A method, device, and storage medium for identifying solid waste of recycled plastic particles

Abstract

The application relates to the field of detection analysis, and discloses a regenerated plastic particle solid waste identification method, device, equipment and storage medium, which are used for quickly and accurately identifying whether a to-be-detected regenerated plastic particle sample belongs to solid waste. The method comprises the following steps: acquiring spectral information data set of the regenerated plastic particles; performing standard normal transformation and second-order derivative joint preprocessing on the spectral data in the training set, so as to optimize the spectral data in the training set; selecting a predetermined pattern recognition algorithm to perform model training on the training set, determining a target condition reaching the highest identification and classification accuracy under a preset condition, and establishing a regenerated plastic particle solid waste identification model; acquiring a near-infrared spectrum of the to-be-detected regenerated plastic particle sample, and inputting the near-infrared spectrum into the regenerated plastic particle solid waste identification model for identification, so as to determine whether the regenerated plastic particle sample belongs to solid waste.

Description

Technical Field

[0001] This invention relates to the field of detection and analysis, and in particular to a method, apparatus, equipment, and storage medium for identifying recycled plastic particle solid waste. Background Technology

[0002] Today, plastic products have become necessities in our daily lives and industrial production. To facilitate production, transportation, and use, the raw materials for producing plastic products are generally in granular form. Plastic raw materials are classified into virgin materials and recycled materials based on their source. Virgin materials refer to high-molecular polymers formed through polymerization of byproducts from fossil fuels such as petroleum. Recycled materials refer to plastic raw materials obtained by recycling used or waste plastics through processes such as sorting, crushing, and granulation; they are also called recycled plastic granules. Compared to producing virgin materials, recycling waste plastics into recycled materials achieves both energy recycling and reduced environmental pollution.

[0003] To compensate for the shortage of domestic plastic raw materials, my country imports a large amount of recycled plastic pellets every year. Some companies import solid waste generated during the production process, such as waste materials, machine head and tail materials, or substandard products, as recycled plastic pellets. Although my country has completely banned the import of solid waste, the increasing volume of recycled plastic pellet imports increases the probability of solid waste entering the country. Therefore, strictly controlling the inflow of solid waste is one of the key tasks of customs at present.

[0004] Currently, the identification of recycled plastic pellet solid waste usually requires using the appearance, infrared spectroscopy, ash content, differential scanning and other indicators of the sample as detection indicators, and then judging whether it belongs to solid waste according to the relevant clauses in GB 34330-2017 "General Rules for Identification of Solid Waste". The whole identification process is difficult and time-consuming.

[0005] With the development of foreign trade and my country's deepening participation in global competition, the public's call for trade facilitation is growing, placing new demands on expedited customs clearance. To fully leverage the role of customs technical departments in promoting customs efficiency through science and technology, provide technical support to frontline customs inspectors, and find suitable methods for rapid identification of imported recycled plastic pellets as solid waste, thus providing a theoretical basis for strengthening port enforcement, is of significant practical importance.

[0006] Based on the aforementioned inability to quickly and accurately identify whether recycled plastic particles belong to solid waste, this invention provides a method for identifying recycled plastic particles as solid waste, which can quickly determine whether recycled plastic particles belong to solid waste. Summary of the Invention

[0007] This invention provides a method, apparatus, device, and storage medium for identifying recycled plastic pellet solid waste. Based on the relationship between near-infrared spectral absorbance information and material information, it fully utilizes the infrared absorption spectral characteristics of recycled plastic pellets to establish identification models for multiple categories of recycled plastic pellet solid waste. This solid waste identification model enables rapid and accurate differentiation of whether a sample to be tested belongs to solid waste, distinguishing between recycled plastic pellets and solid waste plastic pellets, thereby providing a quick detection and judgment conclusion and improving the efficiency and accuracy of solid waste identification for samples to be tested.

[0008] To achieve the above objectives, the first aspect of the present invention provides a method for identifying recycled plastic pellet solid waste, comprising:

[0009] Obtain a spectral information dataset of recycled plastic particles, the spectral information dataset including a training set and a test set;

[0010] The spectral data in the training set are preprocessed using the standard normal transformation algorithm, and the spectral data obtained after the standard normal transformation are subjected to second-order derivative processing based on window optimization selection to optimize the spectral data in the training set.

[0011] A predetermined pattern recognition algorithm is selected to train the model on the near-infrared spectrum in the training set, and the target conditions for achieving the highest recognition and classification accuracy are determined under preset conditions. Under the target conditions, a solid waste identification model for recycled plastic particles is established.

[0012] The near-infrared spectrum of the recycled plastic particle sample to be tested is obtained and input into the recycled plastic particle solid waste identification model for identification, thereby determining whether the recycled plastic particle sample belongs to solid waste.

[0013] Optionally, in another embodiment of the method for identifying recycled plastic particles as solid waste, before acquiring the spectral information dataset of the recycled plastic particles, the method further includes:

[0014] Near-infrared spectra of a predetermined number of recycled plastic granule samples were collected using a predetermined diffuse reflectance method to obtain the near-infrared spectra of the recycled plastic granules.

[0015] Sample categories with fewer than a preset value of spectra in the near-infrared spectrum are removed to obtain a spectral information dataset of recycled plastic pellet samples. The spectral information dataset of recycled plastic pellet samples includes each sample category, the near-infrared spectrum corresponding to each sample category, and the spectral data contained in the near-infrared spectrum, as well as whether the sample belongs to solid waste.

[0016] Using a systematic sampling method with a set interval, near-infrared spectra are selected from the spectral information dataset for each sample category according to a first predetermined ratio as training set and test set, respectively, so that the spectral information dataset is divided into a training set for training the model and a test set for testing the recognition effect of the model according to a second predetermined ratio.

[0017] Optionally, in another embodiment of the method for identifying recycled plastic particle solid waste, after removing sample categories with fewer than a preset value of spectral counts in the near-infrared spectrum to obtain a dataset of spectral information for recycled plastic particle samples, the method further includes:

[0018] Near-infrared spectra containing saturated absorption spectra are removed from the spectral information dataset of the recycled plastic particle samples.

