Portable plastic type identification system based on near infrared spectrum

Abstract

The invention discloses a portable plastic type identification system based on a near infrared spectrum, which realizes rapid identification of different plastic types by combining a hardware circuit with a classification identification model, and comprises a ZYNQ processing system, an STM32 processing system, a near infrared spectrometer module and an interactive display module, the STM32 processing system receives an operation instruction and controls the near-infrared spectrometer module to collect near-infrared spectrum intensity data of a plastic sample in real time, then reflectivity data is obtained through operation and then transmitted to the ZYNQ processing system for data processing, and therefore the reflectivity data is used for iterative training and prediction of a classification and recognition model, and finally a result is displayed. The system has the characteristics of high integration level, small size, real-time spectrum acquisition and processing and the like, so that spectrum acquisition and processing, model iteration training prediction and result display can be carried out in real time in one hardware circuit board, and the identification system is enabled to meet the application requirements of various plastic recovery sites.

Description

Technical Field

[0001] The invention belongs to the application field of near infrared spectrum technology for plastic variety identification, in particular to a portable plastic variety identification system based on near infrared spectrum. Background Art

[0002] With the popularity of plastic products in modern society, plastic has become an indispensable material in production and life in all walks of life. From daily necessities such as packaging bags and bottles and cans to mechanical parts and electronic product casings in the industrial field, the application of plastic is everywhere. With its characteristics of lightness, corrosion resistance, and easy processing, plastic has demonstrated irreplaceable advantages in many scenarios. However, the popularity of plastic has also brought huge challenges in terms of environment and resources. In particular, due to the non-biodegradability of plastic, plastic waste has had a long-term and widespread negative impact on the natural environment. Therefore, how to efficiently and accurately identify different types of plastics in real time has become one of the key technologies in the field of plastic recycling.

[0003] There are many types of plastics on the market, and with the continuous development of science and technology, the types continue to increase. The usage scenarios and performance characteristics of each plastic are different. If plastics are not sorted and mixed for recycling, the quality of the final product may be reduced, resulting in waste of resources and even safety hazards in serious cases. At present, traditional plastic recycling identification methods mainly rely on manual sorting, mechanical identification and optical identification technology. Manual sorting has problems such as high labor costs, low efficiency and high error rate; although mechanical identification has improved efficiency, it still faces difficulties in identifying some similar plastic materials; optical identification technology, such as near-infrared spectroscopy-based identification, has been widely used in plastic recycling due to its non-destructive, contactless and rapid advantages.

[0004] Through patent search, it was found that the publication number is: CN116106256A, which discloses an optimization system and method based on near-infrared plastic classification. The system includes a near-infrared spectrum acquisition module, a computer, a separation nozzle, etc. The near-infrared spectrum acquisition module includes a sensor, a light source, and a spectrometer module. The collected spectrum is sent to the computer. The computer establishes a plastic classification model based on the information, performs feature extraction and preprocessing, and then realizes plastic sorting through the separation nozzle.

[0005] Publication number: CN119323691A, disclosed is a method, device, computer equipment and storage medium for identifying microplastic species, the method comprising: receiving a fluorescence spectrum image to be classified sent by a user terminal; performing a preprocessing operation on the fluorescence spectrum image to obtain a preprocessed spectrum image; calling a trained spectrum classification model to perform a species identification operation to obtain a species identification result; but the above methods all rely on computers for data processing and model building, lack of highly integrated hardware, making the system volume large and requiring external equipment support, and low convenience and flexibility of on-site operation. To solve the above problems, the system proposed in this application adopts a highly integrated hardware circuit (including ZYNQ and STM32 processing systems), so that spectrum acquisition, data processing, prediction and result display can all be implemented on a compact circuit board. This high integration brings a smaller size and higher portability, which is better than the previous two inventions, especially more flexible and efficient in practical applications. Summary of the invention

[0006] The present invention aims to solve the deficiencies of the above-mentioned prior art and proposes a portable plastic type identification system based on near-infrared spectroscopy, in order to improve the efficiency, identification accuracy and automation level of plastic recycling through the collaborative processing of ZYNQ and STM32 main control systems, thereby reducing manual intervention and reducing operating costs, and providing an efficient and intelligent solution for the plastic recycling industry.

