Black resin material determination method, black resin material determination device, waste plastic sorting system, and black resin material determination program

Abstract

To provide a technique capable of discriminating the kind of a black resin material by using near-infrared region spectrum data, and securing necessary and sufficient discrimination accuracy even in that case.SOLUTION: The method includes a step (S102) of classifying a material type of a treatment target into a first group to which a material type in which a feature amount appears belongs and a second group to which a material type other than the material type in which the feature amount appears belongs, and a step (S102, S103) of handling the material type belonging to the second group as one group and determining whether the material type of the treatment target belongs to the second group or corresponds to any material type belonging to the first group.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a black resin material discrimination method, a black resin material discrimination device, a waste plastic piece sorting system, and a black resin material discrimination program. [Background technology]

[0002] In recent years, with the growing awareness of recycling, the reuse of waste plastics has been attracting attention. However, some types of plastics are not suitable for reuse, for example, they emit harmful gases when incinerated. Therefore, when reusing waste plastics, it is necessary to identify the material type of the waste plastic to be reused.

[0003] Some waste plastics that are eligible for recycling are made of black resin materials. For waste plastics made of black resin materials, for example, the absorption characteristics of carbon black used as a colorant result in low resin-specific reflection characteristics. Therefore, it has been proposed to identify the material type using a reflectance spectrum obtained with a mid-infrared sensor (see, for example, Patent Document 1). Furthermore, because highly accurate identification is essential when recycling waste plastics as an alternative fuel to coal, it has also been proposed to identify the material type of black resin waste plastics using near-infrared data, mid-infrared data, and Raman scattering (see, for example, Patent Document 2). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2021-128062 [Patent Document 2] Patent Publication No. 2023-069528 Summary of the Invention [Problem to be solved by the invention]

[0005] From the viewpoint of promoting the widespread reuse of waste plastics, it is desirable to be able to accurately distinguish the type of material of black resin materials using a low-cost system configuration. However, as disclosed in Patent Document 1, for example, when using a reflectance spectrum obtained by a mid-infrared sensor, there are concerns that the use of dedicated equipment will increase costs because mid-infrared sensors are more expensive and less widely used than near-infrared sensors.Furthermore, there is a risk that it will be difficult to set up the environment for acquiring mid-infrared spectral data. Furthermore, if near-infrared spectral data is used, it may be possible to perform highly accurate determination by also using mid-infrared data and Raman scattering, as disclosed in Patent Document 2, for example. However, in this case, there is a risk that the system configuration will become more complex and larger, resulting in higher costs.

[0006] The present invention provides a technology that enables the material type of black resin material to be determined using near-infrared spectral data, while still ensuring sufficient accuracy in the determination. [Means for solving the problem]

[0007] A first aspect of the present invention is A black resin material discrimination method for discriminating the material type of a processing object made of a black resin material using near-infrared spectrum data, comprising: a step of classifying a plurality of material types assumed to be the black resin material into a first group to which material types in which the feature amounts appear in the near-infrared spectrum data belong, and a second group to which other material types belong, based on the feature amounts appearing in the near-infrared spectrum data; a step of treating material types belonging to the second group as a single group when determining the material type of the processing object using the near-infrared spectrum data obtained from the processing object, and determining whether the material type of the processing object belongs to the second group or any of the material types belonging to the first group; The black resin material discrimination method includes the steps of:

[0008] A second aspect of the present invention is a step of performing a predetermined preprocessing on the near-infrared spectral data obtained from the processing object determined to belong to the second group; determining whether the material type of the processing object corresponds to any of the material types belonging to the second group using the near-infrared spectrum data after the preprocessing; The black resin material discrimination method according to the first aspect includes:

[0009] A third aspect of the present invention is The preprocessing includes scaling and noise removal for the near-infrared spectral data. 10 is a black resin material discrimination method according to a second aspect.

[0010] A fourth aspect of the present invention is The first group of materials includes at least one of polyvinyl chloride, polymethyl methacrylate, rubber, and polyurethane foam. The black resin material discrimination method according to any one of the first to third aspects.