[0019] Optionally, in another embodiment of the method for identifying recycled plastic pellet solid waste, the preprocessing of the spectral data in the training set using a standard normal transformation algorithm, and the second-order derivative processing of the spectral data obtained after the standard normal transformation based on window optimization, specifically includes:

[0020] Each spectrum in the spectral data is subjected to a standard normal transformation according to a preset formula to obtain the spectral data after standard normal transformation. The preset formula is: in m is the number of spectral spot lengths, and k = 1, 2, ..., m;

[0021] The spectral data obtained after standard normal transformation is fitted with polynomial least squares on the spectral points within the window. The data of equidistant points within the window are fitted into a second-order polynomial according to the preset window size, and the spectral data after second-order differentiation is obtained.

[0022] Optionally, in another embodiment of the method for identifying recycled plastic particle solid waste, the step of using a predetermined pattern recognition algorithm to train a model on the near-infrared spectrum in the training set, and determining the target conditions to achieve the highest identification and classification accuracy under preset conditions, specifically includes establishing a recycled plastic particle solid waste identification model under the target conditions:

[0023] The nearest neighbor algorithm was used to train the model on the near-infrared spectrum in the training set, and different K values ​​were selected using ten-fold cross-validation to verify the classification accuracy of the model.

[0024] The K value corresponding to the highest accuracy is selected as the final K value, and a first identification model for recycled plastic pellet solid waste is established based on this K value.

[0025] Based on the test set used to test the recognition effect of the model, the classification accuracy of the first recycled plastic particle solid waste recognition model is verified to obtain the verified recycled plastic particle solid waste recognition model.

[0026] Optionally, in another embodiment of the method for identifying recycled plastic particle solid waste, the process of training a model on the near-infrared spectrum in the training set using a predetermined pattern recognition algorithm and determining the target conditions for achieving the highest identification and classification accuracy under preset conditions, specifically further includes:

[0027] The partial least squares classification algorithm was used to train the model on the near-infrared spectrum in the training set, and the maximum number of latent variables was set to 10. The ten-fold cross-validation method was used to select different numbers of latent variables to verify the classification accuracy of the model.

[0028] The optimal number of latent variables is selected based on the number of latent variables corresponding to the highest accuracy, and a second identification model for recycled plastic pellet solid waste is established based on this number of latent variables.

[0029] Based on the test set used to test the recognition effect of the model, the classification accuracy of the second recycled plastic particle solid waste recognition model is verified to obtain the verified recycled plastic particle solid waste recognition model.

[0030] Optionally, in another embodiment of the method for identifying recycled plastic particle solid waste, the step of using a predetermined pattern recognition algorithm to train a model on the near-infrared spectrum in the training set, and determining the target conditions to achieve the highest identification and classification accuracy under preset conditions, specifically further includes establishing a recycled plastic particle solid waste identification model under the target conditions:

[0031] The near-infrared spectrum in the training set is trained using the partial least squares support vector machine algorithm. The multiple categories in the training set are divided into several binary categories. Eight pairs of combinations are formed by using four pairwise methods (error correction output coding, minimum output coding, one-to-one and one-to-many) and two optimization methods (grid search and simplex search).

[0032] The classification accuracy of the eight combinations was compared and verified. The combination with the highest accuracy was selected as the optimal combination. A third solid waste identification model for recycled plastic particles was established based on the optimal combination.

[0033] Based on the test set used to test the recognition effect of the model, the classification accuracy of the third recycled plastic particle solid waste recognition model is verified to obtain the verified recycled plastic particle solid waste recognition model.

[0034] A second aspect of the present invention provides a device for identifying recycled plastic pellet solid waste, comprising:

[0035] The spectral information dataset acquisition module is used to acquire the spectral information dataset of recycled plastic particles, which includes a training set and a test set.

[0036] The spectral preprocessing module is used to preprocess the spectral data in the training set using the standard normal transformation algorithm, and to perform second-order derivative processing on the spectral data obtained after standard normal transformation based on window optimization selection, so as to optimize the spectral data in the training set.

[0037] The solid waste identification model building module is used to select a predetermined pattern recognition algorithm to train the model on the near-infrared spectrum in the training set, and determine the target conditions to achieve the highest identification and classification accuracy under preset conditions, and build a solid waste identification model for recycled plastic particles under the target conditions.

[0038] The identification and determination module is used to acquire the near-infrared spectrum of the recycled plastic particle sample to be tested, and input the near-infrared spectrum into the recycled plastic particle solid waste identification model for identification, and determine whether the recycled plastic particle sample belongs to solid waste.

[0039] Optionally, in another embodiment of the recycled plastic pellet solid waste identification device, the device further includes:

[0040] The spectral acquisition module is used to acquire the near-infrared spectra of a preset number of recycled plastic granule samples using a predetermined diffuse backscan method, thereby obtaining the near-infrared spectra of the recycled plastic granules.

[0041] The sample rejection module is used to reject sample categories with fewer than a preset value of spectral count in the near-infrared spectrum, thereby obtaining a spectral information dataset of recycled plastic particle samples. The spectral information dataset of recycled plastic particle samples includes each sample category, the near-infrared spectrum corresponding to each sample category, and the spectral data contained in the near-infrared spectrum, as well as whether the sample belongs to solid waste.

[0042] The dataset partitioning module is used to select near-infrared spectra for each sample category in the spectral information dataset according to a first predetermined ratio, using a systematic sampling method with a set interval distance, so as to divide the spectral information dataset into a training set for training the model and a test set for testing the recognition effect of the model according to a second predetermined ratio.

[0043] Optionally, in another embodiment of the recycled plastic pellet solid waste identification device, the device further includes:

[0044] The saturated absorption spectrum removal module is used to remove near-infrared spectra containing saturated absorption spectra from the spectral information dataset of the recycled plastic particle sample.

[0045] Optionally, in another embodiment of the recycled plastic particle solid waste identification device, the spectral preprocessing module specifically includes:

[0046] The standard normal variable transformation processing unit is used to perform a standard normal variable transformation on each spectrum in the spectral data according to a preset formula, to obtain the spectral data after standard normal transformation processing. The preset formula is: in m is the number of spectral spot lengths, and k = 1, 2, ..., m;

[0047] The second-order derivative processing unit is used to perform polynomial least squares fitting on the spectral points within the window after the standard normal transformation of the spectral data. The data of the equidistant points within the window are fitted into a second-order polynomial according to the preset window size, and the spectral data after second-order derivative processing is obtained.