[0007] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme:

[0008] The portable plastic type identification system based on near infrared spectroscopy of the present invention is characterized in that it is a system composed of a ZYNQ processing system, an STM32 processing system, a near infrared spectrometer module, and an interactive display module, wherein the ZYNQ processing system includes: a data processing module, a classification identification module, a PS end data communication module, and a PL end data communication module; the STM32 processing system includes: a spectrum data acquisition module, a spectrum data calculation module, and a data communication module; the near infrared spectrometer module is composed of a near infrared spectrometer and a calibration whiteboard; the plastic type identification system is performed according to the following steps:

[0009] S1: The interactive display module obtains the working status configuration instruction of the near-infrared spectrometer and sends it to the spectral data acquisition module, which analyzes the configuration instruction, obtains the configuration signal and sends it to the near-infrared spectrometer through serial communication, so as to configure the gain register and the repeated acquisition times register in the internal register of the near-infrared spectrometer;

[0010] S2: After the configuration is completed, the measurement port of the near-infrared spectrometer is aligned with the calibration whiteboard, and the interactive display module obtains the calibration instruction and sends it to the spectral data acquisition module, and the spectral data acquisition module analyzes the calibration instruction, obtains the calibration signal and sends it to the near-infrared spectrometer through serial communication, so that the near-infrared spectrometer controls the sampling accuracy according to the gain value configured in the gain register and collects the near-infrared spectrum intensity of the calibration whiteboard within the set range according to the number of repetitions configured in the repetition acquisition number register =( ),in, Indicates that the calibration white plate is The intensity value at each position, , Indicates the total number of bands;

[0011] S3: Align the measuring port of the near-infrared spectrometer with the plastic sample, and at the same time, the interactive display module obtains the spectrum data acquisition instruction and sends it to the spectrum data acquisition module, so that the spectrum data acquisition module analyzes the acquisition instruction, obtains the acquisition signal and sends it to the near-infrared spectrometer through serial communication, so as to control the near-infrared spectrometer to collect the near-infrared spectrum intensity data of a plastic sample within the set range. =( ),in, Indicates that the plastic sample is The intensity value at each position;

[0012] S4: The spectrum data calculation module is based on and , using formula (1) to obtain the near-infrared spectral reflectance data of the plastic sample at the i-th position , thus obtaining the near-infrared spectral reflectance data vector of the plastic sample =( );

[0013] (1)

[0014] In formula (1), D represents the dark background spectrum;

[0015] S5: The spectral data calculation module uses the data communication module to obtain the near infrared spectral reflectance data vector After format conversion, it is sent to the ZYNQ processing system to eliminate sample surface scattering and remove sample noise, and the near-infrared spectral reflectance data vector is obtained after data processing. ,in, Represents the near-infrared spectral reflectance data of a plastic sample at the i-th position after data processing;

[0016] S6: Obtain according to the process of S1-S5 After the near-infrared spectral reflectance data of the plastic samples are processed, the near-infrared spectral reflectance set after data processing is recorded as ,in, and They represent the near-infrared spectral reflectance data of the ath and bth plastic samples after data processing, ;make Collection of tags ,in, and They are and The corresponding real plastic type label;

[0017] S7: The classification and recognition module constructs a support vector machine classification and recognition model for the plastic type recognition system, and classifies the near infrared spectral reflectance set and Processing is performed to obtain the optimal normal vector of the decision hyperplane of the support vector machine and the optimal bias term ;

[0018] S8: Obtain the near-infrared spectral reflectance data vector of the plastic sample to be predicted according to the process of S1-S5 , and put it into formula (10), we get Predicted plastic type labels ;

[0019] (10)

[0020] In formula (10), Yes Representation mapped to high-dimensional feature space; T represents transposition;

[0021] S9: The interactive display module obtains the save data instruction and sends it to the data processing module, which is used to convert the near-infrared spectral reflectance data vector of the plastic sample to be predicted. By predicted plastic type label The plastic samples are numbered and saved in the SD card to collect the data collected by the hardware. Finally, the classification and recognition module transmits the data to the PL end data communication module. Sent to the interactive display module for display.

