Detection method of broken plastics and microplastics based on RGB and hyperspectral image fusion

Abstract

The present invention relates to a method for detecting broken plastics and microplastics based on RGB and hyperspectral image fusion, comprising the following steps: obtaining broken plastics and microplastics; mixing them with solid waste to obtain a solid-phase matrix; pre-treating the obtained solid-phase matrix to obtain a material; drying the material to remove some moisture, coating it on a quartz window, drying it until the moisture is completely removed, and flattening it with another quartz window to obtain a test material; using a high-resolution color image scanner and a hyperspectral camera to obtain an RGB image and a hyperspectral image of the obtained test material, respectively; fusing the obtained RGB image and hyperspectral image; and automatically classifying and identifying broken plastics and microplastics using a supervised classification model. Compared with existing technologies, the method provided by the present invention does not require separation steps such as density flotation, which can reduce the loss of broken plastics and microplastics during pre-treatment; it shortens detection time and can achieve high-throughput analysis of large sample sizes; and it can effectively broaden the size range of broken plastics and microplastics and improve recognition accuracy.

Description

Technical Field

[0001] The present invention relates to the field of microplastic detection methodology, and in particular to a method for detecting broken plastics and microplastics based on RGB and hyperspectral image fusion. Background Art

[0002] Microplastics are plastic fragments smaller than 5mm in size. As a new pollutant, they are widely distributed in natural water systems such as lakes and oceans. The abundance of terrestrial microplastics is 4 to 23 times that of the ocean. They are the main source of microplastics in the aquatic environment, so it is crucial to block the release of terrestrial microplastics at the source. There is a large amount of plastic garbage in solid waste. The microplastics and plastic fragments larger than or equal to 5mm (hereinafter referred to as crushed plastics) generated after disposal are enriched in the products. With the subsequent resource utilization of the products, crushed plastics and microplastics can escape and accumulate in the soil environment, posing potential safety risks.

[0003] At present, there are two main methods for detecting broken plastics and microplastics in solid matrices: (1) Referring to the detection technology of microplastics in soil: digesting and flotating solid waste samples to separate microplastics, and then using vibration spectroscopy to identify the broken plastics and microplastics. (2) Referring to the detection method of impurity content in the quality standards of organic waste resource processing products in countries such as Germany, the United Kingdom and Austria, gradient screening solid waste and manually selecting broken plastics and microplastics. The above methods are time-consuming, have extremely low throughput, and high error rates. Due to the heterogeneity of solid waste, the detection of small sample amounts results in extremely high coefficients of variation. In view of the characteristics of solid waste with large differences in organic matter content, complex types of inorganic impurities, and a wide range of sizes of broken plastics and microplastics (from centimeters to micrometers), this patent will develop a high-throughput method for detecting broken plastics and microplastics.

[0004] Hyperspectral images have high spectral resolution and can effectively identify the chemical composition of the target object. However, due to the low spatial resolution of the hyperspectral camera, it is difficult to obtain the morphological and texture details of the target object, which makes it difficult for the hyperspectral image to identify small-sized or fibrous microplastics. RGB images, i.e., images of three primary colors, have only three channels but have high spatial resolution and can obtain clear morphological characteristics of microplastics. Patent CN 108489910 A discloses a method for rapid detection of microplastics in oysters based on hyperspectral technology, the steps of which are as follows: (1) using visible near-infrared spectroscopy to quickly distinguish microplastics from oyster tissues and other impurities in the body; (2) using spectral technology combined with supervised classification methods such as support vector machines to identify different types of microplastics in oysters; (3) combining hyperspectral imaging technology with spectral technology to visualize the spatial distribution of microplastics in oysters; however, this method only uses a hyperspectral camera to detect microplastics, and the classification accuracy of various microplastics is low and the identification size range is small.

[0005] In summary, there is an urgent need to provide a method for detecting broken plastics and microplastics to improve the classification accuracy of various microplastics and broaden the identification size range. Summary of the Invention

[0006] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a method for detecting broken plastics and microplastics based on the fusion of RGB and hyperspectral images. The image fusion technology is used to resample the hyperspectral image and RGB image so that the generated image has both high spatial resolution and multispectral characteristics. The fused results are used to train a supervised classification model to achieve high-throughput detection of broken plastics and microplastics in solid matrices, which can effectively improve the classification accuracy of various microplastics and broaden the recognition size range.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] The purpose of the present invention is to provide a method for detecting broken plastics and microplastics based on RGB and hyperspectral image fusion, comprising the following steps:

[0009] S1. The plastic raw material is frozen and crushed with liquid nitrogen, and then gradient sieved to cut into microplastics and crushed plastics with a size ranging from 100 μm to 50 mm. The mixed plastics and crushed plastics are then added to the solid waste to obtain a solid phase matrix;