[0048] Optionally, in another embodiment of the recycled plastic pellet solid waste identification device, the solid waste identification model building module specifically includes:

[0049] The nearest neighbor algorithm training unit is used to select the nearest neighbor algorithm to train the model on the near-infrared spectrum in the training set, and to select different K values ​​using ten-fold cross-validation to verify the classification accuracy of the model.

[0050] The K-value determination and model building unit is used to select the K-value corresponding to the highest accuracy as the final K-value, and to build a first identification model for recycled plastic pellet solid waste based on the K-value.

[0051] The first model verification unit is used to verify the classification accuracy of the first recycled plastic particle solid waste identification model based on the test set used to test the model recognition effect, so as to obtain the verified recycled plastic particle solid waste identification model.

[0052] Optionally, in another embodiment of the recycled plastic pellet solid waste identification device, the solid waste identification model building module specifically includes:

[0053] The partial least squares classification and discrimination training unit is used to train the model on the near-infrared spectrum in the training set using the partial least squares classification and discrimination algorithm, and sets the maximum number of latent variables to 10. Different numbers of latent variables are selected using ten-fold cross-validation to verify the classification accuracy of the model.

[0054] The optimal number of latent variables determination and model building unit is used to select the number of latent variables corresponding to the highest accuracy as the optimal number of latent variables, and to build a second recycled plastic particle solid waste identification model based on the number of latent variables.

[0055] The second model verification unit is used to verify the classification accuracy of the second recycled plastic particle solid waste identification model based on the test set used to test the model recognition effect, so as to obtain the verified recycled plastic particle solid waste identification model.

[0056] Optionally, in another embodiment of the recycled plastic pellet solid waste identification device, the solid waste identification model building module specifically includes:

[0057] The Partial Least Squares Support Vector Machine (PLS Support Vector Machine) training unit is used to train the near-infrared spectrum in the training set using the PLS Support Vector Machine algorithm. It divides the multiple categories in the training set into several binary classifications and forms eight pairs of combinations by using four pairwise methods (error correction output coding, minimum output coding, one-to-one and one-to-many) and two optimization methods (grid search and simplex search).

[0058] The optimal combination determination and model building unit is used to compare and verify the classification accuracy of the eight combinations, select the combination with the highest accuracy as the optimal combination, and build a third recycling plastic particle solid waste identification model based on the optimal combination.

[0059] The third model verification unit is used to verify the classification accuracy of the third recycled plastic particle solid waste identification model based on the test set used to test the model recognition effect, so as to obtain the verified recycled plastic particle solid waste identification model.

[0060] A third aspect of the present invention also provides a device for identifying recycled plastic pellet solid waste, wherein the device includes a memory and at least one processor, the memory stores instructions, and the memory and the at least one processor are interconnected via a circuit; the at least one processor invokes the instructions in the memory to cause the device for identifying recycled plastic pellet solid waste to perform any of the above-described methods for identifying recycled plastic pellet solid waste.

[0061] A fourth aspect of the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method for identifying recycled plastic particle solid waste as described in any of the preceding claims.

[0062] The technical solution provided by this invention involves acquiring a dataset of spectral information of recycled plastic particles; performing standard normal transformation and second derivative joint preprocessing on the spectral data in the training set to optimize the spectral data; selecting a predetermined pattern recognition algorithm to train the model on the training set, and determining the target conditions to achieve the highest recognition and classification accuracy under preset conditions to establish a solid waste identification model for recycled plastic particles; acquiring the near-infrared spectrum of the recycled plastic particle sample to be tested, and inputting the near-infrared spectrum into the solid waste identification model for identification to determine whether the recycled plastic particle sample belongs to solid waste. This invention enables rapid and accurate identification of whether a sample to be tested belongs to solid waste. Specifically, it utilizes the near-infrared absorption spectral characteristics of recycled plastic particles to establish a solid waste identification model for multiple categories of recycled plastic particles. This solid waste identification model can distinguish whether the sample to be tested is recycled plastic particles or solid waste plastic particles, thereby quickly and accurately obtaining the detection result. This improves the efficiency and accuracy of solid waste identification of samples to be tested, helps improve the efficiency of customs clearance for imported goods, provides technical support for strengthening customs enforcement, and has significant practical implications. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 This is a schematic diagram of one embodiment of the method for identifying recycled plastic granules as solid waste according to the present invention;

[0065] Figure 2 This is a schematic diagram of one embodiment of the solid waste identification device for recycled plastic particles in this invention;

[0066] Figure 3 This is a schematic diagram of one embodiment of the solid waste identification device for recycled plastic particles in this invention. Detailed Implementation

[0067] This invention provides a method, apparatus, device, and storage medium for identifying recycled plastic particles as solid waste, enabling rapid and accurate identification of the type of recycled plastic particles in a sample to be tested. This allows for quick determination of whether the sample is solid waste plastic particles, and significantly improves the efficiency and accuracy of sample testing.

[0068] To enable those skilled in the art to better understand the present invention, the embodiments of the present invention will be described below with reference to the accompanying drawings.

[0069] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0070] This application provides a method for identifying solid waste from recycled plastic pellets, which is described in detail below. (See attached document.) Figure 1 One embodiment of the method for identifying recycled plastic pellet solid waste in this invention includes:

[0071] Step 101: Obtain the spectral information dataset of recycled plastic particles, wherein the spectral information dataset includes a training set and a test set;

[0072] Step 102: Preprocess the spectral data in the training set using the standard normal transformation algorithm, and then perform second-order derivative processing based on window optimization selection on the spectral data obtained after the standard normal transformation to optimize the spectral data in the training set.

[0073] Step 103: Select a predetermined pattern recognition algorithm to train the model on the near-infrared spectrum in the training set, and determine the target conditions to achieve the highest recognition and classification accuracy under preset conditions, and establish a solid waste identification model for recycled plastic particles under the target conditions.

[0074] Step 104: Obtain the near-infrared spectrum of the recycled plastic particle sample to be tested, and input the near-infrared spectrum into the recycled plastic particle solid waste identification model for identification to determine whether the recycled plastic particle sample belongs to solid waste.