[0022] The portable plastic type identification system based on near infrared spectroscopy of the present invention is also characterized in that the format conversion, sample surface scattering elimination and sample noise removal operations in S5 include:

[0023] S51: The spectral data operation module converts the spectral data into Convert to a reflectivity string collection =( ),in, represents the reflectivity string of the plastic sample at the i-th position, and then the data communication module transmits Send to ZYNQ processing system;

[0024] S52: The PS-side data communication module in the ZYNQ processing system receives And send it to the data processing module;

[0025] S53: The data processing module uses formula (2) to obtain the corrected near-infrared spectral reflectance data of the plastic sample at the i-th position, thereby obtaining a corrected near-infrared spectral reflectance data set ;

[0026] (2)

[0027] In formula (2), for The mean of ; for The standard deviation of ; and are the reference mean and reference standard deviation, respectively;

[0028] S54: The data processing module uses formula (3) to obtain the near-infrared spectral reflectance data of the plastic sample at the i-th position after data processing: , thus obtaining the near-infrared spectral reflectance data set after data processing ;

[0029] (3)

[0030] In formula (3), m is the half window size of the filter, is the corrected near-infrared spectral reflectance data of the i+jth position in the window, is the coefficient of the jth position in the window.

[0031] Furthermore, S7 includes:

[0032] S71: The classification and recognition module uses formula (4) to construct a polynomial function of the classification and recognition model ( , );

[0033] (4)

[0034] In formula (4), is a constant term, It is the degree;

[0035] S72: Use formula (5) to construct the objective function ;

[0036] (5)

[0037] In formula (5), is the normal vector of the hyperplane, is the bias term, yes The Lagrange multiplier of ; T represents the transpose;

[0038] S73: Using equations (6) and (7), we can obtain the dualized objective function: And its constraints and solve them, we get Near infrared spectral reflectance data at ;

[0039] (6)

[0040] St: (7)

[0041] In formula (6), yes The Lagrange multiplier of ;

[0042] S74: Using equations (8) and (9), we can obtain the optimal normal vector of the decision hyperplane of the classification recognition model: and the optimal bias term ;

[0043] (8)

[0044] (9)

[0045] In formula (8) and formula (9), yes Representation mapped to a high-dimensional feature space.

[0046] The electronic device of the present invention includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the plastic identification system, and the processor is configured to execute the program stored in the memory.

[0047] The present invention provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and the computer program executes the steps of the plastic identification system when the computer program is executed by a processor.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. The present invention integrates functions such as human-computer interaction, spectral data acquisition, preprocessing, spectral data transmission, storage and model classification prediction; thereby avoiding the use of PC acquisition software and dependence on PC software algorithms, making the recognition system suitable for various plastic recycling site application requirements, and at the same time having efficient and real-time feedback for plastic recognition.

[0050] 2. The present invention adopts a software and hardware combination of the ZYNQ processing system and the STM32 processing system, which improves the flexibility and scalability of the system; at the same time, the two independent processing system designs are integrated into the same hardware circuit board, and an integrated power supply method is adopted, which enhances the stability and reliability of the system, and also makes the system hardware small in size, low in power consumption, and easy to carry and move.

[0051] 3. The present invention interacts with users through an interactive display interface, which significantly improves the system's usability, operating efficiency, plastic identification efficiency and spectral data scheduling capabilities, and improves the system's intelligence level and enhances market competitiveness.

[0052] 4. The present invention can identify the types of plastics in real time online, realize automatic work, and can quickly identify different types of plastics and classify them. Compared with the traditional manual identification method, the automatic detection system can work continuously, reduce the hysteresis effect in identification, reduce manual errors, and improve work efficiency.

[0053] 5. The support vector machine classification and recognition model of the present invention can accurately identify different types of plastic materials according to the molecular structure characteristics by extracting multi-dimensional features, so it has a high classification accuracy for different types of plastics. Whether it is PE, PVC, PP, PC and other types of plastics, the classification and recognition model can be applied in a large range, and plastic recognition shows strong robustness.

