a graph showing spectral results and PS of black plastic applied to a pretreatment-based intelligent classification method for sorting black plastic by Raman spectroscopy according to an embodiment of the present invention; FIG.
8 is a graph showing spectral results and PET of black plastic applied to a pretreatment-based intelligent classification method for sorting black plastic by Raman spectroscopy according to an embodiment of the present invention.
9 is a graph showing a final peak selection graph for black plastic of PET, PP, and PS applied to a pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention.
10 illustrates a principal component analysis method applied to a pretreatment-based intelligent classification method for sorting black plastic by Raman spectroscopy according to an embodiment of the present invention.
11 is a view illustrating an example of points selected for input variables of black plastic of PS applied to a pre-processing-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention.
12 to 14 show experimental results of a classifier and a comparative classifier applied to a pretreatment-based intelligent classifying method for sorting black plastic by Raman spectroscopy according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, in order that those skilled in the art can easily carry out the present invention. In the following detailed description of the preferred embodiments of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. In the drawings, like reference numerals are used throughout the drawings.

In addition, in the entire specification, when a part is referred to as being 'connected' to another part, it may be referred to as 'indirectly connected' not only with 'directly connected' . Also, to "include" an element means that it may include other elements, rather than excluding other elements, unless specifically stated otherwise.

FIG. 1 is a flow chart of a pre-processing-based intelligent classification method for sorting black plastic by Raman spectroscopy according to an embodiment of the present invention. FIG. FIG. 3 is a diagram illustrating a process of extracting characteristic peaks and generating input parameters in a preprocessing-based intelligent classification method for sorting. FIG. 3 is a diagram illustrating a preprocessing-based intelligent classification for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention. FIG. 4 is a diagram illustrating a classification process of black plastic arbitrarily inputted in the pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention. to be. As shown in FIG. 1 to FIG. 4, the pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention includes spectral data of PET, PP, PS, A step S130 of obtaining spectral data for each black plastic material, a step S120 of preparing spectral data for each black plastic material S120, a step S130 learning chemical characteristics of each black plastic material, and a step of classifying black plastics based on their chemical characteristics (S140).

In step S110, spectrum data of the black plastic material (PET, PP, PS) can be obtained using the Raman spectroscopy equipment. In step S110, characteristic peaks in which the peak point areas are not overlapped in consideration of the chemical characteristics of the black plastic materials (PET, PP, PS) are selected from the spectral data obtained by the black plastic materials (PET, PP, PS) And may further include a process of being extracted. In addition, in step S110, the process further includes a step of generating a high-dimensional input parameter considering up to the intensity values of the five wavelength ranges on both sides of the characteristic peak of each of the black plastic materials (PET, PP, PS) .

FIG. 5 is a diagram showing a Raman scattering process, a Raman scattering process, and a chemical formula for black plastic applied to a pretreatment-based intelligent classification method for sorting black plastic by Raman spectroscopy according to an embodiment of the present invention. 5 (a) shows Raman scattering, FIG. 5 (b) shows a Raman scattering process, and FIG. 5 (c) shows a chemical structure of black plastic PP, PS and PET Respectively. In general, Raman spectroscopy was first discovered by CVRaman and KSKrishnan in 1928, observing that when light energy is shot into a medium, light is scattered by vibrations between atoms and interaction between electrons Method. Three types of Raman scattering are shown in FIG. 5 (a), namely, Rayleigh scattering, Stokes Raman scattering, and Anti-Stokes Raman scattering. Rayleigh scattering means that the energy state of the incident light and molecule is not changed but is just as it is. However, unlike Rayleigh scattering, there may be cases such as Stokes and Anti-Stokes in which phonons are absorbed / emitted due to lattice vibration depending on the material, and scattered light is emitted from the air. Stokes is the scattering of energy as low as phonon energy as the phonon is released. It is called Anti-Stokes that phonons are absorbed and energy as high as phonon energy is scattered. FIG. 5 (b) shows the Raman scattering process. In the present invention, characteristics of materials are analyzed with reference to chemical theory in order to find out features that can be classified by materials (PP, PS, PET) of black plastic. As shown in FIG. 5 (c), the chemical structure of PP, PS, and PET in black plastic is chemically the largest in the methyl group (-Ch3) In the case of PS, double bonds (CC) and Benzene ring breathing mode are characteristics of PS compared to PP and PET. In the case of PET, CO-C stretching, carbonyl, ring C1-C4 stretching, and group (C = O) stretching are prominent features.

