CN109766909B - Analysis method for aging behavior of shore environment microplastic based on spectrogram fusion - Google Patents

Abstract

The present invention provides a method for analyzing the aging behavior of microplastics in coastal environments based on spectrum fusion, which includes the following steps: 1. Collect microplastic samples in coastal environments and separate and prepare them. 2. Use an infrared spectrometer to obtain micro-area morphological images and infrared spectrum information of microplastic samples. 3. Extract the relevant characteristic spectral data matrix from the infrared spectrum information. 4. Extract texture features from micro-area image information. 5. Obtain the surface yellowness characteristics from the micro-area image information. 6. Establish a feature fusion prediction model based on sample surface morphology-molecular spectrum. 7. Calibrate the prediction model to obtain the correction model. 8. Extract the feature fusion matrix of the sample to be tested, and input the feature fusion matrix of the sample to be tested into the correction model to obtain the corresponding aging analysis results. This solution achieves rapid analysis of the aging degree of microplastics in complex environments by introducing parameters such as spectral carbonyl ratio and yellowness of texture and shape.

Description

Analysis method of aging behavior of microplastics in coastal environment based on spectral fusion

Technical field

The invention relates to a method for analyzing microplastic pollution in complex coastal environments, in particular a method for analyzing the aging behavior of microplastics in coastal environments based on spectral fusion.

Background technique

In recent years, microplastics have become a new type of pollutant that has attracted much attention in the global marine and coastal environment. The aging behavior of microplastics is an important parameter for the traceability analysis of microplastic pollution sources and the dynamic growth and decline model of microplastics in the environment. During the aging process, the surface of microplastic samples shows obvious yellowing, surface cracks and powdering, and the absorption intensity of functional groups related to aging behavior will show a certain coupling relationship as the degree of aging increases. However, due to weathering, the surface of microplastics separated from environmental samples is uneven or a large number of environmental impurities are attached, which will result in the inability to obtain high signal-to-noise ratio infrared spectra during the measurement process. At the same time, the surface color of the sample during the aging process is also related to the concentration of residual antioxidants inside the sample, and the surface crack shape is related to the volatilization degree of the plasticizer in the sample.

Since the spectrum and surface morphology of aged samples are affected by the surface roughness of the sample and residual additives inside the sample, it is difficult to accurately analyze the aging behavior of microplastic samples under complex environmental backgrounds if a single infrared spectral feature or single morphological feature modeling is used.

Contents of the invention

In order to overcome the above shortcomings and shortcomings of the existing technology, a method for accurately analyzing the aging behavior of microplastic samples in coastal environments is provided, which uses a method based on visible-infrared spectrum fusion to achieve the aging behavior of environmental samples in typical coastal environments. Non-destructive analysis, convenient method and high reliability of detection results.

A method for analyzing the aging behavior of microplastics in coastal environments based on spectral fusion includes the following steps:

a. Coastal environment collects microplastic samples with different degrees of aging and a particle size of more than 1 mm and separates and prepares them.

b. Use a Fourier transform infrared instrument to collect micro-area image information and infrared spectrum information of the microplastic samples. The microplastic samples are randomly divided into correction samples and prediction samples.

c. Use standard normal variables to correct the infrared spectral information, then extract the optimal characteristic wavelengths related to aging behavior and establish a characteristic spectral data matrix K.

d. Use the commonly used method of describing texture on the obtained micro-area image information: extract texture features from the gray-level co-occurrence matrix, and calculate the energy C1 and contrast C2 of the gray-level spatial correlation characteristic co-occurrence matrix respectively as texture quantification indicators that characterize texture features.

Among them, i and j are the pixel gray coordinates respectively, d is the distance between pixels, θ is the error, and P(i,j,d,θ) is the probability.

e. Use the HSB system, that is, the color model, to obtain the hue, brightness, and saturation parameters of the acquired micro-area image information to characterize the yellowness characteristics. In HSB mode, H (hues) represents hue, S (saturation) represents saturation, and B (brightness) represents brightness.

f. Use steps c, d and e to obtain the characteristic spectral data matrix K, texture feature C, and yellowness feature matrix P respectively for feature layer fusion to obtain the fusion matrix M, M=[K C P].

g. Establish a feature fusion prediction model based on surface morphology-molecular spectrum by introducing parameters such as carbonyl ratio, texture morphology and yellowness:

The aging characteristic value of _the spectral image of a certain sample is expressed _as X ₁ , X ₂ to Xn; assuming that the possible feature matrix after fusion is M, consisting of X 1 , Weight can be regarded as a measure of the contribution rate of different feature vectors.

h. Use the correction samples to correct the prediction model described in step f to obtain the correction model.

i. Use a Fourier transform infrared instrument to collect the image spectrum information of the microplastic to be measured, use the fusion matrix obtained in step f to input it into the fusion prediction model obtained in step g, and obtain the corresponding aging analysis results.

