A plastic PET identification method based on hyperspectral technology

Abstract

The invention discloses a plastic PET identification method based on hyperspectral technology, which can realize accurate identification of plastic PET, save manpower and material resources, and reduce investment costs. The method comprises: obtaining hyperspectral original image data of mixed plastics based on different equipment; processing the hyperspectral original image data to obtain hyperspectral reflectance image data; obtaining spectral wavelengths for distinguishing plastic PET from non-plastic PET by analyzing the variation law of spectral reflectance curves of different plastics in the hyperspectral reflectance image data; calculating normalized indexes and ratio indexes of different plastic pixels in the hyperspectral reflectance image data; obtaining normalized index thresholds, ratio index thresholds, and spectral reflectance thresholds for distinguishing plastic PET from non-plastic PET; extracting a plastic PET area in the hyperspectral reflectance image data to obtain an identification effect diagram of the plastic PET.

Description

Technical Field

[0001] The invention relates to the technical field of plastic PET identification, and in particular to a plastic PET identification method based on hyperspectral technology. Background Art

[0002] The role and significance of plastic recycling are very important. In today's increasingly deteriorating environment with increasingly scarce resources, plastic recycling has a place. It is not only beneficial to the protection of the environment and human health, but also beneficial to the production of the plastic industry and the sustainable development of the country. The outlook for plastic recycling is also optimistic. From the perspective of today's environmental and social needs, plastic recycling is the best way to deal with plastics that consume a lot of oil, are difficult to decompose, and damage the environment.

[0003] Polyethylene terephthalate (PET) is the most common plastic in plastic recycling. It has the characteristics of strength-to-weight ratio, chemical resistance, transparency, shatterproof, non-toxic and odorless. It is widely used in film sheets, packaging bottles, electronic appliances, auto parts, mechanical equipment and other fields. Since PET is difficult to degrade naturally, it pollutes the environment. At the same time, due to the high price of PET raw materials, about 7,000 yuan / t to 13,000 yuan / t, its recycling price is about 5,000 yuan / t. Therefore, sorting and recycling PET has high environmental protection significance and economic value.

[0004] The national standard "Technical Specification for Waste Plastic Recycling" (GB / T 39171-2020) was officially released on October 11, 2020 and officially implemented on May 1, 2021. The standard clearly requires that plastic sorting should follow the principles of stability and no secondary pollution, and that products with a single component purity of 90% after sorting can be used for high-value utilization in the later stage. At present, there are many methods for sorting and recycling plastic PET, such as electrostatic sorting, dissolution sorting, density sorting, manual sorting and identification, photoelectric sorting (using infrared / near-infrared spectroscopy), etc.

[0005] Among them, the electrostatic sorting method consumes a lot of electricity; the dissolution sorting method and density sorting method are accompanied by the discharge and treatment of waste solution, which increases environmental pressure and enterprise costs; manual sorting and identification are highly dependent on the experience of workers and are subject to errors caused by human errors; the photoelectric sorting method uses the characteristic differences of different plastics in the near-infrared spectrum to distinguish between plastic PET and non-plastic PET. The energy consumption of the sorting process is lower than that of the first three methods, and there are no additional emissions to be dealt with. However, the current prices of plastic PET photoelectric sorting machines at home and abroad range from 200,000 to more than 1 million. The price is expensive and the cost is high, which limits its application in small and medium-sized enterprises. Summary of the invention

[0006] In view of the above problems existing in the prior art, the present invention provides a plastic PET identification method based on hyperspectral technology, which can realize accurate identification of plastic PET, save manpower and material resources, reduce investment costs, and reduce pressure on environmental protection.