[0075] In existing technologies, different types of plastic particles have different chemical compositions, resulting in different near-infrared spectra. By utilizing pattern recognition methods from chemometrics, combined with optimization techniques such as preprocessing, a model capable of identifying as many samples as possible can be established. This invention aims to establish a solid waste identification model for recycled plastic particles for import and export customs inspection, enabling rapid and accurate batch detection of the type of recycled plastic particles in the sample. If the model cannot identify the type, it can be determined as solid waste plastic particles.

[0076] Furthermore, in another embodiment of the method for identifying recycled plastic particle solid waste, before acquiring the spectral information dataset of the recycled plastic particles, the method further includes:

[0077] Near-infrared spectra of a predetermined number of recycled plastic granule samples were collected using a pre-defined diffuse back-scan method to obtain near-infrared spectra of the recycled plastic granules.

[0078] Sample categories with fewer than a preset value of spectra in the near-infrared spectrum are removed to obtain a spectral information dataset of recycled plastic granule samples. The spectral information dataset of recycled plastic granule samples includes each sample category, the near-infrared spectrum corresponding to each sample category, and the spectral data contained in the near-infrared spectrum, as well as whether the sample belongs to solid waste.

[0079] Using a systematic sampling method with a set interval, near-infrared spectra are selected from the spectral information dataset for each sample category according to a first predetermined ratio as training set and test set, respectively, so that the spectral information dataset is divided into a training set for training the model and a test set for testing the recognition effect of the model according to a second predetermined ratio.

[0080] Specifically, it is necessary to collect near-infrared spectra of the plastic particles. Optionally, this invention uses a predetermined diffuse back-scan method to collect the near-infrared spectra of a predetermined number of recycled plastic particle samples, thereby obtaining the near-infrared spectrum of the recycled plastic particles. In specific implementation, a portable near-infrared spectrometer is used to collect the spectra of the plastic particles. This spectrometer used to collect the near-infrared absorption of the sample is preferably a portable near-infrared spectrometer including a tungsten lamp light source, a spectroscopic system, a detector, and a control and readout system. This invention can select the MicroNIR Pro ES1700 model near-infrared spectrometer. In specific implementation, a sample cup for loading the plastic particle sample to be tested can be self-made as a measurement accessory. This sample cup can be a cylindrical cup with a quartz glass bottom and a stainless steel rim. The sample cup is placed on the spectral acquisition unit of the near-infrared spectrometer. Light emitted from the tungsten lamp shines on the sample cup and interacts with the sample through the quartz glass base of the sample cup. The unabsorbed light returns to the near-infrared spectrometer through diffuse reflection. The diffuse reflected light is then split by a linear gradient filter. The light of different wavelengths after splitting enters the array for detection, thereby detecting the light intensity of different wavelengths. The light intensity signal can be transmitted to a computer or other terminal via a USB interface.

[0081] To establish the accuracy of recycled plastic particle sample collection, this invention employs diffuse reflectance scanning with a data acquisition interval of 7 nm, a scanning range of 950 nm-1650 nm, and a calibration white background. Each plastic particle sample is scanned by loading it into a sample cup, then reloading and repeating the acquisition process three times.

[0082] In one specific implementation, a total of 842 recycled plastic pellet samples were collected, mainly including ABS, PC-ABS, HDPE, HIPS, PA6, PA66, PBT, PC, PET, PETG, PMMA, POM, PP, PPA, and LCP, with a total of 2526 near-infrared spectra collected. Due to the large differences in sample numbers between different categories, and the fact that fewer samples were not statistically significant, sample categories with fewer than 40 spectra were ignored. A total of 15 categories of plastics were included, with the fewest having 41 spectra and the largest having 385; the total number of spectra used to establish the model was 2114. A total of 202 solid waste samples were collected, with a total of 608 near-infrared spectra collected. During the implementation of this project, recycled plastic pellets were classified into one category, and recycled plastic solid waste was classified into another. A summary information table is shown in Table 1.

[0083] Table 1. Summary Table of Recycled Plastic Pellet Sample Information

[0084]

[0085]

[0086] Furthermore, after obtaining the spectral information dataset of all samples, in order to train and test the solid waste identification model of recycled plastic particles, the present invention proposes to adopt a systematic sampling method with a set interval distance. In the spectral information dataset, for each sample category, near-infrared spectra are selected as training set and test set respectively according to a first predetermined ratio, so as to divide the spectral information dataset into a training set for training the model and a test set for testing the recognition effect of the model according to a second predetermined ratio.

[0087] Taking the spectral information dataset of the aforementioned 15 sample categories as an example, it can be divided into training and test sets in ratios of 8:2, 7:3, and 6:4. The training set is used to train the model, while the test set is not used for modeling but rather for testing the model's performance to avoid overfitting. This invention preferably uses a systematic sampling method (distance of three), selecting training and test sets for each category, with a ratio of approximately 3:1 between the training and test sets.

[0088] Because dark-colored plastics are very dark, they are highly likely to cause saturation absorption in the near-infrared spectrum. This results in very little near-infrared spectral information and makes the near-infrared spectra of dark-colored plastic samples very similar. Therefore, in this embodiment of the invention, models are established using all samples (including dark-colored plastics) and spectra with all saturated absorption removed, and the impact of removal on the accuracy of the models is compared. Therefore, in another embodiment of the method for identifying recycled plastic particle solid waste, after removing sample categories with fewer than a preset value of spectra in the near-infrared spectrum to obtain the spectral information dataset of recycled plastic particle samples, the method further includes:

[0089] Near-infrared spectra containing saturated absorption spectra are removed from the spectral information dataset of the recycled plastic particle samples.

[0090] Therefore, after removing near-infrared spectra containing saturated absorption spectra, the sample information summary table is shown in Table 2.

[0091] Table 2. Summary Table of Recycled Plastic Pellet Sample Information

[0092]

[0093] Furthermore, to improve data mining and analysis, reduce time and costs, preprocessing methods are frequently employed to enhance spectral quality and model performance. Specifically, to minimize the impact of uneven plastic particle size and non-specific scattering from particle surfaces, a standard normal transformation preprocessing method was first chosen. Moreover, since near-infrared spectral acquisition is susceptible to baseline drift and background interference, derivative spectroscopy was also employed in this invention to effectively eliminate baseline drift and background interference, providing higher resolution and clearer spectral profile changes than the original spectra.