FIG. 6 is a graph showing spectral results and graphs of PP among black plastics applied to a pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention, and FIG. FIG. 8 is a graph showing spectral results and graphs of PS among black plastics applied to a pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment. FIG. Spectral results and graphs of PET among black plastics applied to pretreatment-based intelligent classification methods for black plastic sorting by spectroscopy. The upper table of FIG. 6 shows the infrared spectra of PP using Raman spectroscopy equipment, and the lower graph shows the average of the spectra of the entire PP data. The upper table in FIG. 7 shows the results of infrared spectra of PS using Raman spectroscopy equipment, and the lower graph shows the average of spectra of the entire PS data. The upper table of FIG. 8 shows the results of infrared spectra of PET using Raman spectroscopy equipment, and the lower graph shows the average of the data spectra of all PETs. In other words, the table of FIG. 6 is a reference data of a paper characterizing the PP using the Raman spectroscopy equipment, and the lower graph is an example of a graph showing 100 data averages for PP. The table of FIG. 7 is a reference data of a paper analyzing the characteristics of PS using Raman spectroscopy equipment, and the lower graph is an example of a graph obtained by averaging 100 data for PS. The table of FIG. 8 is a reference data of a paper characterizing the PET using Raman spectroscopy equipment, and the lower graph is an example of a graph obtained by averaging 100 data for PET.

9 is a graph showing a final peak selection graph for black plastic of PET, PP, and PS applied to a pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention. FIG. 9 is a graph showing peaks of PET, PP, and PS finally defined by analyzing the data of the tables and graphs shown in FIGS. 6 to 8. FIG. In other words, a total of 9 peaks are selected for each black plastic material so that the peak point areas do not overlap with each other considering the chemical characteristics of each material of the black plastic as much as possible. That is, in the case of PP 410㎝ -1 (ωCH 2 + δCH ), 841㎝ -1 (γCH 2 + νC-CH 3), 2871㎝ -1 (νCH 3 sym) selecting a region as a characteristic peak, and 410 - 1 , the graph of the extracted PP data is analyzed and not selected through chemical analysis. In the case of PS, the area of 1001 cm -1 (Benzene ring breathing mode), 1032 cm -1 (CH in plane bending mode), and 1155 cm -1 (CC stretching vibration) are selected as characteristic peak points. The area of 1062 cm -1 corresponding to double bonds (C = C), which is one of the characteristics of PS, is excluded because it is close to the peak of PET of 1613 cm -1 . In addition, regions of 1032 cm -1 and 1155 cm -1 are selected so that regions of the PS graph analysis do not overlap with regions of other materials. In addition, we selected for the case of PET, the 1289㎝ -1, 1613㎝ -1, 1725㎝ -1 peak area as a characteristic point. The characteristics include the 795㎝ -1, -1 858㎝ of PET, but it is the chemical properties of the region close to PP were configured not to be considered a characteristic peak point of PET. In the present invention, spectral data is obtained based on the chemical properties of the materials of the black plastics of PP, PS, and PET, and is used as data for classification and learning.

In step S120, spectral data of black plastic material obtained through step S110 may be preprocessed through a principal component analysis (PCA) algorithm. In this step S120, the preprocessing process of extracting the high-order input variables and converting the selected high-dimensional input variables into the low-dimensional input variables can be performed through consideration of the chemical characteristics of the black plastic materials in step S110 through the principal component analysis algorithm. This principal component analysis is a technique for reducing the high dimensional feature vector to a low dimensional feature vector while maintaining the data information as much as possible. When the input of the large data size is used as the learning and testing of the pattern classifier, It is used to solve the problem of slowing down. Also, since high-order data includes information that may interfere with classification such as noise, the present invention uses the PCA algorithm.