In order to improve the above solution, the present invention is further configured as follows: in step b, 1/3 of the microplastic samples are randomly selected as prediction samples, and 2/3 of the microplastic samples are used as correction samples.

In order to improve the above solution, the present invention is further configured as follows: the analysis method of the micro-area morphological image and infrared spectrum information of the microplastic sample in step c is: using wavelet denoising processing, multivariate scattering correction, and standard normal variable correction methods. Preprocess raw spectral data to obtain high-precision raw data.

In order to improve the above solution, the present invention is further configured to: in step c, screen the aging behavior characteristic wavelengths of wave numbers near 1450 cm ^-1 and 1750 cm ^-1 for the transparent polyethylene sample.

In order to improve the above solution, the present invention is further set as follows: the correction model accuracy root mean square error RMSEP in step g is <0.15, and the correction coefficient of determination R ² >0.92.

This invention builds a feature fusion model based on surface morphology-spectral characteristics by introducing parameters such as carbonyl ratio, texture morphology yellowness, etc., to achieve rapid analysis of the aging degree of microplastics in complex environments. It solves the problem that it is difficult to analyze the aging behavior of samples in actual complex systems using a single aging feature. The microscopic infrared technology used non-destructively detects the spectral image information of trace samples, ensuring the non-destructive repeated testing requirements of sample aging stages and improving the utilization of sample information. This patented method can achieve non-destructive, rapid and accurate analysis of microplastic samples collected from complex coastal environments, provide scientific support for the supervision of microplastic pollution in my country's coastal environment, and is of great significance to the management of marine microplastic pollution.

The present invention will be described in further detail below with reference to the accompanying drawings.

Description of the drawings

Figure 1 is a schematic flow chart of the detection method of the present invention;

Figure 2 is an infrared spectrum carbonyl index diagram of an aged sample according to an embodiment of the present invention;

Figure 3 is a chromaticity characteristic diagram of the micro-area surface of the aged sample according to the embodiment of the present invention.

Detailed ways

Below, the present invention is described in detail through exemplary embodiments. It is to be understood, however, that features of one embodiment may be beneficially combined in other embodiments without further recitation.

A method to analyze the aging behavior of microplastics in coastal environments based on spectral fusion, using transparent polyethylene, which is the most widely distributed in the environment, as a representative sample to study the aging behavior of microplastics in complex coastal environments. Includes the following steps:

a. Collect microplastic samples from the surface environment of Longwan tidal flats in Zhejiang Province. Collect sediments about 5cm thick along the latest high tide line. Screen them through a steel sieve to collect a representative and sufficient number of microplastic samples. Put them into sealed bags and transport them back. laboratory. Microplastic samples were rinsed with deionized water. After filtering and screening with glass microfiber filter paper, it is placed in a metal tray and dried in an oven at 60 degrees Celsius. Finally, it is put into a sealed bag and placed in a clean and light-proof place to obtain transparent polyethylene particles.

b. Divide the transparent polyethylene samples into two categories. From the total sample, 1/3 samples are randomly selected as prediction samples, and 2/3 samples are used as correction samples. Specifically: 30 samples are randomly selected, a total of 10 are used as prediction samples, and the remaining 20 are used as correction samples.

The micro-infrared spectroscopy instrument used in the experiment is a HYPERION Fourier transform infrared instrument, equipped with an infrared detector and an infrared band 20X lens to collect information from different microplastic samples. The number of scans was 20, the recorded wave number range was 4000cm ^-1 –600cm ^-1 , and the spectral resolution was 4cm ^-1 . A total of infrared spectra and micro-area image information of 30 samples were obtained.

c. Use standard normal variable correction to preprocess the calibration sample spectral data to obtain high-precision spectral data.