[0094] In practical implementation, preprocessing the spectral data before modeling can improve the accuracy of modeling and the model recognition effect. The combined preprocessing method used in this invention yields relatively ideal results. Therefore, this invention combines the standard normal transformation algorithm with the second derivative preprocessing method. Thus, in another embodiment of the method for identifying recycled plastic granules as solid waste, the preprocessing of the spectral data in the training set using the standard normal transformation algorithm, and the subsequent second-order derivative processing of the spectral data obtained after the standard normal transformation, specifically includes:

[0095] Each spectrum in the spectral data is subjected to a standard normal transformation according to a preset formula to obtain the spectral data after standard normal transformation. The preset formula is: in m is the number of spectral points, k = 1, 2, ..., m; in practice, the standard normal transformation preprocessing algorithm built into The Unscrambler X can be used for calculation.

[0096] The spectral data obtained after standard normal transformation is fitted with a polynomial least squares method to the spectral points within a window. The data of equidistant points within the window are fitted into a second-order polynomial according to a preset window size, resulting in spectral data after second-order derivative processing. Specifically, the Savitzky-Golay (SG) window-shifting polynomial least squares fitting method is used, which can be achieved using the second-order derivative (SG) preprocessing method built into The Unscrambler X. The preferred window size in this invention is 7.

[0097] Furthermore, in terms of modeling methods, this invention selects three pattern recognition algorithms, including the nearest neighbor algorithm (KNN), the partial least squares classification and discrimination algorithm (PLS-DA), and the least squares support vector machine algorithm (LS-SVM).

[0098] Therefore, optionally, in another embodiment of the method of the present invention, the step of selecting a predetermined pattern recognition algorithm to train the model on the near-infrared spectra in the training set, and determining the target conditions to achieve the highest recognition and classification accuracy under preset conditions, specifically includes establishing a solid waste identification model for recycled plastic particles under the target conditions:

[0099] The nearest neighbor algorithm was used to train the model on the near-infrared spectra in the training set, and different K values ​​were selected using ten-fold cross-validation to verify the classification accuracy of the model.

[0100] The K value corresponding to the highest accuracy is selected as the final K value, and a first identification model for recycled plastic pellet solid waste is established based on this K value.

[0101] Based on the test set used to test the recognition effect of the model, the classification accuracy of the first recycled plastic particle solid waste recognition model is verified to obtain the verified recycled plastic particle solid waste recognition model.

[0102] The specific implementation uses Matlab 2019a software. The nearest neighbor algorithm KNN is implemented using the built-in ClassificationLearner module in Matlab 2019a. With the help of 10-fold cross-validation, different K values ​​will correspond to different classification accuracies of the model. The K value corresponding to the highest accuracy is selected as the final K value. After determining the final K value, the KNN model is built.

[0103] Optionally, in another embodiment of the method for identifying recycled plastic particle solid waste, the process of training a model on the near-infrared spectra in the training set using a predetermined pattern recognition algorithm and determining the target conditions for achieving the highest identification and classification accuracy under preset conditions, specifically further includes:

[0104] The partial least squares classification algorithm was used to train the model on the near-infrared spectra in the training set. The maximum number of latent variables was set to 10. Ten-fold cross-validation was used to select different numbers of latent variables to verify the classification accuracy of the model.

[0105] The optimal number of latent variables is selected based on the number of latent variables corresponding to the highest accuracy, and a second identification model for recycled plastic pellet solid waste is established based on this number of latent variables.

[0106] Based on the test set used to test the recognition effect of the model, the classification accuracy of the second recycled plastic particle solid waste recognition model is verified to obtain the verified recycled plastic particle solid waste recognition model.

[0107] In practice, the Partial Least Squares (PLS-DA) classification algorithm is implemented using the `classification_toolbox_4.0-1` package. The most crucial parameter in the PLS-DA algorithm is determining the optimal number of latent variables. To avoid overfitting, the number of latent variables should not be too large; in this invention, the maximum number of latent variables is set to 10. The optimal number of latent variables is determined using 10-fold cross-validation, i.e., the number of latent variables corresponding to the highest accuracy is selected as the optimal number of latent variables for the model. After determining the optimal number of latent variables, the PLS-DA model is established.

[0108] Optionally, in another embodiment of the method for identifying recycled plastic particle solid waste, the step of using a predetermined pattern recognition algorithm to train a model on the near-infrared spectra in the training set, and determining the target conditions to achieve the highest identification and classification accuracy under preset conditions, specifically further includes establishing a recycled plastic particle solid waste identification model under the target conditions:

[0109] The partial least squares support vector machine algorithm is used to train the model on the near-infrared spectra in the training set. The multiple categories in the training set are divided into several binary classifications. Eight pairs of combinations are formed by using four pairwise methods (error correction output coding, minimum output coding, one-to-one and one-to-many) and two optimization methods (grid search and simplex search).

[0110] The classification accuracy of the eight combinations was compared and verified. The combination with the highest accuracy was selected as the optimal combination. A third solid waste identification model for recycled plastic particles was established based on the optimal combination.

[0111] Based on the test set used to test the recognition effect of the model, the classification accuracy of the third recycled plastic particle solid waste recognition model is verified to obtain the verified recycled plastic particle solid waste recognition model.

[0112] In specific implementation, the Least Squares Support Vector Machine (LS-SVM) algorithm utilizes the LSSVMlabv1_8_R2009b_R2011a algorithm package. The LS-SVM algorithm employs a pairwise strategy, dividing multiple categories into several binary classifications and achieving classification of multiple categories through pairwise reclassification. In this invention, four pairwise methods are selected: Error Correcting Output Coding (ECOC), Minimum Output Coding (MOC), one-to-one coding, and one-to-all coding. Two optimization methods are also used: grid search and simplex search, resulting in eight combinations. The classification accuracy of the model under these eight combinations is compared, and the optimal combination is determined based on the highest accuracy, thus establishing the corresponding LS-SVM classification model.