[0011] A fifth aspect of the present invention is The second group of materials includes at least one of polyethylene, polypropylene, polycarbonate, polystyrene, and acrylonitrile butadiene styrene. The black resin material discrimination method according to any one of the first to third aspects.

[0012] A sixth aspect of the present invention is a data acquisition unit that acquires near-infrared spectrum data from a processing object made of a black resin material; a material type determination unit that determines the material type of the processing object using the near-infrared spectrum data, The material type determination unit A plurality of material types assumed to be the black resin material are classified into a first group to which material types in which the feature amounts appear in the near-infrared spectrum data belong, and a second group to which other material types belong, based on the feature amounts appearing in the near-infrared spectrum data, and When determining the material type of the processing object, the material types belonging to the second group are treated as a single group, and it is determined whether the material type of the processing object belongs to the second group or any of the material types belonging to the first group. This is a black resin material discrimination device.

[0013] A seventh aspect of the present invention is The black resin material discrimination device according to the sixth aspect, A conveying device that conveys waste plastic as the object to be treated; a sorting device that sorts the waste plastics transported by the transport device according to the material type discrimination result by the black resin material discrimination device; This is a waste plastic sorting system equipped with the above.

[0014] An eighth aspect of the present invention is On the computer, acquiring near-infrared spectrum data from a processing object made of a black resin material; a step of classifying a plurality of material types assumed to be the black resin material into a first group to which material types in which the feature amounts appear in the near-infrared spectrum data belong, and a second group to which other material types belong, based on the feature amounts appearing in the near-infrared spectrum data; a step of treating material types belonging to the second group as a single group when determining the material type of the processing object using the near-infrared spectrum data obtained from the processing object, and determining whether the material type of the processing object belongs to the second group or any of the material types belonging to the first group; The black resin material discrimination program executes the above. [Effects of the Invention]

[0015] According to the present invention, it is possible to determine the material type of a black resin material using near-infrared spectral data, while still ensuring the necessary and sufficient accuracy of determination. [Brief explanation of the drawings]

[0016] [Figure 1] 1 is an explanatory diagram showing a configuration example of a waste plastic sorting system according to an embodiment of the present invention. [Figure 2] FIG. 1 is a flowchart showing an example of a procedure of a black resin material discrimination method according to an embodiment of the present invention. [Figure 3] 1A and 1B are explanatory diagrams showing a specific example of a scaling process performed in a black resin material discrimination method according to an embodiment of the present invention, in which FIG. 1A shows spectral data before the scaling process, and FIG. 1B shows spectral data after the scaling process. [Figure 4] 1A and 1B are explanatory diagrams showing a specific example of a noise removal process performed in a black resin material discrimination method according to an embodiment of the present invention, in which FIG. 1A shows spectral data before the noise removal process, and FIG. 1B shows spectral data after the noise removal process. [Figure 5] 1A and 1B are explanatory diagrams showing examples of evaluation of the discrimination results obtained by the black resin material discrimination method according to one embodiment of the present invention, in which (a) is a diagram showing an evaluation example of a comparative example, (b) is a diagram showing an evaluation example by group classification, and (c) is a diagram showing an evaluation example after preprocessing. DETAILED DESCRIPTION OF THE INVENTION

[0017] Hereinafter, a black resin material discrimination method, a black resin material discrimination device, a waste plastic sorting system, and a black resin material discrimination program according to one embodiment of the present invention will be described with reference to the drawings.

[0018] <System configuration example> First, a configuration example of a waste plastic sorting system according to this embodiment will be described. FIG. 1 is an explanatory diagram showing an example of the configuration of a waste plastic sorting system according to this embodiment.

[0019] The waste plastic sorting system (hereinafter also simply referred to as "system") 1 exemplified in this embodiment is used when recycling waste plastics, and sorts the waste plastics 2 to be recycled according to their material type. In particular, in the system 1 of this embodiment, the waste plastics 2 made of black resin material are the objects to be processed for material type sorting. Details of the black resin material will be provided later.

[0020] In order to sort the material types of such processing objects, the system 1 of this embodiment is configured to include at least a conveying device 10, a material identifying device 20, and a sorting device 30. Furthermore, the system 1 of this embodiment may also include a pre-processing device and a cleaning device (not shown) as needed.

