METHOD FOR CLASSIFYING PLASTICS BY MATERIAL USING ATR FT-IR SPECTROSCOPY AND RBFNN PATTERN CLASSIFIER
The present invention refers to a classification search method relates to plastic, more specifically ATF fT a-iR (Attenuated Total Reflection Fourier transform provided infra Red) spectroscopy and RBFNN (Radial Basis Function Neural Network) pattern classification groups depending on the material classification using plastic to the method are disclosed. Recently, as petroleum resources reserves reduces the number of plastic raw material is fallen short of severe environmental pollution and waste plastic article transmissions by the disk is waiting vehicles are level, plastic recycling of the electric abruptly increases in research and development and interest are disclosed. Such plastic recycling plastic classification techniques in the field of plastic recycling system should be applied for implementing advanced valuable technology of pewter. However, chamber number on a conveyer belt and domestic use mainly in plastic classification system there are plastic cladding ends hinged are provided which classification by using a level. Automated plastic classification system for the near infrared ray spectrometry (near provided infrared Reflectance Spectroscopy, NIRS) according to classification using the waste plastic silicone based life but contribute to the pivot bearing, therefore have passivation absorbing black plastic, near infrared spectrometry by using only black plastic silicone according to respective classification were used. Such a plastic classification in conjunction, Public Patent Notification number 10 - 2013 - 0019818 call (title of the invention: visible-range plastic discriminating device and plastic classification system, date of publication: 27 February 2013 years) disclosure like corrosion disclosed. The present invention refers to basing said number of loose method such as door number points provided in order to solve loose number, ATR fT a-iR spectroscopy spectrum data are obtained by using plastic material, plastic material RBFNN pattern classification according to the acquired data input for each plastic material by learning with respect to each, black plastic material classification according to an input plastic can be more accurately and rapidly, and ATF fT-a iR spectroscopy using a classification number the method depending on the material plastic RBFNN pattern classification groups or the under public affairs intended. According to features of the present invention for achieving said purposes, ATF fT a-iR spectroscopy and RBFNN pattern classification groups the method depending on the material classification using a plastic, (1) ATR fT a-iR spectroscopy spectral data is obtained step by using plastic material; (2) said step (1) plastic material is pretreated by spectral data obtained through step; (3) said step (2) which has been pretreated to input data through RBFNN pattern classification, said RBFNN pattern classification with respect to learning step is by said plastic material; and (4) optionally plastic for input, said step (1) to step (3) learned through using RBFNN pattern classification groups, depending on the material classification step including said input electrode of plastic characterized. Preferably, said step (2) is, (2 - 1) said step (1) in each spectrum obtained through plastic material, each plastic material characterized by extracting peak point, said point corresponding to the extracted features peak frequency (wavenumber) extracting a value; and (2 - 2) said step (1) in each spectrum obtained through plastic material, each material in the presence of frequency area and characterized by peak point, said each frequency value extracted in the frequency range corresponding to the transmission value can be extracted. More preferably, said step (2 - 1) is, (2 - 1 - 1) said step (1) by plastic material based on spectrum obtained through all peak point, said plastic material by the steps of estimating approximate spectrum primary function; (2 - 1 - 2) said step (2 - 1 - 1) obtained by plastic material through said primary function transmission value by spectrum approximate estimation step (1) by the rear portion of the transmission spectrum obtained through the steps of plastic material; (2 - 1 - 3) said step (2 - 1 - 2) on first transmission value difference, extracting a peak point corresponding to a predetermined threshold; and (2 - 1 - 4) said step (2 - 1 - 3) extracted through said peak point can obtain the corresponding frequency value. Preferably, said step (3) in, Said RBFNN pattern classification used in the FCM (Fuzzy C provided an upper and a lower) based RBFNN and pattern classification group, differential evolution algorithm (Differential Evolution, hereinafter 'DE') can optimize a parameter of said RBFNN using pattern classification. Preferably, said step (4) is, (4 - 1) any plastic input step; (4 - 2) through said ATR fRANKLIN a-iR spectroscopy (4 - 1) plastics on the basis of spectral data is obtained via a step; (4 - 3) said step (4 - 2) is pretreated plastics on the basis of spectral data obtained via step; and (4 - 4) said step (4 - 3) for the pre-processed data through said step (1) to step (3) learned through