[0007] A method for identifying plastic PET based on hyperspectral technology, characterized in that the method comprises:

[0008] S1. Spread the mixed plastics on a black background and use different hyperspectral imagers to collect the hyperspectral raw image data of the mixed plastics respectively;

[0009] S2, preprocessing the hyperspectral raw image data to obtain hyperspectral reflectance image data;

[0010] S3, obtaining a spectral wavelength for distinguishing plastic PET from non-plastic PET according to the difference in spectral reflectance of different plastics in the hyperspectral reflectance image data;

[0011] S4. Calculate the normalized index and ratio index of different plastic pixels based on the spectral reflectance of different plastics in the hyperspectral reflectance image data;

[0012] S5, obtaining a normalized index threshold, a ratio index threshold, and a spectral reflectance threshold for distinguishing plastic PET from non-plastic PET;

[0013] S6. Extract the plastic PET area in the hyperspectral reflectance image data based on the spectral reflectance threshold, the normalized index threshold, and the ratio index threshold, obtain a recognition effect diagram of the plastic PET, and realize plastic PET recognition.

[0014] Its further characteristic is that

[0015] Before collecting the hyperspectral raw image data, a hyperspectral imager for collecting the hyperspectral raw image data is selected, and the spectral range of the hyperspectral imager is 900nm to 2350nm;

[0016] The hyperspectral imager includes: at least one short-wave infrared camera using linear gradient filter imaging technology, and at least two short-wave infrared cameras using transmission grating imaging technology;

[0017] Step S1, the plastic mixture includes plastic PET, plastic PP, plastic PE, plastic PVC, plastic ABS and heat-resistant resin;

[0018] Step S2, preprocessing the hyperspectral original image, that is, converting the digital quantization value of the hyperspectral original image data into spectral reflectance, and the conversion formula is:

[0019]

[0020] Among them, R ref Represents the reflectivity value, R ref Less than or equal to 1, DN raw Represents the brightness value of the image pixel of the original data, DN white is the image pixel brightness value of the reference plate, DN dark represents the instrument dark background noise, R white represents the spectral reflectance of the reference plate;

[0021] Step S3, the spectral wavelength used to distinguish plastic PET from non-plastic PET is a dual-channel wavelength, including: 1630nm and 1700nm;

[0022] Step S4, the normalized index NI is calculated as follows: NI = (R ref1700 -R ref 1630 ) / (R ref 1700 +R ref 1630 );

[0023] Among them, R ref 1700 Indicates the spectral reflectance of plastic PET or non-plastic PET at a wavelength of 1700, R ref 1630 Indicates the spectral reflectance of plastic PET or non-plastic PET at a wavelength of 1630;

[0024] The calculation formula of the ratio index RI is: RI = R ref 1700 / R ref 1630 ;

[0025] In step S5, the normalized index threshold is determined according to the frequency distribution difference of the normalized index NI of the plastic PET and the non-plastic PET in the hyperspectral reflectance image data; the ratio threshold is determined according to the difference between the dynamic change range of the ratio index of the plastic PET and the dynamic change range of the ratio index of the non-plastic PET; the spectral reflectance threshold is obtained based on the frequency distribution difference of the spectral reflectance values ​​of the plastic PET and the non-plastic PET pixels at wavelengths of 1630nm and 1700nm in the hyperspectral reflectance image data;

[0026] The spectral reflectance value of 0.1 at a wavelength of 1630 nm is the spectral reflectance threshold for distinguishing the plastic PET from the non-plastic PET, the normalized index threshold is 0.05, and the ratio index threshold is 1.2.

[0027] The following beneficial effects can be achieved by adopting the above method of the present invention: the plastic PET identification method of the present application adopts a comprehensive judgment method of spectral reflectance threshold, normalized index threshold, and ratio index threshold to identify the plastic PET in the plastic mixture. The method can accurately obtain the area where the plastic PET is located, thereby realizing accurate identification of the plastic PET. The identification is based on hyperspectral technology, and there is no problem of discharge and treatment of waste solution, human error, etc., which saves manpower and material resources, thereby reducing investment costs and pressure on environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. The drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0029] Figure 1 The spectral reflectance curves of the plastic PET, plastic PP, plastic PE, plastic PVC, plastic ABS, heat-resistant resin and background of the present invention;