[0021] The pretreatment device performs processes such as crushing and classifying the waste plastic 2 to be recycled to produce waste plastic pieces of a predetermined size. Since the crushing and classifying processes can be performed using known techniques, detailed explanations will be omitted here. The cleaning device cleans the waste plastic 2 to be reused, removing dirt and foreign matter adhering to the surface.

[0022] The conveying device 10 conveys waste plastic 2 to be recycled (waste plastic pieces of a predetermined size if a pre-treatment device is provided, or waste plastic after cleaning if a cleaning device is provided; the same applies in the following explanation). The waste plastic 2 can be conveyed, for example, by using a belt conveyor, but this is not necessarily limited to this, and other known technologies can also be used.

[0023] The material determination device 20 determines the type of material (quality) that constitutes each of the waste plastics 2 transported by the transport device 10. More specifically, the material determination device 20 determines the type of black resin material that constitutes the waste plastics 2, which are the objects to be processed, transported by the transport device 10. The material determination device 20 determines the type of material of the objects to be processed using near-infrared spectral data, and the specific details of the material type determination will be described later in detail.

[0024] The sorting device 30 sorts the waste plastics 2 transported by the conveying device 10 according to the results of the material type determination by the material determination device 20. The sorting of the waste plastics 2 can be performed, for example, using a plurality of storage containers 31 prepared for each material type and an air injection nozzle 32 that changes the trajectory of the waste plastics 2 discharged from the conveying device 10 to determine which storage container 31 the waste plastics 2 are to be stored in, but this is not necessarily limited to this, and other known techniques may also be used.

[0025] With the above-described configuration, in the system 1 of this embodiment, the material determination device 20 determines the material type of the waste plastic 2 to be processed while it is being transported by the transport device 10. Then, the sorting device 30 sorts the waste plastic 2 by material type according to the material type determination result by the material determination device 20. Therefore, according to the system 1 of this embodiment, even if various types of waste plastic 2 are mixed in the supplied waste plastic 2, by going through sorting according to the material type determination result, it is possible to exclude waste plastic 2 of a material type that is not suitable for reuse from being subject to reuse.

[0026] <Configuration example of material discrimination device> Next, a description will be given of an example of the configuration of the material determining device 20 in the above-described system 1. The material determining device 20 corresponds to a specific example of the "black resin material determining device" according to the present invention.

[0027] As shown in FIG. 1, the material determining device 20 is configured to include at least an irradiation source 21, a detector 22, and a determiner .

[0028] The irradiation source 21 is configured, for example, by a halogen lamp, and irradiates light onto the waste plastic 2 (i.e., the object to be processed by the material determination device 20) being transported by the transport device 10. Note that the irradiation source 21 is not necessarily limited to a halogen lamp, and may be configured using other known technologies, as long as it irradiates light onto the object to be processed.

[0029] The detector 22 is configured, for example, by a hyperspectral camera, and can acquire spectral data of multiple (specifically, 100 or more) wavelength bands of light by dispersing light into wavelength bands (hereinafter, wavelength bands are also referred to as "bands") and capturing images. More specifically, the detector 22 is positioned to capture images of the waste plastic 2 as the processing object, and acquires spectral data, particularly in the near-infrared wavelength band, as the imaging results (hyperspectral image data) of the processing object using reflected light from the processing object. In other words, the detector 22 corresponds to a specific example of a "data acquisition unit" according to the present invention, which acquires near-infrared spectral data from the processing object. The near-infrared wavelength band referred to here refers to, for example, a wavelength band of 780 nm or more and 2.5 μm or less. Note that the detector 22 may acquire spectral data of other wavelength bands in addition to near-infrared spectral data, as long as it acquires near-infrared spectral data.

[0030] The classifier 23 is configured by a computer device that combines hardware resources such as a CPU (Central Processing Unit), RAM (Random Access Memory), ROM (Read Only Memory), etc. Then, by executing a predetermined program (software) that has been pre-installed, information processing by the software is specifically realized using the hardware resources. As a result, the classifier 23 realizes a function for determining the material type of the waste plastic 2 as the processing object, using the near-infrared spectral data acquired by the detector 22. In other words, the classifier 23 corresponds to a specific example of a "material type determination unit" according to the present invention, which determines the material type of the processing object using near-infrared spectral data.

