Research On Rapid Detection Technology Of Marine Microplastics

A large amount of plastic waste is gradually formed into microplastics by the process of physicochemical and biological effects,these bio-toxic microplastics gradually accumulate in the ocean,meanwhile absorb various toxic substances,and would be enriched with the food chain,causing great damage to marine life and ecological environment,posing a serious threat to the global marine environment.In recent years,microplastics pollution has received increasing attention.Detection technology is a basic tool for studying microplastics,and designing an accurate and efficient microplastics detection technology is of great significance.Traditional microplastics detection techniques have certain limitations.The visual method for physical characterization detection of microplastics,which requires experienced researchers to operate.Statistical results could easily be influenced by subjective factors.Thermal analysis techniques could destroy the test samples.Fourier infrared spectroscopy has certain requirements for the water content and size of samples.With its development flexibility and high-cost efficiency,machine vision is developing rapidly,and the use of computers instead of manual detection to identify the physical characterization of microplastics could significantly improve the efficiency of detection and ensure the objectivity of the test results.Raman spectroscopy is a nondestructive detection method,which is suitable for small size and low containing water volume.Different components of microplastics have corresponding characteristic peaks of Raman spectra,and from the peak positions people can infer the specific type of sample under test.In this thesis,a rapid detection method for marine microplastics was proposed,which can realize the physical characterization and the species information of microplastics on-site integrated detection.The system includes a computerized identification method for the physical characterization of microplastics by microscopic imaging,combining laser Raman detection technology with neural network algorithms for microplastic species identification,and completed the development of the detection system and conducted field tests.The main research elements of this thesis are as follows:(1)Designed a marine microplastic detection platform,the main body adopts a cage structure to enhance the stability of the optical path,part of the optical path is shared between the visible band and the Raman laser infrared band to simply the architecture.the microscopic imaging part adopts a transmissive visible light source,the 150 W light source is coupled by optical fiber out,and a dichroic mirror is used to reduce the loss and ensure the normal operation of the camera in the visible band and the Raman laser in the infrared band.(2)Microplastic physical characterization and species identification method using computer image processing techniques to analyze the physical characterization of microplastic samples.After pre-processing and threshold segmentation,edge profile information of microscopic images was obtained and further obtained microplastic morphological information discriminated by aspect ratio and other information to improve the detection efficiency and ensure the objectivity of the detection results.The microplastic Raman spectroscopy dataset was produced,and the neural network model was designed and optimized for the datasets with different processing methods.The ten times accuracy of the optimized neural network model for the original Raman spectroscopy dataset reached 97.13%,and the highest accuracy was96.45%.The impact of different sizes of 2D Raman spectral datasets on the training results was compared,and a 2D spectral image of 256×256 size was selected,and the ten times accuracy of the neural network model trained based on the Raman spectral image dataset reached 97.52%,and the highest accuracy also reached 97.72%.Compared with several classical machine learning and deep learning models,the neural network model proposed in this thesis requires shorter training time and has higher accuracy.(3)Developed a system control upper computer software,which can realize rapid detection of microplastics.The system was tested in Yantai coastal zone,and 40 samples of 6 kinds of microplastics were successfully detected,the most types were polyethylene,and the number of fiber/linear and fragment shapes were similar,the largest was 3.3 mm long and the smallest was18 μm,the maximum detection time for a single microplastic sample is 8 minutes,and the microplastic identification accuracy reached 93.94%,which proved that the microplastic detection system designed in this paper can effectively identify microplastic samples and provide a reference for the field of rapid identification of marine microplastics.