[0030] Figure 2 A frequency distribution diagram of normalized indices of plastic PET pixels and non-plastic PET pixels in the hyperspectral reflectance image of the present invention;

[0031] Figure 3 A frequency distribution diagram of the ratio index of plastic PET pixels and non-plastic PET pixels in the hyperspectral reflectance image of the present invention;

[0032] Figure 4 It is a frequency distribution diagram of the spectral reflectance values ​​of plastic PET pixels and non-plastic PET pixels at 1630nm in the high spectral reflectance image of the present invention;

[0033] Figure 5 This is a diagram showing the recognition effect of the hyperspectral reflectance image collected by the camera Lambda-NIR and the plastic PET in the hyperspectral reflectance image of the present invention;

[0034] Figure 6 This is a diagram showing the recognition effect of the plastic PET in the hyperspectral reflectance image acquired by the camera GaiaField Pro-N17E and the hyperspectral reflectance image of the present invention;

[0035] Figure 7 This is a diagram showing the recognition effect of the plastic PET in the hyperspectral reflectance image acquired by the camera GaiaField Pro-N235E-XE and the hyperspectral reflectance image in the present invention;

[0036] Figure 8It is a flow chart of the plastic PET identification method of the present invention. DETAILED DESCRIPTION

[0037] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0038] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product or equipment comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0039] Plastic PET is an environmentally friendly polymer material. According to its form and main use, the product can be divided into fiber-grade PET, bottle-grade PET and film-grade PET. Fiber-grade PET is mainly used in the downstream textile industry, while bottle-grade PET and film-grade PET are mainly used in the downstream packaging industry. In recent years, global demand has grown steadily. From 2014 to 2020, the global market demand for bottle-grade PET increased from 19.34 million tons to 28.76 million tons, with a compound annual growth rate of 6.84%, which is generally higher than the global bottle-grade PET production capacity and output growth rate in the same period. It has shown an accelerated growth trend since 2015, indicating that bottle-grade PET has a good market prospect as a new type of environmentally friendly packaging material. Therefore, it is of great significance and economic value to sort and recycle plastic PET quickly and accurately. However, the current electrostatic sorting method, dissolution sorting method, density sorting method, manual sorting and identification, photoelectric sorting method, etc. used for plastic PET identification have problems such as environmental pollution, high investment in manpower and material resources, or low identification accuracy due to human errors. In view of the above technical problems existing in the prior art, a specific embodiment of a plastic PET identification method based on hyperspectral technology is provided below. Before the reference plate image is collected, a hyperspectral imager for collecting hyperspectral image data of a variety of mixed plastics is selected, and the spectral range of the hyperspectral imager is 900nm to 2350nm. In this embodiment, the selected hyperspectral imager includes: a short-wave infrared camera based on linear gradient filter imaging technology, model Lambda-NIR, and two short-wave infrared cameras based on transmission grating imaging technology, model GaiaField Pro-N17E and GaiaField Pro-N235E-XE, and the parameters of each camera are shown in Table 1:

[0040] Table 1 Lambda-NIR, GaiaField Pro-N17E and GaiaField Pro-N235E-XE parameters

[0041]

[0042] The three hyperspectral imagers (i.e. cameras), Lambda-NIR, GaiaField Pro-N17E and GaiaField Pro-N235E-XE, all acquire hyperspectral image data of different plastics through specview acquisition and analysis software.

[0043] See Figure 8 The specific steps to achieve plastic PET identification based on the above three hyperspectral imagers include:

[0044] S1. Automatically expose the above three cameras to a reference plate (in this embodiment, the reference plate is a white plate), collect the reference plate image as a calibration image (in this embodiment, the background is a black cloth), and obtain the pixel value information of the calibration image and the dark background pixel value information of each hyperspectral imager; then lay the mixed plastics to be detected flat on a black background, and use a hyperspectral imager to collect and obtain the hyperspectral original image data of a variety of mixed plastics; the various mixed plastics include plastic PET, plastic PP, plastic PE, plastic PVC, plastic ABS and heat-resistant resin.