[0054] 6. The present invention is a non-destructive detection method. Compared with the traditional chemical detection method, it does not involve the use of chemical reagents. This method not only meets the environmental protection requirements, but also can reduce the potential hazards in identification, and is therefore more environmentally friendly. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a schematic diagram of the connection of the identification system device of the present invention;

[0056] Figure 2 This is a diagram of the plastic type identification system of the present invention;

[0057] Figure 3 The hardware-based collection flow chart of the plastic spectrum data of the present invention;

[0058] Figure 4 It is the average spectrum of seven plastics of the present invention. DETAILED DESCRIPTION

[0059] In this embodiment, a portable plastic type identification system based on near infrared spectroscopy is constructed in the following manner: Figure 1 The plastic spectral data acquisition, processing, communication and storage system device shown is composed of a ZYNQ processing system, an STM32 processing system, a near-infrared spectrometer module, and an interactive display module, wherein the near-infrared spectrometer module is composed of a near-infrared spectrometer and a calibration whiteboard; the interactive display module has the functions of controlling the hardware system to perform data acquisition, data processing and result display; the entire system is powered by a 9V-3A power adapter, and the system hardware is powered by voltage matching, such as: ZYNQ and STM32 processing systems use 9V power supply, the interactive display module uses 5V power supply, and the near-infrared spectrometer uses 3.3V power supply.

[0060] Specifically, Figure 2 As shown, the ZYNQ processing system includes: data processing module, classification and recognition module, PS-side data communication module, and PL-side data communication module; the STM32 processing system includes: spectral data acquisition module, spectral data calculation module, and data communication module; the data communication modules all use the UART serial communication protocol for data exchange; the plastic type recognition system is carried out in the following steps:

[0061] S1: Turn on the system power, the system automatically performs initialization operations, the ZYNQ processing system and the STM32 processing system load the application program from the SPI Flash respectively, and the interactive display module loads the executable file from the SD card. After the system initialization is completed, the completion information will be displayed on the interactive display module;

[0062] The near infrared spectroscopy data collection process is as follows Figure 3 As shown, the interactive display module obtains the working status configuration instruction of the near-infrared spectrometer and sends it to the spectral data acquisition module. The spectral data acquisition module analyzes the configuration instruction, obtains the configuration signal and sends it to the near-infrared spectrometer through UART1 serial port communication, which is used to configure the gain register and the repeated acquisition times register in the internal register of the near-infrared spectrometer, wherein the gain register value is configured to 64, and the repeated acquisition times register value is configured to 6.

[0063] S2: After the configuration is completed, the measurement port of the near-infrared spectrometer is aligned with the calibration whiteboard, and the interactive display module obtains the calibration instruction and sends it to the spectral data acquisition module. The spectral data acquisition module analyzes the calibration instruction, obtains the calibration signal and sends it to the near-infrared spectrometer through serial communication, so that the near-infrared spectrometer controls the sampling accuracy according to the gain value configured in the gain register and collects the near-infrared spectrum intensity of the calibration whiteboard in the range of 950nm to 1680nm according to the number of repetitions configured in the repetition acquisition number register. =( ) ,in, Indicates that the calibration white plate is The intensity value at each position, , Indicates the total number of bands, which is set to 228.

[0064] S3: The measuring port of the near-infrared spectrometer is aimed at the plastic sample, and the interactive display module obtains the spectrum data acquisition instruction and sends it to the spectrum data acquisition module, so that the spectrum data acquisition module analyzes the acquisition instruction, obtains the acquisition signal and sends it to the near-infrared spectrometer through serial communication, which is used to control the near-infrared spectrometer to collect near-infrared spectrum intensity data of a plastic sample in the range of 950nm to 1680nm =( ),in, Indicates that the plastic sample is The intensity value at each position.