[0113] After establishing the identification model for recycled plastic pellet solid waste using the above three modeling methods, it is necessary to test the identification effect of the model through the above test set, and verify the classification accuracy of the first, second and third recycled plastic pellet solid waste identification models respectively, so as to obtain the verified recycled plastic pellet solid waste identification model.

[0114] Specifically, taking Table 2, which summarizes the above 842 recycled plastic particle samples and 202 recycled plastic particle solid waste samples, as an example, the accuracy of the data is compared with that obtained by using the technical solution of the present invention, or by using the original spectral modeling, or by using the standard normal transformation / second derivative to preprocess the spectral data alone, and by using a combination of preprocessing methods to process the spectral data. The classification results including saturated absorption spectra are shown in Table 3.

[0115] Table 3. Summary of Classification Results Including Saturated Absorption Plastic Spectra

[0116]

[0117] After removing all saturated absorption spectra, the models established using the same joint preprocessing method and three pattern recognition algorithms were compared to obtain the classification results of the removed saturated absorption spectra, as shown in Table 4.

[0118] Table 4. Summary of Classification Results for Removing Saturated Absorption Plastic Spectra

[0119]

[0120] To fully utilize the spectral characteristics of recycled plastic particles, over 800 recycled plastic particle samples were collected in the aforementioned specific embodiment. As shown in the comparison data of solid waste identification accuracy in Tables 3 and 4, various solid waste identification models for recycled plastic particles were established using multiple preprocessing methods (standard normal transformation, second derivative, and standard normal transformation combined with second derivative) and three pattern recognition methods (KNN, PLS-DA, and LS-SVM). Among these, when saturated absorption spectra are included, the KNN algorithm achieved an accuracy of 70%-80%, indicating a mediocre classification effect; the PLS-DA algorithm achieved an overall accuracy of only about 45-50%, showing poor classification performance. The LS-SVM algorithm achieved an accuracy of approximately 80%-100%, with the optimal model achieving 100% accuracy on the training set and 85.27% accuracy on the test set. With acceptable accuracy, the LS-SVM model can be used for the rapid identification of different types of recycled plastic particles.

[0121] Since dark-colored plastics often exhibit saturation absorption in the near-infrared spectrum, a recognition pattern was established by combining preprocessing and pattern recognition algorithms after removing all saturation absorption spectra. The accuracy of the models was significantly improved for KNN, PLS-DA, and LS-SVM, indicating that samples with saturation absorption (dark-colored plastics) have a significant impact on different types of recycled plastic samples. Specifically, KNN improved from 70-80% to 90-95%; PLS-DA from 45-50% to 65-70%; and LS-SVM from 80-100% to 95-100%. After removing all saturation absorption spectra, both KNN and LS-SVM can identify different types of recycled plastic particles with high accuracy.

[0122] In summary, the embodiments of the present invention can quickly and accurately distinguish whether a sample to be tested belongs to solid waste. Specifically, based on the relationship between near-infrared spectral absorbance information and material information, it fully utilizes the infrared absorption spectral characteristics of recycled plastic particles to establish a solid waste identification model for multiple categories of recycled plastic particles. This solid waste identification model can quickly and accurately distinguish the category of recycled plastic particles in the sample to be tested, differentiating whether the sample is recycled plastic particles or solid waste plastic particles. This allows for rapid and accurate detection results in customs inspections, improving the efficiency and accuracy of solid waste identification for samples to be tested.

[0123] The above describes the method for identifying recycled plastic pellet solid waste in the embodiments of the present invention. The following describes the device for identifying recycled plastic pellet solid waste in the embodiments of the present invention. Please refer to [link to relevant documentation]. Figure 2 One embodiment of the solid waste identification device for recycled plastic pellets in this invention includes:

[0124] The spectral information dataset acquisition module 11 is used to acquire the spectral information dataset of recycled plastic particles, which includes a training set and a test set.

[0125] The spectral preprocessing module 12 is used to preprocess the spectral data in the training set using the standard normal transformation algorithm, and to perform second-order derivative processing on the spectral data obtained after standard normal transformation based on window optimization selection, so as to optimize the spectral data in the training set.

[0126] The solid waste identification model building module 13 is used to select a predetermined pattern recognition algorithm to train the model on the near-infrared spectrum in the training set, and determine the target conditions to achieve the highest identification and classification accuracy under preset conditions, and build a solid waste identification model for recycled plastic particles under the target conditions.

[0127] The identification and determination module 14 is used to acquire the near-infrared spectrum of the recycled plastic particle sample to be detected, and input the near-infrared spectrum into the recycled plastic particle solid waste identification model for identification, and determine whether the recycled plastic particle sample belongs to solid waste.

[0128] Optionally, in another embodiment of the recycled plastic pellet solid waste identification device, the device further includes:

[0129] The spectral acquisition module is used to acquire the near-infrared spectra of a preset number of recycled plastic granule samples using a predetermined diffuse backscan method, thereby obtaining the near-infrared spectrum of the recycled plastic granules.

[0130] The sample rejection module is used to reject sample categories whose number of spectra in the near-infrared spectrum is less than a preset value, thereby obtaining a spectral information dataset of recycled plastic particle samples. The spectral information dataset of recycled plastic particle samples includes each sample category, the near-infrared spectrum corresponding to each sample category, and the spectral data contained in the near-infrared spectrum, and whether the sample category belongs to solid waste.

[0131] The dataset partitioning module is used to select near-infrared spectra for each sample category in the spectral information dataset according to a first predetermined ratio, using a systematic sampling method with a set interval distance, so as to divide the spectral information dataset into a training set for training the model and a test set for testing the recognition effect of the model according to a second predetermined ratio.

[0132] Optionally, in another embodiment of the recycled plastic pellet solid waste identification device, the device further includes:

[0133] The saturated absorption spectrum removal module is used to remove near-infrared spectra containing saturated absorption spectra from the spectral information dataset of the recycled plastic particle sample.

[0134] Optionally, in another embodiment of the recycled plastic particle solid waste identification device, the spectral preprocessing module specifically includes:

[0135] The standard normal variable transformation processing unit is used to perform a standard normal variable transformation on each spectrum in the spectral data according to a preset formula, to obtain the spectral data after standard normal transformation processing. The preset formula is: in m is the number of spectral spot lengths, and k = 1, 2, ..., m;

[0136] The second-order derivative processing unit is used to perform polynomial least squares fitting on the spectral points within the window after the standard normal transformation of the spectral data. The data of the equidistant points within the window are fitted into a second-order polynomial according to the preset window size, and the spectral data after second-order derivative processing is obtained.