[0010] S2. Pre-treating the solid matrix obtained in S1 according to the degree of agglomeration and gradation of the solid waste described in S1 to obtain a material;

[0011] S3, drying the material obtained in S2 to a moisture content of 60-75 wt%, coating it on a quartz window, and further drying it to completely remove the moisture to obtain a dry material. Using another quartz window, flatten the obtained dry material to obtain a material to be tested with a thickness of 0.5-1.5 mm;

[0012] S41, flipping the combination of the material to be tested obtained in S3 and the two quartz windows, using a high-resolution color image scanner to capture an RGB image of the material to be tested obtained in S3, selecting "Advanced Mode", setting the resolution and color bit, and adjusting the optical density value to obtain a high-resolution RGB image;

[0013] S42, using a hyperspectral camera to collect a hyperspectral image of the material to be tested obtained in S3, adjusting the object distance, exposure time, frame number range, and light source intensity to obtain the clearest hyperspectral image;

[0014] S5, adding layer coordinates to the RGB image obtained in S41 to obtain a reference RGB image, using it as a reference image for image registration, selecting 7-15 constraint points in the obtained reference RGB image and the hyperspectral image obtained in S42, respectively, generating a registered hyperspectral image using a cubic convolution interpolation method, and selecting a Gram-schmidt method to fuse the obtained reference RGB image and the obtained registered hyperspectral image to obtain a fused image;

[0015] S6. Select the regions of interest of the broken plastics and microplastics described in S1 and the solid phase matrix described in S1 on the fused image obtained in S5, obtain standardized spectral curves of various regions of interest, use supervised classification methods to identify broken plastics and microplastics, calculate the precision, recall rate and F1 score, and calculate the detection rate based on the dosage and test results.

[0016] Furthermore, the solid waste in S1 includes fine-graded low-agglomeration materials, coarse-graded high-agglomeration materials, and coarse-graded low-agglomeration materials.

[0017] Furthermore, the fine-graded low-agglomeration material includes dried sewage plant sludge or biogas residue.

[0018] Furthermore, the coarse-graded high-agglomeration material includes kitchen waste, other waste or compost.

[0019] Furthermore, the coarse-graded low-agglomeration material includes landfill, slag or soil contaminated by agricultural film.

[0020] Further preferably, the pretreatment in S2 comprises the following steps:

[0021] When the solid waste described in S1 is a fine-graded low-agglomeration material, a dispersion is added and then ultrasonically treated. When the solid waste described in S1 is a coarse-graded high-agglomeration material, a digester is added, and the reaction is carried out at 45°C for 12 hours, and then a dispersion is added, and then ultrasonically treated. When the solid waste described in S1 is a coarse-graded low-agglomeration material, after passing through a sieve with a pore size of 2 mm, the undersize material and the dispersion are mixed and then ultrasonically treated. After the oversize material is sorted out of centimeter-level broken plastics, the remaining oversize material is mixed with a digester, and the reaction is carried out at 45°C for 12 hours, and then a dispersion is added, and then ultrasonically treated.

[0022] Further preferably, the digester is a mixed solution of hydrogen peroxide and ammonia water, wherein the concentration of hydrogen peroxide is 125 g / L and the concentration of ammonia water is 105 g / L.

[0023] Further preferably, the dispersion liquid is an ethanol solution with an ethanol volume fraction of 65-75%.

[0024] Furthermore, the plastic raw material in S1 includes one or more of polyethylene, polypropylene, polyvinyl chloride, polystyrene, polyethylene terephthalate, polyamide, and polybutylene adipate / terephthalate resin.

[0025] Furthermore, the shapes of the crushed plastics and microplastics described in S1 include one or more of block, film, fiber and sphere.

[0026] Furthermore, the resolution in S41 is 1200 dpi, the color bit is 24 bits, and the optical density value is 3.8.

[0027] The reason for flipping the combination of the test material obtained in S3 and the two quartz windows in S41 is that the high-resolution color image scanner places the quartz window on the scanning plate and shoots (scans) from bottom to top, and the hyperspectral camera shoots from top to bottom on the object placement platform. In order to ensure subsequent alignment, the quartz window needs to be flipped.

[0028] Furthermore, the object distance in S42 is 15-20 cm, the exposure time is 2.2-3.8 ms, the frame rate range is 40-70 Hz, and the light source intensity is 2000-2500 lux.

[0029] Furthermore, the supervised classification method described in S6 includes support vector machine and random forest.

[0030] Furthermore, the crushed plastic is plastic fragments with a size greater than or equal to 5 mm; the microplastic is plastic fragments with a size less than 5 mm.

[0031] Furthermore, the RGB image in step S41 is not limited to RGB images, but also includes full-color spectrum images, grayscale images taken by industrial CCD (charge coupled device) cameras, and other high-resolution images containing target texture details.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1) The method provided by the present invention does not require separation steps such as density flotation, which can reduce the loss of microplastics during the pretreatment process.