10 is a diagram illustrating a principal component analysis method applied to a pretreatment-based intelligent classification method for sorting black plastic by Raman spectroscopy according to an embodiment of the present invention. In step S120 of the principal component analysis method as shown in FIG. 10, a preprocessing process using an algorithm of the principal component analysis method, as shown in FIG. 3, comprises steps S121 and S122 of constructing a set of learning data S for recognition , and the step (S122) that the data set from a set of learning data (S) calculates the average (Ψ), and the step (S123) of calculating the average of the difference (Φ i) of the learning data and learning data, for recognition Calculating an eigenvalue matrix (?) And an eigenvector matrix U of the covariance matrix (C) through eigenvalue analysis, calculating a covariance matrix (C) in the "step (S125), and the respective data and the specific number (M for selecting the transformation matrix having the eigenvectors) (W = [ω 1, ω 2, ω 3, ..., ω M ']) (M)' And extracting the reduced feature data by the linear transformation (S126).

Equation (1) below represents a relational expression constituting a set (S) of learning data.

Figure pat00001

Equation (2) below represents a relational expression for calculating the average (Ψ) of the data set in the set of learning data (S).

Figure pat00002

Equation (3) below expresses a relational expression for calculating the difference (? I ) between the average of the learning data and the learning data.

Figure pat00003

Equation (4) below expresses a relational expression for calculating a covariance matrix C from learning data for recognition.

Figure pat00004

Equation (5) below calculates a eigenvalue matrix (?) And an eigenvector matrix U of the covariance matrix (C) through eigenvalue analysis and selects a specific number (M ' ) having the largest eigenvalue .

Figure pat00005

[Equation 6] below is reduced due to the linear transformation in the respective data and the "transformation matrix with eigenvectors of the (W = [ω 1, ω 2, ω 3, ..., ω M specific number (M)"]) And extracts the feature data.

Figure pat00006

In step S130, the preprocessed data is input to a SVM pattern classifier, and the SVM pattern classifier learns the chemical properties of the black plastic material. In this step S130, a 5-fold cross validation for the 5-fold cross validation to improve the reliability of the classifier performance of the SVM pattern classifier can be used.

In step S140, the inputted black plastics may be classified based on the chemical characteristics of each material by using the SVM pattern classifier learned through steps S110 to S130 for arbitrarily inputted black plastics. Step S140 includes step S141 of inputting any plastic, step S142 of acquiring spectral data of black plastic inputted through step S141 through the Raman spectroscopic equipment, (Step S143) in which the preprocessing process of dimension reduction of the black plastic spectral data obtained through step S142 is performed, and the preprocessed data through step S143 are input to the SVM pattern classifier learned through steps S110 to S130, And step S144 in which the input black plastic is sorted by material.

11 is a view illustrating an example of points selected for input parameters for black plastic of PS applied to the pre-processing-based intelligent classification method for black plastic sorting by Raman spectroscopy according to an embodiment of the present invention. That is, the data used in the present invention are three types of black plastics such as PET, PP, and PS collected at an actual recycling center. Data of 100 samples are extracted using Raman spectroscopy equipment for each material. 11 shows the peaks selected by chemical theory and data analysis for each black plastic material. Three characteristic peaks were selected for each material, and information about a total of nine peaks was used as an input parameter of the classifier. For example, when the 1155 cm -1 portion selected as the characteristic peak of PS is enlarged, a certain range is specified as the input variable, not by using only one characteristic peak point as an input variable but by a characteristic peak point side. Raman spectroscopic measurement by the equipment because it can be minutely move (shift) in the present invention, by specifying a range on either side by 5㎝ -1 based on the characteristic peak point enter the Intensity values that finally corresponds to the 1150 ~ 1160㎝ -1 It is used as a variable. We have implemented a simulation to use 99 input values corresponding to 9 peaks as inputs to the classifier, using 11 input values per characteristic peak point.