[0137] Optionally, in another embodiment of the recycled plastic pellet solid waste identification device, the solid waste identification model building module specifically includes:

[0138] The nearest neighbor algorithm training unit is used to select the nearest neighbor algorithm to train the model on the near-infrared spectrum in the training set, and to select different K values ​​using ten-fold cross-validation to verify the classification accuracy of the model.

[0139] The K-value determination and model building unit is used to select the K-value corresponding to the highest accuracy as the final K-value, and to build a first identification model for recycled plastic pellet solid waste based on the K-value.

[0140] The first model verification unit is used to verify the classification accuracy of the first recycled plastic particle solid waste identification model based on the test set used to test the model recognition effect, so as to obtain the verified recycled plastic particle solid waste identification model.

[0141] Optionally, in another embodiment of the recycled plastic pellet solid waste identification device, the solid waste identification model building module specifically includes:

[0142] The partial least squares classification and discrimination training unit is used to train the model on the near-infrared spectra in the training set using the partial least squares classification and discrimination algorithm. The maximum number of latent variables is set to 10, and different numbers of latent variables are selected using ten-fold cross-validation to verify the classification accuracy of the model.

[0143] The optimal number of latent variables determination and model building unit is used to select the number of latent variables corresponding to the highest accuracy as the optimal number of latent variables, and to build a second recycled plastic particle solid waste identification model based on the number of latent variables.

[0144] The second model verification unit is used to verify the classification accuracy of the second recycled plastic particle solid waste identification model based on the test set used to test the model recognition effect, so as to obtain the verified recycled plastic particle solid waste identification model.

[0145] Optionally, in another embodiment of the recycled plastic pellet solid waste identification device, the solid waste identification model building module specifically includes:

[0146] The Partial Least Squares Support Vector Machine (PLS Support Vector Machine) training unit is used to train the model on the near-infrared spectra in the training set using the PLS Support Vector Machine algorithm. The training set is divided into several binary categories, and eight pairs are formed by combining four pairwise methods (error correction output coding, minimum output coding, one-to-one and one-to-many) and two optimization methods (grid search and simplex search).

[0147] The optimal combination determination and model building unit is used to compare and verify the classification accuracy of the eight combinations, select the combination with the highest accuracy as the optimal combination, and build a third recycling plastic particle solid waste identification model based on the optimal combination.

[0148] The third model verification unit is used to verify the classification accuracy of the third recycled plastic particle solid waste identification model based on the test set used to test the model recognition effect, so as to obtain the verified recycled plastic particle solid waste identification model.

[0149] It should be noted that the apparatus in the embodiments of the present invention can be used to implement all the technical solutions in the above method embodiments. The functions of each functional module can be specifically implemented according to the methods in the above method embodiments. The specific implementation process can be referred to the relevant descriptions in the above examples, and will not be repeated here. Figure 2 The solid waste identification device for recycled plastic particles in the embodiments of the present invention will be described in detail from the perspective of modular functional entities. The solid waste identification device for recycled plastic particles in the embodiments of the present invention will be described in detail from the perspective of hardware processing.

[0150] Figure 3This is a schematic diagram of the structure of a solid waste identification device for recycled plastic particles provided in an embodiment of the present invention. The solid waste identification device 300 for recycled plastic particles can vary considerably due to different configurations or performance. It may include one or more central processing units (CPUs) 301 (e.g., one or more processors) and a memory 309, and one or more storage media 308 (e.g., one or more mass storage devices) for storing application programs 307 or data 306. The memory 309 and storage media 308 can be temporary or persistent storage. The program stored in the storage media 308 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations stored in Boolean variables for graph computation. Furthermore, the processor 301 may be configured to communicate with the storage media 308 and execute the series of instruction operations in the storage media 308 on the solid waste identification device 300 for recycled plastic particles.

[0151] The recycled plastic pellet solid waste identification device 300 may also include one or more power supplies 302, one or more wired or wireless network interfaces 303, one or more input / output interfaces 304, and / or one or more operating systems 305, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 The structure of the solid waste identification device for recycled plastic pellets shown does not constitute a limitation on the solid waste identification device for recycled plastic pellets. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0152] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0153] In the embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between devices or units through some interfaces, and may be electrical, mechanical, or other forms.

[0154] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0155] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0156] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium, which can be non-volatile or volatile. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0157] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying solid waste from recycled plastic pellets, characterized in that, The method for identifying recycled plastic pellet solid waste includes: Obtain a spectral information dataset of recycled plastic particles, the spectral information dataset including a training set and a test set; The spectral data in the training set are preprocessed using the standard normal transformation algorithm, and the spectral data obtained after the standard normal transformation are subjected to second-order derivative processing based on window optimization selection to optimize the spectral data in the training set. A predetermined pattern recognition algorithm is selected to train the model on the near-infrared spectrum in the training set, and the target conditions for achieving the highest recognition accuracy are determined under preset conditions. Under the target conditions, a solid waste identification model for recycled plastic particles is established. The predetermined pattern recognition algorithm includes the nearest neighbor algorithm, the partial least squares classification algorithm, and the least squares support vector machine algorithm. Using a preset test set, the first, second, and third recycling plastic particle solid waste identification models established by the three predetermined pattern recognition algorithms were tested respectively, and the classification accuracy of each model was verified based on the recognition effect of each model, thereby obtaining the verified recycling plastic particle solid waste identification model. The near-infrared spectrum of the recycled plastic particle sample to be tested is obtained, and the near-infrared spectrum is input into the verified recycled plastic particle solid waste identification model for identification, to determine whether the recycled plastic particle sample belongs to solid waste.