[0031] <Examples of material types to be processed> Here, a specific example will be described regarding the material type of the waste plastic 2 that is the processing target of the material determination device 20. As described above, the material determination device 20 of this embodiment determines the material type of the waste plastic 2 that is made of black resin material.

[0032] A black resin material refers to a resin material that has a black appearance or an appearance that can be considered black. The black appearance color may be the color of the resin material itself, or it may be a colored resin. The following types of black resin materials are envisioned. Specifically, the following nine types of black resin materials may be considered as potential materials for waste plastic 2: polyvinyl chloride (PVC), polymethyl methacrylate (PMMA), rubber, polyurethane foam (PUR), polyethylene (PE), polypropylene (PP), polycarbonate (PC), polystyrene (PS), and acrylonitrile butadiene styrene (ABS). While the black resin materials are not necessarily limited to the nine types listed here, the following description will use these nine types as examples of black resin materials.

[0033] Among these types of black resin materials, there are some that have low resin-specific reflectance characteristics due to, for example, the absorption characteristics of carbon black used as a colorant. Therefore, even if near-infrared spectral data is acquired, the characteristic features are not apparent in the near-infrared spectral data, unlike spectral data for non-black resins. This is one of the reasons why it is difficult to identify the material type of black resin materials using near-infrared spectral data. In light of this, conventional methods have used mid-infrared spectral data instead of near-infrared spectral data (see, for example, Patent Document 1), or mid-infrared data and Raman scattering in addition to near-infrared spectral data (see, for example, Patent Document 2).

[0034] The present inventors conducted extensive research into this issue and came up with the following idea. As described above, black resin materials have small reflectance characteristics in near-infrared spectral data, and therefore some black resin materials have features that are difficult to reveal in near-infrared spectral data. Therefore, it is difficult to uniformly determine which of the nine types of black resin materials constituting the processing target object corresponds to using near-infrared spectral data. However, since feature values ​​are not hidden for all nine types, the nine material types can be classified into two groups: one containing material types that do not reveal feature values ​​and the other containing material types that can reveal feature values. Based on this, the inventors came up with the novel idea that by grouping the nine material types and then performing a step-by-step discrimination process on each group, it is possible to discriminate the material type of black resin materials with sufficient discrimination accuracy even when using near-infrared spectral data.

[0035] To realize the above-mentioned idea, the material determining device 20 of this embodiment is configured to determine the material type of black resin material in accordance with the processing procedure described below.

[0036] The processing procedure described below is realized by the discriminator 23, which serves as a computer device in the material discrimination device 20, executing a predetermined program (software). In other words, the predetermined program that realizes the processing procedure corresponds to one embodiment of the "black resin material discrimination program" according to the present invention.

[0037] In this case, the specified program that realizes the processing procedure may be provided by being stored on a recording medium (e.g., a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc.) that can be read by the discriminator 23, as long as it can be installed in the discriminator 23, or may be provided from outside via a network such as the Internet or a dedicated line.

[0038] <Processing procedure for determining the material type of black resin material> Next, an example of the processing operation of the material discrimination device 20 of this embodiment will be described. Here, we will mainly specifically explain the procedure for discriminating the material type of the black resin material that constitutes the waste plastic 2. The procedure given as an example here corresponds to one specific example of the "black resin material discrimination method" according to the present invention.

[0039] FIG. 2 is a flowchart showing an example of the procedure of the black resin material discrimination method according to this embodiment.

[0040] In this embodiment, when determining the material type of the black resin material that makes up the waste plastic 2, the detector 22 first acquires near-infrared spectral data from the waste plastic 2 (step 101, hereinafter step will be abbreviated as "S").

[0041] Once the near-infrared spectral data is acquired, the classifier 23 can extract features from the near-infrared spectral data. The features can be extracted by appropriately combining processes using known techniques, such as normalization of the spectral data, feature selection using a machine learning model (Random Forests), and learning using a multilayer perceptron (MLP). However, the use of these techniques is not necessarily limited, and other known techniques may also be used to extract features.