[0034] 2) The method provided by the present invention can detect broken plastics and microplastics in solid matrices in situ with low damage.

[0035] 3) The method provided by the present invention is time-saving and can simultaneously obtain the morphological and material information of broken plastics and microplastics, enabling high-throughput analysis of large sample volumes.

[0036] 4) The method provided by the present invention has high classification accuracy and can identify a wider range of sizes than the image classification method obtained by a single detector. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of the method for detecting broken plastics and microplastics based on RGB and hyperspectral image fusion provided by the present invention.

[0038] Figure 2 This is the classification result diagram of broken plastics and microplastics in Example 1. DETAILED DESCRIPTION

[0039] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments will help those skilled in the art further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that those skilled in the art may make several variations and improvements without departing from the scope of the present invention. These all fall within the scope of protection of the present invention.

[0040] Any features such as preparation methods, materials, structures or composition ratios that are not clearly described in this technical solution shall be deemed to be common technical features disclosed in the prior art.

[0041] Example 1

[0042] like Figure 1 As shown, the method for detecting broken plastics and microplastics based on RGB and hyperspectral image fusion in this embodiment includes the following steps:

[0043] S1. Polypropylene, polyethylene, polyethylene terephthalate, polyvinyl chloride, polystyrene, and polybutylene adipate / terephthalate (all purchased from Meryer) were used as plastic raw materials. The plastic raw materials were frozen and crushed with liquid nitrogen and then gradient sieved to obtain microplastics and crushed plastics with sizes of 100-250 μm, 250-500 μm, 250-500 μm, 500-1000 μm, 1-2 mm, 2-5 mm, and 5-50 mm. Three crushed plastics and microplastics were selected from each size class and mixed and added to the biogas residue to obtain a solid phase matrix;

[0044] S2. Add 65% (in terms of ethanol volume fraction) ethanol solution of the dispersion to the solid matrix obtained in S1, and then ultrasonicate at 340W for 10 minutes to obtain a material;

[0045] S3. Pour the material obtained in S2 into a Petri dish and dry it at 45°C to a moisture content of 60 wt%. Then, apply it to a quartz window and further dry it until the moisture is completely removed to obtain a dry material. Use another quartz window to flatten the obtained dry material to obtain a material to be tested with a thickness of 1 mm.

[0046] S41. Turn over the quartz window carrying the material to be tested obtained in S3, and use a high-resolution color image scanner to capture an RGB image of the material to be tested obtained in S3. Select "Advanced Mode", set the resolution to 1200 dpi, the color bit to 24 bit, and adjust the optical density value to 3.8 to obtain a high-resolution RGB image.

[0047] S42. Use a hyperspectral camera to collect a hyperspectral image of the material to be tested obtained in S3. After white calibration using a white acrylic rod, the object distance is 18 cm, the exposure time is 2.8 ms, the frame rate range is 55 Hz, and the light source intensity is 2200 lux to obtain the clearest hyperspectral image.

[0048] S5, adding layer coordinates to the RGB image obtained in S41 to obtain a reference RGB image, using it as a reference image for image registration, selecting 12 constraint points in the obtained reference RGB image and the hyperspectral image obtained in S42, respectively, generating a registered hyperspectral image using a cubic convolution interpolation method, and selecting a Gram-schmidt method to fuse the obtained reference RGB image and the obtained registered hyperspectral image to obtain a fused image;

[0049] S6. Select regions of interest (ROIs) of the plastic fragments and microplastics described in S1 and the solid matrix described in S1 from the fused image obtained in S5, obtain standardized spectral curves for each region of interest, and use a supervised classification method using a support vector machine to identify the plastic fragments and microplastics with an accuracy exceeding 90%, a recall rate exceeding 85%, and an F1 score exceeding 85%. The detection rate is the ratio of the number of plastic fragments and microplastics identified to the number of plastic fragments and microplastics added in S1, with a detection rate exceeding 90% for all types of plastic fragments and microplastics.

[0050] The classification results are as follows Figure 2 As shown, the black part is the dispersed sludge, and the white box is the identified microplastics and broken plastics.

[0051] The feasibility of this method was verified using traditional detection methods. Microplastics were extracted from biogas residues through digestion and flotation, then transferred to a filter membrane. Suspected microplastic particles were detected using a microscope in infrared transmission mode. The ratio of the number of identified broken plastics and microplastics to the number of added broken plastics and microplastics was calculated. The detection rate for all types of broken plastics and microplastics was between 85% and 95%.

[0052] The above description of the embodiments is intended to facilitate understanding and use of the invention by those skilled in the art. It will be apparent that those skilled in the art can readily make various modifications to these embodiments and apply the general principles described herein to other embodiments without requiring inventive effort. Therefore, the present invention is not limited to the above-described embodiments. Improvements and modifications made by those skilled in the art based on the disclosure of the present invention, without departing from the scope of the present invention, should be within the scope of protection of the present invention.