For the obtained infrared spectrum of the polyethylene sample, the optimal characteristic wavelength related to the aging behavior was extracted and a spectral data matrix was established. The specific method is: based on the full-band spectral information, screen the aging behavior characteristic wavelengths of wave numbers near 1450cm ^-1 and 1750cm ^-1 . The absorption intensity of the characteristic carbonyl functional groups in the above vicinity shows a linear coupling relationship with the degree of aging.

d. For the obtained micro-area image information, use the gray level co-occurrence matrix to extract texture morphological features, and calculate the energy C1, contrast C2, and correlation of the corresponding co-occurrence matrix as texture quantification indicators to characterize the texture features, where i and j are respectively is the pixel gray coordinate, d is the distance between pixels, θ is the error, and P(i,j,d,θ) is the probability:

e. The obtained micro-area image information uses the HSB system to obtain color features such as H (hues) hue, S (saturation), and B (brightness) brightness to represent the yellowness characteristics.

v=max.

Since the correlation between saturation and brightness is significant, only two characteristic parameters, hue H and brightness B, are considered when extracting color characteristics. The data shows that the yellow sample surface has a hue of 55-75 and a brightness greater than 40. The hue of the brown-yellow sample is similar to yellow, but the brightness is only 10-20. The black sample surface extracts the highest hue, but the lowest brightness value.

f. Use steps c, d and e to obtain the characteristic spectral data matrix K, the texture feature C, and the yellowness feature matrix P respectively to perform feature layer fusion to obtain the fusion matrix M, M=[K, C, P].

g. Establish a feature fusion prediction model based on surface morphology-molecular spectrum by introducing parameters such as carbonyl ratio, texture morphology and yellowness:

The aging feature values of the spectral image of a certain sample are expressed as X1, X2 to Xn; assuming that the possible feature values after fusion are M, consisting of X1, X2... A measure of vector accuracy.

Among the accuracy of the prediction model established in this embodiment, the prediction root mean square error RMSEP=0.122.

h. Use the correction samples to correct the feature fusion prediction model to obtain the correction model.

i. Use a Fourier transform infrared instrument to collect the image spectrum information of the microplastic to be measured, use the fusion matrix obtained in step f to input it into the fusion prediction model obtained in step g, and obtain the corresponding aging analysis results.

This specific embodiment is only an explanation of the present invention, and it is not a limitation of the present invention. Those skilled in the art can make modifications to this embodiment without creative contribution as needed after reading this description, but as long as the rights of the present invention are All requirements are protected by patent law.

Claims (1)

1. A method for analyzing the aging behavior of microplastics in coastal environments based on spectral fusion, which is characterized by including the following steps: a. Collect microplastic samples with different degrees of aging and particle sizes above 1 mm from the coastal environment and separate and prepare them; b. Use a Fourier transform infrared instrument to collect micro-area image information and infrared spectrum information of the microplastic sample, and the microplastic sample is randomly divided into correction samples and prediction samples; c. Use the standard variable method to correct the obtained infrared spectral information, then extract the optimal characteristic wavelength related to the aging behavior of the sample and establish the characteristic spectral data matrix K, and then obtain the carbonyl ratio through the characteristic spectral data matrix K; d. Extract texture features from the obtained micro-area image information by using the gray-level co-occurrence matrix, a common method for describing texture, and calculate the energy C1 and contrast C2 of the gray-level spatial co-occurrence matrix respectively as texture quantification indicators that characterize texture features, Among them, i and j are the pixel gray coordinates respectively, d is the distance between pixels, θ is the error, and P(i,j,d,θ) is the probability; e. Use RGB to convert the obtained micro-area image information into the HSB system to obtain the hue, brightness and saturation parameters and extract the yellowness feature matrix P. In the HSB system, H represents hue, S represents saturation, and B represents brightness. ; f. Use steps c, d and e to obtain the characteristic spectral data matrix K, the texture feature matrix C, and the yellowness feature matrix P respectively, and fuse the three matrices at the feature layer to obtain the fusion matrix: M=[K C P]; g. Establish a multi-source feature fusion prediction model based on surface morphology-molecular spectrum by introducing carbonyl ratio, texture features and yellowness features: The aging characteristic values of the spectral image of a certain sample are expressed as X ₁ , X ₂ to Xn; assuming that the possible feature matrix after fusion _is M, consisting of X ₁ , β _i is a measure of the contribution rate of different feature vectors; h. Use the correction samples to correct the prediction model obtained in step g to obtain the correction model; i. Use a Fourier transform infrared instrument to collect the image spectrum information of the microplastic to be measured, input the fusion matrix obtained in step f into the fusion prediction model obtained in step g, and obtain the corresponding aging analysis results.