[0065] S4: Spectral data calculation module reads out in RAM and , and use formula (1) to obtain the near-infrared spectral reflectance data of the plastic sample at the i-th position , thus obtaining the near-infrared spectral reflectance data vector of the plastic sample =( );

[0066] (1)

[0067] In formula (1), D represents the dark background spectrum. The dark background spectrum is used to deduct the electronic noise baseline of the near-infrared spectrometer (mainly from the InGaAs detector and circuit board inside the near-infrared spectrometer). The acquisition method is to avoid the plastic sample by placing the measurement port of the near-infrared spectrometer in the air and measure it downwards towards the air. The dark background spectrum is obtained.

[0068] S5: The spectral data operation module uses the data communication module to generate the near-infrared spectral reflectance data vector After format conversion, it is transmitted to the ZYNQ processing system to eliminate sample surface scattering and remove sample noise, and obtain the near-infrared spectral reflectance data vector after data processing. ,in, Represents the near-infrared spectral reflectance data of a plastic sample at the i-th position after data processing;

[0069] S51: The spectral data operation module converts the Convert to a reflectivity string collection =( ),in, Represents the reflectivity string of the plastic sample at the i-th position, and then the data communication module transmits Sent to the ZYNQ processing system, where the data stream sent by UART2 is converted using a TTL to RS485 chip with a baud rate of 115200 to ensure transmission efficiency and accuracy;

[0070] S52: PS data communication module receiving in ZYNQ processing system It is stored in the base address 0xFF000000 of the PS serial port controller, with an offset of 0x0030U, and then sent to the data processing module;

[0071] S53: The data processing module uses formula (2) to eliminate the error caused by the surface scattering of the plastic sample, and obtains the corrected near-infrared spectral reflectance data of the plastic sample at the i-th position, thereby obtaining a corrected near-infrared spectral reflectance data set. ;

[0072] (2)

[0073] In formula (2), for The mean of ; for The standard deviation of ; and are the reference mean and reference standard deviation, respectively;

[0074] S54: The data processing module uses formula (3) to remove noise in the spectral data while retaining the spectral characteristics, and obtains the near-infrared spectral reflectance data of the plastic sample at the i-th position after data processing. , thus obtaining the near-infrared spectral reflectance data set after data processing ;

[0075] (3)

[0076] In formula (3), m is the half window size of the filter, is the corrected near-infrared spectral reflectance data of the i+jth position in the window, is the coefficient of the jth position in the window.

[0077] S6: Obtain according to the process of S1-S5 After the near-infrared spectral reflectance data of the plastic samples are processed, the near-infrared spectral reflectance set after data processing is recorded as ,in, and They represent the near-infrared spectral reflectance data of the ath and bth plastic samples after data processing, ;

[0078] make Collection of tags ,in, and They are and Corresponding real plastic type labels; collected The types of plastic samples include the following seven types: polyethylene (PE), polypropylene (PP), polyvinyl chloride (PVC), polycarbonate (PC), nylon (PA), terpolymer (ABS) and polyoxymethylene (POM); the near-infrared spectral data of each type of plastic are averaged, such as Figure 4 As shown, it can be seen that different types of plastics have certain differences in their spectral characteristics in the range of 950–1680 nm. These differences indicate that it is feasible to distinguish different types of plastics based on near-infrared spectral data.

[0079] S7: Classification and recognition module constructs a classification and recognition model for the plastic type recognition system and classifies the near infrared spectral reflectance set and Processing is performed by optimizing the category weight, optimization accuracy, degree of polynomial function and fault tolerance value in the DDR of the ZYNQ processing system, and iterative training is performed to obtain the optimal normal vector of the decision hyperplane of the classification recognition model. and the optimal bias term ;

[0080] S71: The classification and recognition module uses formula (4) to construct a polynomial function of the classification and recognition model ( , );

[0081] (4)

[0082] In formula (4), is a constant term, is the degree. In this embodiment, Set to 0, Set to 3, that is ;

[0083] S72: Use formula (5) to maximize the margin in the high-dimensional feature space through the Lagrange multiplier method to construct the objective function ;

[0084] (5)

[0085] In formula (5), is the normal vector of the hyperplane, is the bias term, yes The Lagrange multiplier of ; T denotes the transpose.