Claims

1. A method for detecting broken plastics and microplastics based on RGB and hyperspectral image fusion, characterized in that: The method comprises the following steps: S1. The plastic raw materials are frozen and crushed with liquid nitrogen, and then gradient sieved to cut into microplastics and plastic fragments with a size ranging from 100 μm to 50 mm. The mixed plastics and crushed plastics are then added to the solid waste to obtain a solid phase matrix; The solid wastes described in S1 include fine-graded low-agglomeration materials, coarse-graded high-agglomeration materials, and coarse-graded low-agglomeration materials; The fine-graded low-agglomeration material includes dried sewage plant sludge or biogas residue; The coarse-graded high-agglomeration material includes kitchen waste, other waste or compost; The coarse-graded low-agglomeration materials include landfills, slag or soil contaminated by agricultural films; S2. Pre-treating the solid matrix obtained in S1 according to the degree of agglomeration and gradation of the solid waste described in S1 to obtain a material; The pre-processing in S2 comprises the following steps: When the solid waste in S1 is a finely graded and low-agglomeration material, a dispersion liquid is added, followed by ultrasonic treatment; When the solid waste in S1 is a coarse-graded and highly agglomerated material, a digester is added, and after reacting at 45°C for 12 hours, a dispersion is added, and then ultrasonic treatment is performed; When the solid waste described in S1 is a coarse-graded low-agglomeration material, after passing through a sieve with a pore size of 2 mm, the undersize material is mixed with the dispersion and then ultrasonically treated. After the oversize material is sorted out to remove centimeter-level plastic fragments, the remaining oversize material is mixed with a digester, reacted at 45°C for 12 hours, and then the dispersion is added and ultrasonically treated. S3, drying the material obtained in S2 to a moisture content of 60-75 wt% and then coating it on a quartz window, further drying it to completely remove the moisture to obtain a dry material, and using another quartz window to flatten the obtained dry material to obtain a material to be tested with a thickness of 0.5-1.5 mm; S41. Flip the combination of the material to be tested obtained in S3 and the two quartz windows, use a high-resolution color image scanner to capture an RGB image of the material to be tested obtained in S3, select "Advanced Mode", set the resolution and color bit, and adjust the optical density value to obtain a high-resolution RGB image; S42, using a hyperspectral camera to collect a hyperspectral image of the material to be tested obtained in S3, adjusting the object distance, exposure time, frame number range, and light source intensity to obtain the clearest hyperspectral image; In S42, the object distance is 15-20 cm, the exposure time is 2.2-3.8 ms, the frame rate range is 40-70 Hz, and the light source intensity is 2000-2500 lux; S5, adding layer coordinates to the RGB image obtained in S41 to obtain a reference RGB image, using it as a reference image for image registration, selecting 7-15 constraint points in the obtained reference RGB image and the hyperspectral image obtained in S42, respectively, generating a registered hyperspectral image using a cubic convolution interpolation method, and selecting a Gram-schmidt method to fuse the obtained reference RGB image and the obtained registered hyperspectral image to obtain a fused image; S6. Selecting regions of interest of the broken plastics and microplastics described in S1 and the solid matrix described in S1 on the fused image obtained in S5, obtaining standardized spectral curves of various regions of interest, using a supervised classification method to identify the broken plastics and microplastics, calculating the precision, recall rate, and F1 score, and calculating the detection rate based on the dosage and the test results; The digester is a mixed solution of hydrogen peroxide and ammonia water, wherein the concentration of hydrogen peroxide is 125 g / L and the concentration of ammonia water is 105 g / L; The dispersion liquid is an ethanol solution with an ethanol volume fraction of 65-75%; The resolution in S41 is 1200 dpi, the color bit is 24 bits, and the optical density value is 3.8; The supervised classification methods described in S6 include support vector machines and random forests.

2. The method for detecting broken plastics and microplastics based on RGB and hyperspectral image fusion according to claim 1, characterized in that: The plastic raw materials described in S1 include one or more of polyethylene, polypropylene, polyvinyl chloride, polystyrene, polyethylene terephthalate, polyamide, and polybutylene adipate / terephthalate resin.

3. The method for detecting broken plastics and microplastics based on RGB and hyperspectral image fusion according to claim 1, characterized in that: The shapes of the crushed plastics and microplastics described in S1 include one or more of block, film, fiber and sphere.

4. The method for detecting broken plastics and microplastics based on RGB and hyperspectral image fusion according to claim 1, characterized in that: The crushed plastics are plastic fragments with a size greater than or equal to 5 mm; The microplastics are plastic fragments smaller than 5 mm in size.

Images