[0042] As described above, in this embodiment, nine types of black resin materials are assumed: PVC, PMMA, rubber, PUR, PE, PP, PC, PS, and ABS. Among these material types, some have obvious feature amounts (i.e., the difference in feature amounts from other materials is clear), while others have small reflection and absorption characteristics for light in the near-infrared range, making it difficult to identify feature amounts. Materials with difficult feature amounts are difficult to distinguish from other materials, which may lead to misidentification of material types (reduced identification accuracy).

[0043] Therefore, in this embodiment, instead of using a uniform criterion to determine which of the nine material types the extracted feature quantities of near-infrared spectrum data belong to, the following group classification is performed and a different discrimination process is applied to each group.

[0044] Specifically, when determining the type of material of the black resin material, the material is classified into a first group to which material types that exhibit characteristic quantities appear in the near-infrared spectrum data belong, and a second group to which other material types belong (S102). Classification into the first and second groups may be performed, for example, as follows.

[0045] First, near-infrared spectral data is acquired from a large number of sample pieces prepared in advance for each of the nine material types, and the feature quantities of the near-infrared spectral data are categorized by material type. Then, for each categorized feature quantity, the feature quantity is classified as either one that is obvious (i.e., one that has a clear difference in feature quantity from other materials) or one that is difficult to obvious. As a result, material types that have obvious feature quantities belong to group 1, and other material types belong to group 2. Such grouping can be performed, for example, using a known machine learning method.

[0046] When grouped in this manner, in this embodiment, the first group of materials includes four types, for example, PVC, PMMA, rubber, and PUR, while the second group of materials includes five types, for example, PE, PP, PC, PS, and ABS.

[0047] It is assumed that information regarding the grouping (e.g., information regarding the material types belonging to each group, information regarding the categorization of the features of the near-infrared spectral data by material type, etc.) is stored in advance in a database or the like that can be accessed by the discriminator 23.

[0048] By accessing the information on such grouping, the discriminator 23 can determine which type pattern the feature of the acquired near-infrared spectral data matches, and then classify the data as belonging to the first group or the second group (S102).

[0049] If the result of the above group classification is that the material belongs to the first group, the material type of the black resin material is determined based on the feature values ​​that appear in the acquired near-infrared spectrum data, since the first group contains material types for which the feature values ​​are apparent (S103). Specifically, the material type of the waste plastic 2 to be processed is determined to be one of the material types belonging to the first group: PVC, PMMA, rubber, or PUR. The method for determining the material type at this time may be, for example, using the learning function of a one-dimensional convolutional neural network, but is not necessarily limited to this, and other known techniques may also be used.

[0050] The discrimination results for each material type in the first group obtained in this way clearly show the features that appear in the near-infrared spectral data, so even when the near-infrared spectral data is used, there is no misidentification of the material type and the discrimination accuracy is sufficient.

[0051] On the other hand, if the material is classified as belonging to the second group as a result of the group classification, the material types belonging to the second group are treated as a single group, and a different discrimination process is applied to each material type in the first group, since the second group contains material types whose features are difficult to reveal.

[0052] Specifically, for waste plastics 2 of a material type determined to belong to the second group, predetermined pre-processing is performed on near-infrared spectrum data obtained from the material type of the black resin material that constitutes the waste plastics 2 (S104). The predetermined pre-processing includes, for example, scaling and noise removal processing for the near-infrared spectrum data to be processed. As long as the pre-processing includes at least these scaling and noise removal processing, it is acceptable for the pre-processing to include other processing.

[0053] The scaling process and noise removal process as preprocessing will be specifically described below.

[0054] The second group includes material types whose features are difficult to reveal. For this reason, near-infrared spectral data classified as belonging to the second group is subjected to scaling processing, for example, for bands with small reflectance characteristics, with the aim of emphasizing the reflectance characteristics. Scaling processing is a process of aligning the scale of values ​​that the spectral data can take between each piece of data, and is performed, for example, using the following equation (1):

[0055]

number

[0056] In the above equation (1), x is the signal intensity value at each wavelength in the spectrum data, min(x) is the minimum value of x, max(x) is the maximum value of x, and x' is the value of x after scaling processing.