2. The method for identifying recycled plastic pellet solid waste according to claim 1, characterized in that, Before obtaining the spectral information dataset of recycled plastic particles, the method further includes: Near-infrared spectra of a predetermined number of recycled plastic granule samples were collected using a predetermined diffuse reflectance method to obtain the near-infrared spectra of the recycled plastic granules. Sample categories with fewer than a preset value of spectra in the near-infrared spectrum are removed to obtain a spectral information dataset of recycled plastic granule samples. The spectral information dataset of recycled plastic granule samples includes each sample category, the near-infrared spectrum corresponding to each sample, and the spectral data contained in the near-infrared spectrum, as well as whether the sample belongs to solid waste. Using a systematic sampling method with a set interval, near-infrared spectra are selected from the spectral information dataset for each sample category according to a first predetermined ratio as training set and test set, respectively, so that the spectral information dataset is divided into a training set for training the model and a test set for testing the recognition effect of the model according to a second predetermined ratio.

3. The method for identifying recycled plastic pellet solid waste according to claim 2, characterized in that, After removing sample categories with fewer than a preset value of spectra in the near-infrared spectrum to obtain the spectral information dataset of recycled plastic particle samples, the method further includes: Near-infrared spectra containing saturated absorption spectra are removed from the spectral information dataset of the recycled plastic particle samples.

4. The method for identifying recycled plastic pellet solid waste according to claim 1, characterized in that, The preprocessing of the spectral data in the training set using the standard normal transform algorithm, and the subsequent second-order derivative processing of the spectral data obtained after the standard normal transform based on window optimization, specifically includes: Each spectrum in the spectral data is subjected to a standard normal transformation according to a preset formula to obtain the spectral data after standard normal transformation. The preset formula is: ,in m is the number of spectral spots, k = 1, 2, ..., m; The spectral data obtained after standard normal transformation is fitted with polynomial least squares on the spectral points within the window. The data of equidistant points within the window are fitted into a second-order polynomial according to the preset window size, and the spectral data after second-order differentiation is obtained.

5. The method for identifying recycled plastic pellet solid waste according to claim 2, characterized in that, The process of selecting a predetermined pattern recognition algorithm to train a model on the near-infrared spectrum in the training set, and determining the target conditions to achieve the highest recognition and classification accuracy under preset conditions, specifically includes establishing a solid waste identification model for recycled plastic particles under the target conditions: The nearest neighbor algorithm was used to train the model on the near-infrared spectrum in the training set, and different K values ​​were selected using ten-fold cross-validation to verify the classification accuracy of the model. The K value corresponding to the highest accuracy is selected as the final K value, and a first identification model for recycled plastic pellet solid waste is established based on this K value. Based on the test set used to test the recognition effect of the model, the classification accuracy of the first recycled plastic particle solid waste recognition model is verified to obtain the verified recycled plastic particle solid waste recognition model.

6. The method for identifying recycled plastic pellet solid waste according to claim 5, characterized in that, The process of training a model on the near-infrared spectrum in the training set using a predetermined pattern recognition algorithm, and determining the target conditions to achieve the highest recognition and classification accuracy under preset conditions, specifically includes establishing a solid waste identification model for recycled plastic particles under the target conditions: The partial least squares classification algorithm was used to train the model on the near-infrared spectrum in the training set, and the maximum number of latent variables was set to 10. The ten-fold cross-validation method was used to select different numbers of latent variables to verify the classification accuracy of the model. The optimal number of latent variables is selected based on the number of latent variables corresponding to the highest accuracy, and a second identification model for recycled plastic pellet solid waste is established based on this number of latent variables. Based on the test set used to test the recognition effect of the model, the classification accuracy of the second recycled plastic particle solid waste recognition model is verified to obtain the verified recycled plastic particle solid waste recognition model.

7. The method for identifying recycled plastic pellet solid waste according to claim 5, characterized in that, The step of using a predetermined pattern recognition algorithm to train a model on the near-infrared spectrum in the training set, and determining the target conditions to achieve the highest recognition and classification accuracy under preset conditions, specifically includes establishing a solid waste identification model for recycled plastic particles under the target conditions: The least squares support vector machine algorithm is used to train the near-infrared spectrum in the training set. The multiple categories in the training set are divided into several binary categories. Eight pairs of combinations are formed by using four pairwise methods (error correction output coding, minimum output coding, one-to-one and one-to-many) and two optimization methods (grid search and simplex search). The classification accuracy of the eight combinations was compared and verified. The combination with the highest accuracy was selected as the optimal combination. A third solid waste identification model for recycled plastic particles was established based on the optimal combination. Based on the test set used to test the recognition effect of the model, the classification accuracy of the third recycled plastic particle solid waste recognition model is verified to obtain the verified recycled plastic particle solid waste recognition model.

8. A device for identifying solid waste of recycled plastic pellets, characterized in that, The device includes: The spectral information dataset acquisition module is used to acquire the spectral information dataset of recycled plastic particles, which includes a training set and a test set. The spectral preprocessing module is used to preprocess the spectral data in the training set using the standard normal transformation algorithm, and to perform second-order derivative processing on the spectral data obtained after standard normal transformation based on window optimization selection, so as to optimize the spectral data in the training set. The solid waste identification model building module is used to select a predetermined pattern recognition algorithm to train the model on the near-infrared spectrum in the training set, and determine the target conditions to achieve the highest identification and classification accuracy under preset conditions. Under the target conditions, a solid waste identification model for recycled plastic particles is built. The predetermined pattern recognition algorithm includes the nearest neighbor algorithm, the partial least squares classification algorithm, and the least squares support vector machine algorithm. The testing and verification module is used to test the first, second, and third recycled plastic particle solid waste identification models established by the three predetermined pattern recognition algorithms using a preset test set, and to verify the classification accuracy of each model based on its recognition effect, thereby obtaining the verified recycled plastic particle solid waste identification model. The identification and determination module is used to acquire the near-infrared spectrum of the recycled plastic particle sample to be tested, and input the near-infrared spectrum into the verified recycled plastic particle solid waste identification model for identification, and determine whether the recycled plastic particle sample belongs to solid waste.

9. A device for identifying solid waste from recycled plastic pellets, characterized in that, The recycled plastic particle solid waste identification device includes a memory and at least one processor. The memory stores instructions, and the memory and the at least one processor are interconnected via a circuit. The at least one processor invokes the instructions in the memory to cause the recycled plastic particle solid waste identification device to perform the recycled plastic particle solid waste identification method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for identifying recycled plastic pellet solid waste as described in any one of claims 1-7.

Images