[0086] S73: Using equations (6) and (7), we can obtain the dualized objective function: And its constraints and solve them, we get Near infrared spectral reflectance data at ;

[0087] (6)

[0088] St: (7)

[0089] In formula (6), yes The Lagrange multiplier of ;

[0090] S74: Using equations (8) and (9), the optimal normal vector of the decision hyperplane of the classification recognition model is obtained by optimizing the class weight, optimization accuracy, degree of the polynomial function, and fault tolerance value in the DDR of the ZYNQ processing system through iterative training. and the optimal bias term ;

[0091] (8)

[0092] (9)

[0093] In formula (8) and formula (9), yes The representation mapped to a high-dimensional feature space makes the data linearly separable in this high-dimensional feature space.

[0094] S8: Obtain the near-infrared spectral reflectance data vector of the plastic sample to be predicted according to the process of S1-S5 , and put it into formula (10), we get Predicted plastic type labels , the tag That is, the prediction results of unknown plastic samples are obtained in real time through spectral data collection, data processing, and model prediction by the hardware system, realizing the rapid identification of different types of plastics;

[0095] (10)

[0096] In formula (10), Yes Representation mapped to a high-dimensional feature space; T represents transpose.

[0097] S9: The cross-validation accuracy of the classification recognition model is 98.9%. The confusion matrix of the relationship between the prediction results of the classification recognition model and the actual categories is calculated to observe which categories perform better in recognition and which categories have more errors.

[0098] The interactive display module obtains the data saving instruction and sends it to the data processing module, so that the data processing module converts the near infrared spectral reflectance data vector of the plastic sample to be predicted into Label by predicted plastic type The samples are numbered and saved in an SD card to collect data sets based on hardware collection of plastic samples;

[0099] The classification recognition module predicts the plastic type label through the PL end data communication module , The near-infrared spectrum curve and model accuracy within the set range are converted into HEX code format and sent to the interactive display module for display.

[0100] In this embodiment, an electronic device includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the above-mentioned identification system, and the processor is configured to execute the program stored in the memory.

[0101] In this embodiment, a computer-readable storage medium stores a computer program on the computer-readable storage medium, and the computer program executes the steps of the above-mentioned identification system when executed by a processor.

Claims

1. A portable plastic type identification system based on near infrared spectroscopy, characterized in that: The system is composed of a ZYNQ processing system, an STM32 processing system, a near-infrared spectrometer module, and an interactive display module. The ZYNQ processing system includes: a data processing module, a classification recognition module, a PS-side data communication module, and a PL-side data communication module; the STM32 processing system includes: a spectral data acquisition module, a spectral data calculation module, and a data communication module; the near-infrared spectrometer module is composed of a near-infrared spectrometer and a calibration whiteboard; the plastic type recognition system is performed according to the following steps: S1: The interactive display module obtains the working status configuration instruction of the near-infrared spectrometer and sends it to the spectral data acquisition module, which analyzes the configuration instruction, obtains the configuration signal and sends it to the near-infrared spectrometer through serial communication, so as to configure the gain register and the repeated acquisition times register in the internal register of the near-infrared spectrometer; S2: After the configuration is completed, the measurement port of the near-infrared spectrometer is aligned with the calibration whiteboard, and the interactive display module obtains the calibration instruction and sends it to the spectral data acquisition module, and the spectral data acquisition module analyzes the calibration instruction, obtains the calibration signal and sends it to the near-infrared spectrometer through serial communication, so that the near-infrared spectrometer controls the sampling accuracy according to the gain value configured in the gain register and collects the near-infrared spectrum intensity of the calibration whiteboard within the set range according to the number of repetitions configured in the repetition acquisition number register =( ) ,in, Indicates that the calibration white plate is The intensity value at each position, , Indicates the total number of bands; S3: Align the measuring port of the near-infrared spectrometer with the plastic sample, and at the same time, the interactive display module obtains the spectrum data acquisition instruction and sends it to the spectrum data acquisition module, so that the spectrum data acquisition module analyzes the acquisition instruction, obtains the acquisition signal and sends it to the near-infrared spectrometer through serial communication, so as to control the near-infrared spectrometer to collect the near-infrared spectrum intensity data of a plastic sample within the set range. =( ),in, Indicates that the plastic sample is The intensity value at each position; S4: The spectrum data calculation module is based on and , using formula (1) to obtain the near-infrared spectral reflectance data of the plastic sample at the i-th position , thus obtaining the near-infrared spectral reflectance data vector of the plastic sample =( ); (1) In formula (1), D represents the dark background spectrum; S5: The spectral data calculation module uses the data communication module to obtain the near infrared spectral reflectance data vector After format conversion, it is sent to the ZYNQ processing system to eliminate sample surface scattering and remove sample noise, and the near-infrared spectral reflectance data vector is obtained after data processing. ,in, Represents the near-infrared spectral reflectance data of a plastic sample at the i-th position after data processing; S6: Obtain according to the process of S1-S5 After the near-infrared spectral reflectance data of the plastic samples are processed, the near-infrared spectral reflectance set after data processing is recorded as ,in, and They represent the near-infrared spectral reflectance data of the ath and bth plastic samples after data processing, ;make Collection of tags ,in, and They are and The corresponding real plastic type label; S7: The classification and recognition module constructs a support vector machine classification and recognition model for the plastic type recognition system, and classifies the near infrared spectral reflectance set and Processing is performed to obtain the optimal normal vector of the decision hyperplane of the support vector machine and the optimal bias term ; S8: Obtain the near-infrared spectral reflectance data vector of the plastic sample to be predicted according to the process of S1-S5 , and put it into formula (10), we get Predicted plastic type labels ; (10) In formula (10), Yes Representation mapped to high-dimensional feature space; T represents transposition; S9: The interactive display module obtains the save data instruction and sends it to the data processing module, which is used to convert the near-infrared spectral reflectance data vector of the plastic sample to be predicted. Label by predicted plastic type The plastic samples are numbered and saved in the SD card to collect the data collected by the hardware. Finally, the classification and recognition module transmits the data to the PL end data communication module. Sent to the interactive display module for display.