[0057] FIG. 3 is an explanatory diagram showing a specific example of the scaling process performed in this embodiment. In the near-infrared spectral data of the material types belonging to the second group, the features are difficult to reveal, as shown in Fig. 3(a) for example. When scaling processing using the above formula (1) is performed on such near-infrared spectral data, for example, for the band between 1000.2 nm and 1663.8 nm, the scaling processing results in near-infrared spectral data in which the magnitude of the signal intensity value for each wavelength is revealed by emphasizing the reflection characteristics, as shown in Fig. 3(b) for example.

[0058] When the scaling process described above is performed, the feature values ​​of the near-infrared spectral data after the scaling process are made apparent, but there is a risk that the noise components may be amplified. Therefore, after the scaling process is performed, a noise reduction process is performed on the near-infrared spectral data after the scaling process in order to remove the noise components.

[0059] The noise removal process involves, for example, performing a fast Fourier transform on the scaling-processed spectrum data to convert it into a frequency signal, and then performing a process of reducing the cutoff frequency of the converted frequency signal using a low-pass filter. The cutoff frequency may be set as appropriate. The data obtained by performing a fast inverse Fourier transform on the frequency signal after reducing the cutoff frequency is used as the noise removal-processed spectrum data.

[0060] FIG. 4 is an explanatory diagram showing a specific example of the noise removal process performed in this embodiment. The near-infrared spectral data after scaling has superimposed noise components, as shown in Figure 4(a). If such near-infrared spectral data is subjected to noise reduction processing, for example, according to the procedure described above, the superimposed noise components are removed by the noise reduction processing, as shown in Figure 4(b), and as a result, near-infrared spectral data with prominent feature values ​​is obtained.

[0061] After the preprocessing including at least the scaling and noise removal processes described above is performed, the material type of the black resin material classified as belonging to the second group is determined based on the feature quantities revealed by the preprocessing for the near-infrared spectrum data after the preprocessing, as shown in Fig. 2 (S105). Specifically, the material type of the waste plastic 2 to be processed is determined to be one of the material types belonging to the second group, namely PE, PP, PC, PS, or ABS. The method for determining the material type at this time is the same as that for the first group.

[0062] The discrimination results for each material type in the second group obtained in this way have sufficient discrimination accuracy without misidentifying the material type, even when near-infrared spectral data is used, because the feature values ​​are made apparent by preprocessing.

[0063] Thereafter, it is determined whether or not the material type has been determined for all waste plastics 2 to be processed (S106), and the above-described series of processes (S101 to S106) is repeated for each waste plastic 2 until all have been determined.

[0064] <Effects of this embodiment> According to the black resin material discrimination method, black resin material discrimination device, waste plastic sorting system, and black resin material discrimination program described in this embodiment, the following effects can be obtained.

[0065] In this embodiment, the material types of black resin materials are classified into a first group to which material types for which feature quantities appear in near-infrared spectral data belong, and a second group to which other material types belong. When determining the material type of the processing object, material types belonging to the second group are treated as a single group, and it is determined whether the material type of the processing object belongs to the second group or to any of the material types belonging to the first group. By treating material types belonging to the second group as a single group in this way, even if the feature quantities of the material types belonging to the second group are difficult to manifest in near-infrared spectral data and it is difficult to distinguish them from other materials, there is no need to distinguish them from other materials when they are treated as a single group, so there is no risk of misidentification of material types (deterioration in discrimination accuracy). That is, in this embodiment, instead of using a uniform standard to determine which of nine assumed material types of black resin materials the material falls under based on the feature amounts of near-infrared spectrum data, the material is classified into first and second groups, and then a different discrimination process is applied to each group. Therefore, according to this embodiment, the nine material types are grouped, and then a discrimination process is performed stepwise for each group, so that misdiscrimination of material types (decrease in discrimination accuracy) can be suppressed even when near-infrared spectrum data is used.