12 to 14 are graphs showing experimental results of a classifier and a comparative classifier applied to a pretreatment-based intelligent classification method for sorting black plastic by Raman spectroscopy according to an embodiment of the present invention. FIGS. 12 to 14 show simulation and results of a pretreatment-based intelligent classification method for sorting black plastic by Raman spectroscopy according to an embodiment of the present invention. The data used in the present invention were obtained by collecting PET, PP, PS samples of black plastic from actual recycling centers and collecting and collecting a total of 300 samples of PET, PP, PS of black plastic, Through data analysis, three characteristic peak points were selected for each material. In addition, when the Raman spectroscopy equipment is used, the peak points are shifted finely, so that the range of the peak point is not set, and all the information in the range is used as the input parameter of the classifier. Also, in the present invention, the classification performance of the FCM-based RBFNN classifier is compared using the input parameters of the same condition for comparison with the SVM classifier. Here, the structure in the FCM-based RBFNN used for comparison is based on a general neural network structure. However, by using the FCM clustering method instead of the existing Gaussian function in the hidden layer portion, So that it can display the shape of radial active function. Also, the fuzzy coefficient is 2.0, and the rule number is fixed to 4.0 when the performance is the best. In the tables shown in FIGS. 12 and 13, the reliability of the classifier performance is enhanced by using 5-fold cross validation. Each test also uses the PCA algorithm and the values listed are the average of the results from 5-fcv, which is the result of four tests. In the table, Training_PCR and Testing_PCR indicate the pattern classification rate of the training data and the pattern classification rate of the test data, and the values indicated by ± indicate Standard Deviation. In the experiment of the simulation in the present invention, the performance of the SVM classifier and the RBFNN classifier were almost similar, but it was confirmed that the SVM classification ratio was higher. That is, the table of FIG. 14 shows the number of errors / classifications for each class in the case where the Training_PCR with the best performance in the table of FIG. 13 is 97.92% and the Testing_PCR is 95.67%. In the case of PET, nine out of 100 data are wrong, all of the PPs are matched, and six of the PS are misclassified. The pre-processing-based intelligent classification method for sorting black plastic by Raman spectroscopy according to an embodiment of the present invention extracts data from the collected samples using Raman spectroscopy equipment and analyzes the characteristic peaks through chemical theory and data analysis And the selected characteristic peaks were used as input variables of the SVM classifier to confirm the classification performance.

The present invention may be embodied in many other specific forms without departing from the spirit or essential characteristics and scope of the invention.

S110: Acquisition of spectral data for each black plastic material (PET, PP, PS)
S111: A process in which three characteristic peaks in which the peak point regions are not overlapped with each other in consideration of the chemical characteristics of the black plastic material (PET, PP, PS) are extracted and extracted
S112: Process of generating high-dimensional input variables considering the intensity values of the five wavelength ranges on both sides of the characteristic peak of the black plastic material
S120: Step of preprocessing spectral data for each black plastic material
S121: Step of constructing a set of learning data for recognition
S122: calculating the average of the data set from the set of learning data
S123: calculating the difference between the average of the learning data and the learning data
S124: Calculating the covariance matrix from the learning data for recognition
S125: calculating eigenvalue matrix and eigenvector matrix of the covariance matrix through eigenvalue analysis and selecting a specific number having the largest eigenvalue
S126: Extracting the reduced feature data by the linear transformation in each of the data and the transformation matrix having a specific number of eigenvectors
Step S130: Learning about the chemical properties of each black plastic material
S140: the black plastics are classified based on the chemical characteristics of each material
S141: Step of inputting any plastic
S142: the spectral data of the black plastic inputted through the Raman spectroscopic equipment is obtained
S143: a step in which spectral data of the obtained black plastic is subjected to a preprocessing process of dimension reduction
S144: the preprocessed data is input to the learned SVM pattern classifier and the arbitrarily inputted black plastic is classified into materials

Claims (7)