[0045] S2, preprocessing the hyperspectral raw data, and mixing the hyperspectral reflectance image data of the plastic, specifically: using the following conversion formula to convert the digital quantization values ​​of the background image and the hyperspectral raw image data into spectral reflectance,

[0046]

[0047] Among them, R ref Represents the reflectivity value, R ref Less than or equal to 1, DN raw Represents the brightness value of the image pixel of the original data, DN white is the image pixel brightness value of the reference plate, DN dark represents the dark background noise of the hyperspectral imager, R whiteRepresents the reflectivity of the reference plate. DN is the abbreviation of Digital Number, which indicates the brightness value of the image pixel, records the grayscale value of the image, has no unit, and is an integer value. The value is related to the radiation resolution of the sensor (i.e., the hyperspectral imager), the emissivity of the ground object, the atmospheric transmittance and scattering rate, etc. R is the abbreviation of Reflectance, which represents the percentage of the radiation energy reflected by the object (the object in this application is a plastic mixture) to the total radiation energy. Its size range is always less than or equal to 1. The nature of the object can be judged by the spectral reflectivity.

[0048] S3. According to the difference in spectral reflectance of different plastics in the hyperspectral reflectance image, the spectral wavelength used to distinguish plastic PET from non-plastic PET is obtained. Specifically, the spectral reflectance curves of plastic PET, PP, PE, PVC, ABS, heat-resistant resin and background at wavelengths of 1100nm to 2000nm are drawn. Figure 1 Among them, the spectral reflectance values ​​of plastics ABS, PP, PE and PVC are displayed as the main vertical axis due to their high spectral reflectance, and the horizontal axis represents the high spectral wavelength. The spectral reflectance of plastic PET is low due to its strong transparency and strong light signal penetration. The heat-resistant resin and background are gray and black, and have strong absorption, so the spectral reflectance is also low. Therefore, the spectral reflectance values ​​of plastic PET, heat-resistant resin and background are displayed as the secondary vertical axis to observe the change trend of their spectra. Figure 1 It can be seen that near 1630nm, the spectral reflectance curve of plastic PET has an obvious "trough" position, while the reflectance curves of other plastics and background near the wavelength of 1630nm show a slow downward trend, without an obvious "trough" position. Near the wavelength of 1700nm, the spectral reflectance curve of plastic PET shows an upward trend, and its spectral reflectance value is higher than the reflectance value near the wavelength of 1630nm, while other plastics and background have a more obvious "trough" position at the wavelength of 1700nm, and the spectral reflectance value is lower than the spectral reflectance value near the wavelength of 1630nm. Therefore, the wavelength of 1630nm and the wavelength of 1700nm can be selected as the two-channel combination to distinguish plastic PET and non-PET.

[0049] S4. Based on the spectral reflectance of different plastics in the hyperspectral reflectance image, the normalized index and ratio index of different plastic pixels are calculated, wherein the normalized index is the difference between the spectral reflectance of the near-infrared region and the red region / the sum of the spectral reflectance of the near-infrared region and the red region. The calculation formula for calculating the normalized index (expressed as NI) of the present invention is as follows:

[0050] NI=(R ref1700 -R ref 1630 ) / (R ref 1700 +R ref 1630 );

[0051] The ratio index is the ratio of the spectral reflectance between the near-infrared region and the red region. The calculation formula of the ratio index (expressed by RI) is:

[0052] RI=R ref 1700 / R ref 1630 ;

[0053] Among them, R ref 1700 It represents the spectral reflectance of plastic PET pixel or non-plastic PET pixel at a wavelength of 1700nm, R ref 1630 Indicates the spectral reflectance of plastic PET pixel or non-plastic PET pixel at a wavelength of 1630nm.