2. The portable plastic type identification system based on near infrared spectroscopy as claimed in claim 1, characterized in that: S5 include: S51: The spectral data operation module converts the spectral data into Convert to a reflectivity string collection =( ),in, represents the reflectivity string of the plastic sample at the i-th position, and then the data communication module transmits Send to ZYNQ processing system; S52: The PS-side data communication module in the ZYNQ processing system receives And send it to the data processing module; S53: The data processing module uses formula (2) to obtain the corrected near-infrared spectral reflectance data of the plastic sample at the i-th position, thereby obtaining a corrected near-infrared spectral reflectance data set ; (2) In formula (2), for The mean of ; for The standard deviation of ; and are the reference mean and reference standard deviation, respectively; S54: The data processing module uses formula (3) to obtain the near-infrared spectral reflectance data of the plastic sample at the i-th position after data processing: , thus obtaining the near-infrared spectral reflectance data set after data processing ; (3) In formula (3), m is the half window size of the filter, is the corrected near-infrared spectral reflectance data of the i+jth position in the window, is the coefficient of the jth position in the window.

3. The portable plastic type identification system based on near infrared spectroscopy as claimed in claim 1, characterized in that: S7 includes: S71: The classification and recognition module uses formula (4) to construct a polynomial function of the classification and recognition model ( , ); (4) In formula (4), is a constant term, It is the degree; S72: Use formula (5) to construct the objective function ; (5) In formula (5), is the normal vector of the hyperplane, is the bias term, yes The Lagrange multiplier of ; T represents the transpose; S73: Using equations (6) and (7), we can obtain the dualized objective function: And its constraints and solve them, we get Near infrared spectral reflectance data at ; (6) St: (7) In formula (6), yes The Lagrange multiplier of ; S74: Using equations (8) and (9), we can obtain the optimal normal vector of the decision hyperplane of the classification recognition model: and the optimal bias term ; (8) (9) In formula (8) and formula (9), yes Representation mapped to a high-dimensional feature space.

4. An electronic device, comprising a memory and a processor, characterized in that: The memory is used to store a program that supports a processor to execute the plastic identification system according to any one of claims 1 to 3, and the processor is configured to execute the program stored in the memory.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the plastic identification system according to any one of claims 1 to 3 are executed.

Images