RBFNN pattern classification input for classification according to color from the plastic material can be. In the present invention number fT a-iR spectroscopy and RBFNN pattern classification groups and ATF-in depending on the material classification according to a method using plastic, ATR fT a-iR spectroscopy spectrum data are obtained by using plastic material, plastic material RBFNN pattern classification according to the acquired data input for each plastic material by learning with respect to each, black plastic material classification according to an input plastic can more accurately and rapidly. Figure 1 shows a classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy and therefore RBFNN pattern classification method according to the flow of plastic by the use of silicone shown flow. Figure 2 shows a classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy and therefore RBFNN pattern classification method according in the plastic by the use of silicone, ATF fT a-iR spectroscopy shown by spectrum obtained using the plastic material. Figure 3 shows a classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy and therefore RBFNN pattern classification method according in the plastic by the use of silicone, ATF fT a-iR spectroscopy obtained using plastic spectral data for a through flow of the pre-processing process number 1 method shown. Figure 4 shows a classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy and therefore RBFNN pattern classification method according in the plastic by the use of silicone, ATF fT non-iR spectroscopy of PET plastic spectra data obtained using a 90 is shown. Figure 5 shows a classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy and therefore RBFNN pattern classification method according in the plastic by the use of silicone, a plastic PET ATF fT a-iR spectroscopy obtained using the spectral data pre-processing the decoded signal through the number 2 method shown. Figure 6 shows a classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy and therefore RBFNN pattern classification method according in the plastic by the use of silicone, includes a positive any plastic, plastic flow of the process according to the classification input material shown. In the present invention hereinafter with reference to the attached drawing in a preferred embodiment for the present invention is provided to a person with skill in the art to the embodiment hereinafter is detailed as follows. Only, so that the detail of the present invention preferred embodiment of the SFC, publicly known related function or configuration description is the subject matter of invention specifically breach can be decided to omit other description if the analogy. In addition, similar functionality and action throughout the part in which the same or similar code for the drawing less than 2000. In addition, the entire specification, 'connected' that the synthetic resin when any portion, this' connected directly 'as well as when, the other element interposed therebetween intermediate' indirectly connected ' comprises a unit when. In addition, frames are 'comprising' any configuration element, particularly opposite substrate under the outside number but without other components may further include other components of switched to each other. Figure 1 shows a classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy and therefore RBFNN pattern classification method according to the flow of plastic by the use of silicone shown flow are disclosed. As shown in fig. 1, the pattern classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy is a method using plastic RBFNN and depending on the material classification, (1) ATR fT a-iR spectroscopy spectral data is obtained step (S100) by using plastic material, (2) step (1) plastic material is pretreated by spectral data obtained through step (S200), (3) step (2) which has been pretreated to input data through RBFNN pattern classification, RBFNN pattern classification with respect to learning step (S300) is by plastic material, and (4) optionally plastic for input, step (1) to step (3) learned through using RBFNN pattern 999000093299 9 groups, depending on the material classification input plastic comprising step (S400) can be. In hereinafter, the classification groups according to one embodiment of the invention ATF fT non-iR spectroscopy using the method depending on the material plastic and RBFNN pattern classification step specifically describe each less than 1000. In step S100 ATR fT a-iR spectroscopy spectral data by using plastic material can be obtained. Wherein, the ATR fT a-iR spectroscopy, transparent layer and includes a light source fixed to that of the sample via a beam splitter (Beam Splitter) elapsed in the openings, each concave of optical path difference I made several wavelength detected interferometer (Interferometer) obtained in the detector (Detector), obtained via the interferometer (Fourier Transform) of obtaining a spectrum of a signal via a Fourier transform spectroscopy signal of the fT a-iR, two mirror reflecting infrared sample detected wavelength compared to original wavelength long rods method for clustering a preset legal entity ATR (attenuated total reflection) mixed therein. Figure 2 shows a classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy and therefore RBFNN pattern classification method according in the plastic by the use of silicone, ATF fT a-iR spectroscopy shown by spectrum obtained using the plastic material are disclosed. As shown in fig. 2, in step S100 ATF fT a-iR spectroscopy can be determine by using plastic material. In addition, by using ATF fT a-iR spectroscopy, near-infrared spectroscopy to obtain spectral data of the existing method of plastic material through miserable black can be discriminable acquisition by also for hereinafter. In step S200 step S100 through step S210 and S220 by plastic material obtained through th spectral data pre-processing can be disclosed. In step S100 ATR fT a-iR spectroscopy by spectral data is obtained using plastic material, causes each wavelength (transmittance) is measured at wavelength coverage area of transmission value because the Linear data array format, used as input data for same RBFNN pattern classification in order to pre-process the spirit. Specifically, through pretreatment step S210, obtained through modifying step S100 by plastic material, each plastic material characterized by extracting peak point, extracting and frequency value corresponding to the extracted features peak point, through pretreatment step S220, obtained through modifying step S100 by plastic material, each material in the presence of frequency peak area and each feature point, each frequency value extracted in the frequency range are contacted extracting value are disclosed. The frequency (wavelength) to provide the means by which the reciprocal of the wavelength (wavenumber), representative of the number of leaf and per 1 cm length, unit is cm-1 Mainly less than 2000. More particularly, through pretreatment step S210, step S100 by spectrum based on the peak point obtained through an all plastic material, plastic material (S211) the steps of estimating approximate spectrum by primary function, step S211 by spectrum obtained by plastic material through primary function transmission value obtained through an approximate estimation least S100 (S212) the rear portion of the transmission spectrum plastic material by the steps of, step S212 on first transmission value difference, corresponding peak point (S213) extracting a preset threshold, and step S213 peak point extracted through corresponding frequency can be made via obtain (S214). In hereinafter, with reference to 3 also through step S210, step S100 is pretreated to explain detailed process for each plastic spectral data acquired in less than 1000. Figure 3 shows a classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy and therefore RBFNN pattern classification method according in the plastic by the use of silicone, ATF fT a-iR spectroscopy obtained using plastic spectral data for a through flow of the pre-processing process number 1 method shown are disclosed. Number 1 method wherein method step S210 through pretreated joined substrate. In step S211, 3 (a) and 3 (b) also as shown in also, step S100 plastic spectrum obtained through all peak point is determined, based on the approximate estimation plastic spectrum peak point all primary function ( ) Can be determined. Wherein, the differentiated spectrum peak point all the minimum weighted point be a moving point. In addition, in step S212, step S100 y obtained through plastic spectrum function into a slot, step S211 plastic spectrum obtained through primary function approximate estimation Referred to as surface, plastic spectrum obtained through least 100 primary function transmission value approximate estimation spectrum transmission at the rear portion of plastic can be derived. In step S213, also as shown in 3 (c), step S212 preset threshold (threshold) or more obtained by transmission through the rear portion of the corresponding peak point can be extracted. In addition, step S212 on first transmission portion of the peak point when a preset threshold corresponding to a plurality personal, final peak point based on the magnitude of the value of the rear portion of the transmission can be extracted. Wherein, in the embodiment can be set differently according to the preset thresholds, S212 on first transmission portion of the corresponding peak point when a plurality personal preset threshold, the number of final peak point extracting other processes can be determined also in the embodiment. In the embodiment according, step S212 on first transmission portion of the corresponding peak point when a preset threshold personal 10, magnitude of the value of the rear portion of the transmission chamber is installed to enclose the final peak point extraction of upper 4 peak point and, in the embodiment according to another, magnitude of the value of the rear portion of the transmission chamber is installed to enclose the final peak point extraction of upper peak point 5 may be filled. In step S214, step S213 peak point can be extracted through corresponding frequency values. In hereinafter, with reference to the step S220 also 4 and also 5 through, acquired in step S100 is pretreated plastic spectral data to explain detailed process for each less than 1000. Figure 4 shows a classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy and therefore RBFNN pattern classification method according in the plastic by the use of silicone, ATF fT non-iR spectroscopy of PET plastic spectra data obtained using a 90 shown in the drawing and, according to one embodiment of the invention ATF fT a-iR spectroscopy and Figure 5 the plastic by the use of a classification method according in silicone RBFNN pattern classification groups, ATF fT a-iR spectroscopy obtained using the number 2 method for pretreating process through a PET plastic spectra data shown are disclosed. The method number 2 method step S220 through pretreated joined substrate. As shown in fig. 4, the stator is even of PET plastic called, of modifying the physical properties of a plastic PET 90 obtains the requested frequency value corresponding to each peak point although not all be the same, can be sure that the peak point appears mainly featured in a particular frequency range. The, in step S220, as shown in (a) also 5, PET plastic spectrum peak frequency value appearing in the face and is characterized in that a predominantly point, peak point can be mobile (shift) in view of the error range features, features can be extracted mainly point appearing frequency peak. For example, peak point features appearing in the frequency value predominantly 1713 cm-1 When, peak point features can be moved by considering an error range, 1713 cm-1 About ± 10cm-1 Range of frequency range (1703 cm-1 To 1723 cm-1) Can be extracted. In addition, as shown in (b) also 5, based on the extracted frequency range, each frequency value extracted in the frequency range corresponding to each transmissive value can be extracted. In step S300, step S200 RBFNN pattern which has been pretreated to input data through classification, classification is by plastic material can be RBFNN pattern with respect to learning. Wherein, step S200 is pre-treated to through classification data to be input for RBFNN pattern, which has been pretreated through step S210, step S100 in an acquired spectrum through corresponding frequency value and the extracted features peak point, which has been pretreated through step S220, step S100 through an acquired spectrum is characterized in that a peak point each frequency value is present in the frequency range value corresponding each transmissive, RBFNN pattern classification used in the FCM (Fussy C provided means) based RBFNN pattern classification group. In addition, in step S300, using differential evolution algorithm (Differential Evolution) FCM cluster ring purge programmable gate, and the second half RBFNN rule possibility RBFNN hidden layer (clustering) structure of a pattern of parameters including RBFNN classification (polynomial order) polynomial can be optimized. In step S400, optionally plastic input, step S100 to step S300 learned through RBFNN pattern classification group can be classification depending on the material of plastic. Figure 6 shows a classification groups according to one embodiment of the invention ATF fT a-iR spectroscopy and therefore RBFNN pattern classification method according in the plastic by the use of silicone, includes a positive any plastic, plastic flow of the process according to the classification input material shown are disclosed. As shown in fig. 6, any plastic is inputted (S411), fRANKLIN-a iR spectroscopy via ATR plastics on the basis of spectral data is obtained (S412), spectral data pre-processing (S413) is obtained plastic, data which has been pretreated to step S100 through step S300 learned RBFNN pattern classification (S414) plastic material can be input for each classification. As the device, in the present invention number fT a-iR spectroscopy and RBFNN pattern classification groups and ATF-in depending on the material classification according to a method using plastic, ATR fT a-iR spectroscopy spectrum data are obtained by using plastic material, plastic material RBFNN pattern classification according to the acquired data input for each plastic material by learning with respect to each, black plastic material classification according to an input plastic can more accurately and rapidly. Technology field of the invention the present invention refers to the device in a person with skill in the art and various deformation or by applications, technical idea defined by generated by the range of the present invention according to claim below will. S100: ATR fT a-iR spectroscopy spectral data is obtained step by using plastic material S200: step S100 plastic material is pretreated by spectral data obtained through step S210: obtained through modifying step S100 by plastic material, each plastic material characterized by extracting peak point, peak point frequency (wavenumber) extracting a value corresponding to the extracted features S211: step S211 by spectrum based on the peak point obtained through an all plastic material, plastic material by the steps of estimating approximate spectrum primary function S212: step S211 approximate estimation by spectrum obtained by plastic material through primary function transmission value obtained through the rear portion of the transmission spectrum least S100 plastic material by the steps of S213: step S212 on first transmission value