[0054] S5, obtaining a normalized index threshold, a ratio index threshold, and a spectral reflectance threshold for distinguishing plastic PET from non-plastic PET; specifically, setting a NI threshold (i.e., a normalized index threshold) based on the difference between the NI value dynamic range of the plastic PET area and the NI value dynamic range of the non-plastic PET area in the hyperspectral reflectance image, Figure 2 From the frequency distribution diagram of the NI values ​​of plastic PET pixels and non-plastic PET pixels in the hyperspectral reflectance image, it can be seen that when the NI value is 0.05, the NI value frequency curve of the plastic PET pixel intersects with the NI value frequency curve of the non-plastic PET pixel. When the NI value is greater than 0.05, most of the pixels in the hyperspectral reflectance image are plastic PET pixels. Therefore, the normalized index threshold is determined to be 0.05.

[0055] The RI threshold is set based on the difference between the dynamic range of RI values ​​in the plastic PET area and the non-plastic PET area in the hyperspectral reflectance image. Figure 3 From the frequency distribution diagram of the RI values ​​of plastic PET and non-plastic PET pixels in the hyperspectral reflectance image, it can be seen that when the NI value is 1.2, the RI value frequency curve of the plastic PET pixel intersects with the RI value frequency curve of the non-plastic PET pixel. When the NI threshold (ratio index threshold) is set greater than 1.2, most of the pixels in the hyperspectral reflectance image are plastic PET pixels. Therefore, the ratio index threshold is determined to be 1.2.

[0056] According to the difference between the dynamic range of the spectral reflectance value at a wavelength of 1630nm in the area where the plastic PET is located and the dynamic range of the spectral reflectance value at a wavelength of 1630nm in the non-plastic PET area in the image, the spectral reflectance value at a wavelength of 1630nm is set as the spectral reflectance threshold. Figure 4The frequency distribution diagram of the spectral reflectance values ​​of plastic PET and non-plastic PET pixels at a wavelength of 1630nm in the hyperspectral reflectance image shows that when the spectral reflectance value at a wavelength of 1630nm is less than 0.1, most of the pixels in the hyperspectral reflectance image are plastic PET pixels. Figure 1 It can be seen that at a wavelength of 1700nm, the spectral reflectance values ​​of plastic PP and plastic PE are very close to the spectral reflectance value of plastic PET, so it is impossible to use only the spectral reflectance value at a wavelength of 1700nm to set the spectral reflectance threshold for distinguishing plastic PET from non-PET. Therefore, the spectral reflectance value at a wavelength of 1630nm is used as the spectral reflectance threshold, and the spectral reflectance threshold is determined to be 0.1.

[0057] S6. Extract the plastic PET region in the hyperspectral reflectance image based on the spectral reflectance threshold, normalized index threshold, and ratio index threshold (the region where the spectral reflectance threshold, normalized index threshold, and ratio index threshold intersect is the PET region), obtain the recognition effect diagram of the plastic PET, and realize the recognition of the plastic PET. The specific processing and analysis of the hyperspectral reflectance image in the above steps S2 to S6 are all realized by the processor, and the extracted plastic PET image can be displayed on the display, thereby realizing the visual recognition of the plastic PET.

[0058] Figure 5 , Figure 6 , Figure 7 The effect diagram of identifying plastic PET through short-wave infrared dual channels is given. Figure 5 Figure 5a is a mixed plastic hyperspectral reflectance image collected by a Lambda-NIR camera. Figure 5 Figure 5b is a rendering of the plastic PET extracted using the identification method of this application. Figure 6 Figure 6a is a mixed plastic hyperspectral reflectance image collected by the GaiaField Pro-N17E camera. Figure 6 Figure 6b is a rendering of the plastic PET extracted using the identification method of this application. Figure 7 Figure 7a is a mixed plastic hyperspectral reflectance image collected by the GaiaField Pro-N235E-XE camera. Figure 6 6b is a rendering of the plastic PET extracted by the identification method of this application, from Figure 5 5b, Figure 6 Middle 6b, Figure 7 As can be seen from Figure 7b, the combined effect of the spectral reflectance threshold, normalized index threshold, and ratio index threshold determined by the present application can quickly, accurately, and visually identify plastic PET.