[0066] In this embodiment, for material types determined to belong to the second group, predetermined preprocessing is performed on the near-infrared spectral data, and the near-infrared spectral data after the preprocessing is used to determine whether the material type of the processing target corresponds to any of the material types belonging to the second group. By performing such preprocessing, it becomes possible to realize the visualization of feature amounts for the near-infrared spectral data of material types belonging to the second group. In other words, in this embodiment, by applying a different discrimination process to the material types belonging to the second group than to the first group, it is possible to discriminate the material types based on the feature quantities of the near-infrared spectral data without mis-discriminating the material types (deteriorating discrimination accuracy).

[0067] Moreover, in this embodiment, as pre-processing performed on the near-infrared spectral data of the material type determined to belong to the second group, at least scaling processing and noise removal processing are performed on the near-infrared spectral data. Therefore, in this embodiment, by undergoing at least scaling processing and noise removal processing, it is possible to reliably make the feature quantities appearing in the near-infrared spectral data even for material types belonging to the second group.

[0068] As described above, in this embodiment, the material types of black resin materials are classified into a first group and a second group based on the feature quantities appearing in the near-infrared spectrum data, and then a different discrimination process is applied to each group. In other words, by undergoing group classification and discrimination process for each group, it is possible to discriminate the material types of black resin materials using the near-infrared spectrum data. Therefore, according to this embodiment, compared to when mid-infrared spectral data is used, an increase in equipment costs can be suppressed, and the environmental setting for data acquisition does not become difficult. Furthermore, since it is not necessary to use mid-infrared spectral data, Raman scattering, etc., in addition to near-infrared spectral data, there is no risk of high costs due to the complexity and expansion of the system configuration. Moreover, even when near-infrared spectral data is used for discrimination, as described above, it is possible to ensure sufficient discrimination accuracy without causing erroneous discrimination of material types (deterioration of discrimination accuracy). In other words, according to this embodiment, it is possible to determine the material type of black resin material using near-infrared spectral data, while still ensuring the necessary and sufficient accuracy of determination.

[0069] The effects obtained in this embodiment will be quantitatively described below with specific examples. FIG. 5 is an explanatory diagram showing an example of evaluation of the discrimination result obtained by the black resin material discrimination method according to this embodiment. In Figure 5, the evaluation of the material type discrimination results for black resin materials is performed using the F-measure, which is the harmonic mean of precision and recall.

[0070] The precision is the proportion of data that is actually positive among data predicted to be positive, and can be determined, for example, by the following equation (2).

[0071]

number

[0072] The recall rate is the ratio of predicted positive results to actual positive results, and can be determined, for example, by the following equation (3).

[0073]

number

[0074] In the above formulas (2) and (3), TP is a true positive, FP is a false positive, and FN is a false negative.

[0075] The F-measure, which is the harmonic mean of precision and recall, can be determined, for example, by the following equation (4).

[0076]

number

[0077] When the F value is used to evaluate the results of discriminating the material type of black resin material, the following results are obtained.

[0078] For example, Fig. 5(a) shows the F-values ​​of nine types of black resin materials as an evaluation example of a comparative example, when the material types were determined based on uniform criteria using near-infrared spectral data without undergoing the grouping described in this embodiment. According to the example of Fig. 5(a), the four types of PVC, PMMA, rubber, and PUR are material types for which the features of the near-infrared spectral data are evident, and therefore all have F-values ​​of 0.9 or higher, indicating good discrimination accuracy. However, the five types of PE, PP, PC, PS, and ABS have F-values ​​of approximately 0.7 to 0.8, which may result in misidentification of the material types.

[0079] On the other hand, Fig. 5(b) shows the F-value when material types belonging to the second group are treated as a single group after group classification into the first and second groups, as an evaluation example according to this embodiment. According to the example of Fig. 5(b), the F-values ​​for all four types, PVC, PMMA, rubber, and PUR, are 0.9 or higher, indicating that good discrimination accuracy is achieved. Furthermore, the material types belonging to the second group (labeled "Group" in the figure) that are treated as a single group do not need to be distinguished from other materials when they are treated as a single group, and therefore the F-values ​​are 0.9 or higher, indicating that good discrimination accuracy is achieved.