difference, corresponding peak point extracting a preset threshold S214: step S213 to obtain a corresponding point extraction via at least one peak frequency S300: RBFNN pattern classification which has been pretreated to input data through step S200, RBFNN pattern classification is by plastic material with respect to learning step S400: plastic for optionally input, step S100 to step S300 using learned through RBFNN pattern classification groups, depending on the material classification step input plastic S410: any plastic input step S420: fRANKLIN-a iR spectroscopy ATR through step S410 plastics on the basis of spectral data is obtained via a step S430: plastics on the basis of spectral data obtained through step S420 is pretreated step S440: step S430 step S100 to step S300 learned through pre-processed data through input for classification according to step RBFNN pattern classification plastic material The present invention relates to a method for classifying plastics by material using an ATR FT-IR spectroscopy and an RBFNN pattern classifier. More particularly, the present invention relates to a method including: (1) a step in which spectrum data by plastic material is obtained using the ATR FT-IR spectroscopy; (2) a step in which the spectrum data by plastic material obtained through the step (1) is pre-processed; (3) a step in which the data pre-processed through the step (2) is inputted to the RBFNN pattern classifier, and the RBFNN pattern classifier learns characteristics by plastic material; and (4) a step in which a randomly inputted plastic is classified by the material, using the RBFNN pattern classifier learned through the steps (1)-(3), for the randomly inputted plastic. According to the method for classifying plastics by material using an ATR FT-IR spectroscopy and an RBFNN pattern classifier suggested in the present invention, the inputted plastics including a black plastic can be classified more accurately and rapidly by the material, by obtaining the spectrum data by plastic material using the ATR FT-IR spectroscopy and making the RBFNN classifier learn the characteristics by plastic material by inputting the data obtained by the plastic material. COPYRIGHT KIPO 2016 fT a-iR ATR (Attenuated Total Reflection Fourier transform provided infra Red) spectroscopy and RBFNN (Radial Basis Function Neural Network) pattern classification groups depending on the material classification using plastic as a method, (1) ATR fT a-iR spectroscopy spectral data is obtained step by using plastic material; (2) said step (1) plastic material is pretreated by spectral data obtained through step; (3) said step (2) which has been pretreated to input data through RBFNN pattern classification, said RBFNN pattern classification is by said plastic material with respect to learning step; and (4) optionally plastic for input, said step (1) to step (3) learned through using RBFNN pattern classification groups, said input comprising a wet protein depending on the material classification step, said step (2) is, (2 - 1) said step (1) in each spectrum obtained through plastic material, each plastic material characterized by extracting peak point, said point corresponding to the extracted features peak frequency (wavenumber) extracting a value; and (2 - 2) said step (1) in each spectrum obtained through plastic material, each material in the presence of frequency area and characterized by peak point, said each frequency value extracted in the frequency range including the corresponding transmission value extracting characterized, ATR fT a-iR spectroscopy and RBFNN pattern classification groups depending on the material classification method using a plastic. Back number According to Claim 1, said step (2 - 1) is, (2 - 1 - 1) said step (1) by plastic material based on spectrum obtained through all peak point, said plastic material by the steps of estimating approximate spectrum primary function; (2 - 1 - 2) said step (2 - 1 - 1) obtained by plastic material through said primary function transmission value by spectrum approximate estimation step (1) by the steps of plastic material obtained through the rear portion of the transmission spectrum; (2 - 1 - 3) said step (2 - 1 - 2) on first transmission value difference, extracting a predetermined threshold corresponding to peak point; and (2 - 1 - 4) said step (2 - 1 - 3) to obtain the corresponding frequency value extracted through said peak point characterized as including, ATR fT a-iR spectroscopy and RBFNN pattern classification groups depending on the material classification method using a plastic. Back number According to Claim 1, said step (4) includes, (4 - 1) any plastic input step; (4 - 2) fRANKLIN a-iR spectroscopy ATR through said step (4 - 1) plastics on the basis of spectral data is obtained via a step; (4 - 3) said step (4 - 2) the spectrum data of the obtained via step is pretreated plastic; and (4 - 4) said step (4 - 3) for the pre-processed data through said step (1) to step (3) learned through RBFNN pattern classification input to a step including a plastic material characterized according to classification, ATR fT a-iR spectroscopy and RBFNN pattern classification groups depending on the material classification method using a plastic.