A pretreatment-based intelligent classification method for black plastic sorting by Raman spectroscopy,
(1) obtaining spectral data for each of black plastic materials (PET, PP, PS) using Raman spectroscopy equipment;
(2) preprocessing spectral data of the black plastic material obtained through the step (1) through a principal component analysis (PCA) algorithm;
(3) inputting the preprocessed data through the step (2) into a support vector machine (SVM) pattern classifier, and learning the chemical properties of the black plastic material by the SVM pattern classifier; And
(4) For the black plastic arbitrarily inputted, the input black plastics are classified based on the chemical characteristics of each material by using the SVM pattern classifier learned through the above-mentioned steps (1) to (3) Based classification method for black plastic by Raman spectroscopy.
2. The method according to claim 1, wherein in the step (1)
Three characteristic peaks in which the peak point regions are not overlapped in consideration of the chemical properties of the black plastic material (PET, PP, PS) are selected and extracted from spectral data obtained by the black plastic materials (PET, PP, PS) Further comprising the steps of: (a) pre-processing based intelligent classification for black plastic sorting by Raman spectroscopy.
3. The method according to claim 2, wherein in the step (1)
Dimensional input parameter considering the intensity values of the five wavelength regions on both sides based on the characteristic peaks of the black plastic material (PET, PP, PS) Pretreatment - based intelligent classification method for black plastic sorting by Raman spectroscopy.
4. The method according to any one of claims 1 to 3, wherein in the step (2)
Wherein the preprocessing process is performed by converting the selected high dimensional input variables into low dimensional input variables in consideration of the chemical characteristics of the black plastic material in the step (1) through the principal component analysis algorithm. Pretreatment - based intelligent classification method for black plastic sorting by spectroscopic method.
5. The method according to claim 4, wherein the step (2) is a preprocessing process using an algorithm of the principal component analysis method,
(2-1) constructing a set of learning data for recognition;
(2-2) calculating an average of the data set in the set of learning data;
(2-3) calculating a difference between an average of the learning data and the learning data;
(2-4) calculating a covariance matrix from learning data for recognition;
(2-5) calculating an eigenvalue matrix and an eigenvector matrix of a covariance matrix by analyzing eigenvalues and selecting a specific number having the largest eigenvalue; And
(2-6) extracting the reduced feature data by the linear transformation in each of the data and the transformation matrix having a specific number of eigenvectors, and performing a pre-processing based on the Raman spectroscopic method Intelligent classification method.
5. The method according to claim 4, wherein in the step (3)
Wherein the 5-fold cross validation for 5-fold cross validation is used to increase the reliability of the classifier performance of the SVM pattern classifier.
5. The method of claim 4, wherein step (4)
(4-1) inputting any plastic;
(4-2) obtaining spectral data of the black plastic inputted through the step (4-1) through the Raman spectroscopic equipment;
(4-3) a step of pre-processing the dimension reduction of spectral data of the black plastic obtained through the step (4-2); And
(4-4) In the step (4-3), the preprocessed data is input to the SVM pattern classifier learned through the steps (1) to (3), and the arbitrarily input black plastic is classified into materials Based intelligent classification method for black plastic sorting by Raman spectroscopy.
KR1020160148293A 2016-11-08 2016-11-08 A intelligent classification method for preprocessing based of a black plastic sorting by raman spectroscopy KR20180051239A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
KR1020160148293A KR20180051239A (en) 2016-11-08 2016-11-08 A intelligent classification method for preprocessing based of a black plastic sorting by raman spectroscopy

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1020160148293A KR20180051239A (en) 2016-11-08 2016-11-08 A intelligent classification method for preprocessing based of a black plastic sorting by raman spectroscopy

Publications (1)

Publication Number Publication Date
KR20180051239A true KR20180051239A (en) 2018-05-16

Family

ID=62452307

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1020160148293A KR20180051239A (en) 2016-11-08 2016-11-08 A intelligent classification method for preprocessing based of a black plastic sorting by raman spectroscopy

Country Status (1)

Country Link
KR (1) KR20180051239A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102172914B1 (en) 2019-06-07 2020-11-03 한국생산기술연구원 Fast searching method and apparatus for raman spectrum identification
WO2021107735A3 (en) * 2019-11-29 2021-07-22 정지훈 Method for classifying coffee raw materials, method for providing roasting profile of raw beans using same, and system for providing information on coffee raw materials
CN114674803A (en) * 2022-02-16 2022-06-28 北京化工大学 Quantitative analysis method for components of marine environment micro-plastic mixture
CN114850078A (en) * 2022-05-23 2022-08-05 安徽大学 Dark plastic material sorting unit system based on raman spectrum
CN114965973A (en) * 2022-05-12 2022-08-30 知里科技(广东)有限公司 Method for identifying recycled plastic based on instrument detection and analysis technology combined with multiple chemometrics methods and/or machine learning algorithm
WO2024010120A1 (en) * 2022-07-06 2024-01-11 주식회사 에이트테크 Waste classification system based on vision-hyperspectral fusion data
KR20240009180A (en) 2022-07-13 2024-01-22 잎스 주식회사 Recyclable plastic identifying apparatus