[0080] 5(c) shows, as an evaluation example according to this embodiment, the F-values ​​obtained when the material types of the second group were identified after preprocessing of the near-infrared spectral data. According to the example in FIG. 5(c), the F-values ​​for all five types of PE, PP, PC, PS, and ABS were all 0.8 or higher after preprocessing, demonstrating better identification accuracy than the example in FIG. 5(a).

[0081] As described above, according to this embodiment, even in the quantitative evaluation using the F value, it is possible to ensure sufficient discrimination accuracy in discriminating the material type of black resin material when near-infrared spectral data is used.

[0082] <Modification> Although the embodiments of the present invention have been described above, the above disclosure shows exemplary embodiments of the present invention, and the technical scope of the present invention is not limited to the above exemplary embodiments.

[0083] For example, in the above-described embodiment, nine types of black resin materials are assumed: PVC, PMMA, rubber, PUR, PE, PP, PC, PS, and ABS, but the black resin materials are not necessarily limited to these nine types.

[0084] The same applies to the group classification of each material type; the specific form of group classification described in the above embodiment is merely an example, and even if the group classification is performed in a form different from that of the above embodiment, it is possible to obtain the effects described in the above embodiment by applying different discrimination processes to each group after the group classification. [Explanation of symbols]

[0085] 1...waste plastic sorting system, 2...waste plastic, 10...conveyor device, 20...material discrimination device, 21...irradiation source, 22...detector, 23...discriminator, 30...sorting device, 31...storage container, 32...air injection nozzle

Claims

1. A black resin material discrimination method for discriminating the material type of a processing object made of a black resin material using near-infrared spectrum data, comprising: classifying a plurality of material types assumed to be the black resin material into a first group including material types for which the feature amounts appear in the near-infrared spectrum data are apparent, and a second group including other material types; a step of treating material types belonging to the second group as a single group when determining the material type of the processing object using the near-infrared spectrum data obtained from the processing object, and determining whether the material type of the processing object belongs to the second group or any of the material types belonging to the first group; A method for identifying a black resin material, comprising:

2. a step of performing a predetermined preprocessing on the near-infrared spectral data obtained from the processing object determined to belong to the second group; determining whether the material type of the processing object corresponds to any of the material types belonging to the second group using the near-infrared spectrum data after the preprocessing; The method for identifying a black resin material according to claim 1 , further comprising:

3. The preprocessing includes scaling and noise removal for the near-infrared spectral data. The method for identifying a black resin material according to claim 2 .

4. The first group of materials includes at least one of polyvinyl chloride, polymethyl methacrylate, rubber, and polyurethane foam. The method for determining a black resin material according to claim 1 .

5. The second group of materials includes at least one of polyethylene, polypropylene, polycarbonate, polystyrene, and acrylonitrile butadiene styrene. The method for determining a black resin material according to claim 1 .

6. a data acquisition unit that acquires near-infrared spectrum data from a processing object made of a black resin material; a material type determination unit that determines the material type of the processing object using the near-infrared spectrum data, The material type determination unit A plurality of material types assumed to be the black resin material are classified into a first group to which material types in which the feature amounts appear in the near-infrared spectrum data belong, and a second group to which other material types belong, based on the feature amounts appearing in the near-infrared spectrum data, and When determining the material type of the processing object, the material types belonging to the second group are treated as a single group, and it is determined whether the material type of the processing object belongs to the second group or any of the material types belonging to the first group. Black resin material discrimination device.

7. The black resin material discriminating device according to claim 6 ; A conveying device that conveys waste plastic as the object to be treated; a sorting device that sorts the waste plastics transported by the transport device according to the material type discrimination result by the black resin material discrimination device; A waste plastic sorting system equipped with:

8. On the computer, acquiring near-infrared spectrum data from a processing object made of a black resin material; classifying a plurality of material types assumed to be the black resin material into a first group including material types for which the feature amounts appear in the near-infrared spectrum data are apparent, and a second group including other material types; a step of treating material types belonging to the second group as a single group when determining the material type of the processing object using the near-infrared spectrum data obtained from the processing object, and determining whether the material type of the processing object belongs to the second group or any of the material types belonging to the first group; A black resin material discrimination program that executes the above.

Images