[0059] The above are only preferred embodiments of the present application, and the present invention is not limited to the above embodiments. It is understood that other improvements and changes directly derived or associated by those skilled in the art without departing from the spirit and concept of the invention should be considered to be included in the scope of protection of the invention.

Claims

1. A plastic PET identification method based on hyperspectral technology, characterized in that: The method includes: S1. Spread the mixed plastics on a black background and use different hyperspectral imagers to collect the hyperspectral raw image data of the mixed plastics respectively; S2, preprocessing the hyperspectral raw image data to obtain hyperspectral reflectance image data; S3, obtaining a spectral wavelength for distinguishing plastic PET from non-plastic PET according to the difference in spectral reflectance of different plastics in the hyperspectral reflectance image data; S4. Calculate the normalized index and ratio index of different plastic pixels based on the spectral reflectance of different plastics in the hyperspectral reflectance image data; S5, obtaining a normalized index threshold, a ratio index threshold, and a spectral reflectance threshold for distinguishing plastic PET from non-plastic PET; S6, extracting the plastic PET area in the hyperspectral reflectance image data based on the spectral reflectance threshold, the normalized index threshold, and the ratio index threshold, obtaining a plastic PET recognition effect diagram, and realizing plastic PET recognition; Before the hyperspectral raw image data is collected, a hyperspectral imager for collecting the hyperspectral raw image data is selected, and the spectral range of the camera is 900nm~2350nm; the hyperspectral imager includes: at least one short-wave infrared camera using linear gradient filter imaging technology, and at least two short-wave infrared cameras using transmission grating imaging technology; In step S1, the mixed plastic includes plastic PET, plastic PP, plastic PE, plastic PVC, plastic ABS and heat-resistant resin; In step S2, the hyperspectral original image data is preprocessed, that is, the digital quantization value of the hyperspectral original image data is converted into spectral reflectance, and the conversion formula is: ， Among them, R ref Represents the reflectivity value, R ref Less than or equal to 1, DN raw Represents the brightness value of the image pixel of the original data, DN white is the image pixel brightness value of the reference plate, DN dark represents the camera dark background noise, R white Indicates the reflectivity of the reference plate, which is a white plate; In step S3, the spectral wavelength used to distinguish plastic PET from non-plastic PET is a dual-channel wavelength, including: 1630nm and 1700nm; In step S4, the normalized index NI is calculated by the formula for: NI=(R ref1700 - R ref 1630 ) / ( R ref 1700 + R ref 1630 ) ； The calculation formula of the ratio index RI is: :RI=R ref 1700 / R ref 1630 ; in, R ref 1700 Indicates the spectral reflectance of plastic PET or non-plastic PET at a wavelength of 1700nm. R ref 1630 Indicates the spectral reflectance of plastic PET or non-plastic PET at a wavelength of 1630nm; In step S5, the spectral reflectance value of 0.1 at a wavelength of 1630 nm is the spectral reflectance threshold for distinguishing the plastic PET from the non-plastic PET, the normalized index threshold is 0.05, and the ratio index threshold is 1.

2.

2. The method for identifying plastic PET based on hyperspectral technology according to claim 1, characterized in that: In step S5, the normalized index threshold is determined according to the frequency distribution difference of the normalized index NI of the plastic PET and the non-plastic PET in the hyperspectral reflectance image data; the ratio index threshold is determined according to the difference between the dynamic change range of the ratio index of the plastic PET and the dynamic change range of the ratio index of the non-plastic PET; the spectral reflectance threshold is obtained based on the frequency distribution difference of the spectral reflectance values ​​of the plastic PET and the non-plastic PET pixels at wavelengths of 1630nm and 1700nm in the hyperspectral reflectance image data.

Images