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102172914B1 (en) 2019-06-07 2020-11-03 한국생산기술연구원 Fast searching method and apparatus for raman spectrum identification
WO2021107735A3 (en) * 2019-11-29 2021-07-22 정지훈 Method for classifying coffee raw materials, method for providing roasting profile of raw beans using same, and system for providing information on coffee raw materials
CN114674803A (en) * 2022-02-16 2022-06-28 北京化工大学 Quantitative analysis method for components of marine environment micro-plastic mixture
CN114965973A (en) * 2022-05-12 2022-08-30 知里科技(广东)有限公司 Method for identifying recycled plastic based on instrument detection and analysis technology combined with multiple chemometrics methods and/or machine learning algorithm
CN114850078A (en) * 2022-05-23 2022-08-05 安徽大学 Dark plastic material sorting unit system based on raman spectrum
WO2024010120A1 (en) * 2022-07-06 2024-01-11 주식회사 에이트테크 Waste classification system based on vision-hyperspectral fusion data
KR20240009180A (en) 2022-07-13 2024-01-22 잎스 주식회사 Recyclable plastic identifying apparatus

Similar Documents

Publication Publication Date Title
KR20180051239A (en) A intelligent classification method for preprocessing based of a black plastic sorting by raman spectroscopy
Möller et al. Photometric classification of type Ia supernovae in the SuperNova Legacy Survey with supervised learning
Sheng et al. Classification of iron ores by laser-induced breakdown spectroscopy (LIBS) combined with random forest (RF)
EP1992939A1 (en) A kernel-based method and apparatus for classifying materials or chemicals and for quantifying the properties of materials or chemicals in mixtures using spectroscopic data.
CN107818298B (en) General Raman spectrum feature extraction method for machine learning substance identification algorithm
Samsudin et al. Spectral feature selection and classification of roofing materials using field spectroscopy data
CN102982339B (en) A kind of method that EO-1 hyperion characteristic variable is chosen
Roh et al. Development of intelligent sorting system realized with the aid of laser-induced breakdown spectroscopy and hybrid preprocessing algorithm-based radial basis function neural networks for recycling black plastic wastes
Lavine et al. Wavelets and genetic algorithms applied to search prefilters for spectral library matching in forensics
CN104376325A (en) Method for building near-infrared qualitative analysis model
Huang et al. Oil source recognition technology using concentration-synchronous-matrix-fluorescence spectroscopy combined with 2D wavelet packet and probabilistic neural network
CN112766227A (en) Hyperspectral remote sensing image classification method, device, equipment and storage medium
JP2002090299A (en) Method for distinguishing grade of high-molecular material
Tamin et al. A review of hyperspectral imaging-based plastic waste detection state-of-the-arts
KR101850087B1 (en) Plastic classifying method based on plastic material using laser spectrum method
Liu et al. Hyperspectral band selection based on consistency-measure of neighborhood rough set theory
Höppener et al. Classification of (micro) plastics using cathodoluminescence and machine learning
Cesetti et al. Waste processing: new near infrared technologies for material identification and selection
CN113408616A (en) Spectrum classification method based on PCA-UVE-ELM
Chen et al. Spectroscopic identification of environmental microplastics
Brunner et al. Evaluation of potential emission spectra for the reliable classification of fluorescently coded materials
Bonifazi et al. Quality control by HyperSpectral Imaging (HSI) in solid waste recycling: logics, algorithms and procedures
KR101674915B1 (en) Plastic classifying method based on plastic material using atr ft-ir spectroscopy and rbfnn pattern classifier
Vajna et al. Analysis of car shredder polymer waste with Raman mapping and chemometrics.
Lei et al. Near-infrared spectrum of coal origin identification based on SVM algorithm

Legal Events

Date Code Title Description
A201 Request for examination
E902 Notification of reason for refusal
E601 Decision to refuse application