Search

CN-122016755-A - High-flux rapid detection method for multidimensional features of microplastic

CN122016755ACN 122016755 ACN122016755 ACN 122016755ACN-122016755-A

Abstract

The invention discloses a high-throughput rapid detection method for micro-plastic multi-dimensional characteristics, which not only realizes full-flow automatic processing, multi-mode data fusion synchronous characterization and standardized data integration and report generation by coupling micro-imaging, raman spectrum, deep learning and machine learning to perform micro-plastic multi-dimensional characteristic synchronous detection, but also can automatically output a structured report containing morphological characteristic parameters, supports cross-research data comparison and promotes detection standardization. The method has the advantages of remarkably improving the detection efficiency, shortening the single-sample detection time to 15min, comprehensively ensuring the detection precision, ensuring the material identification accuracy to be more than or equal to 95%, ensuring the shape classification accuracy to be more than or equal to 92%, ensuring the volume prediction average error to be less than 40%, and ensuring the multi-dimensional synchronous analysis capability, namely synchronously outputting a plurality of core shape features by a single process for the first time, and avoiding sample loss and data deviation caused by step detection. Provides a reliable tool for the research of the micro-plastic pollution mechanism and the environmental monitoring, thereby having wide application prospect.

Inventors

  • ZHANG YAN
  • ZHANG ZHANAO

Assignees

  • 南京大学

Dates

Publication Date
20260512
Application Date
20260206

Claims (7)

  1. 1. The high-throughput rapid detection method for the multidimensional features of the microplastic is characterized by comprising the following steps of: S1, acquiring a microscopic image and corresponding coordinates of a micro-plastic sample, acquiring Raman spectrum data of the coordinate point, and determining the material of the micro-plastic through spectrum matching based on the acquired Raman spectrum data; s2, carrying out automatic image segmentation processing on the microscopic image obtained in the step S1, identifying particles of the microplastic and extracting morphological characteristic parameters, wherein the method comprises the following steps: Equivalent circle diameter is taken as particle diameter; Aspect ratio, namely, the minimum circumscribed rectangle long axis/short axis is taken as a shape parameter; the circularity is 4 pi A/perimeter 2 , which is taken as a shape factor; Particle area, namely the number of contour pixel points is multiplied by the calibration coefficient; The number of particles; then automatically classifying according to the length-width ratio and the circularity threshold value, and determining the shape of the micro plastic particles; S3, constructing a volume prediction model based on PyCaret, namely building a machine learning model based on morphological characteristic parameters of each particle, wherein the model takes a shape factor, a particle size and a particle area as input characteristics, takes an actual measurement volume as a training target, trains the machine learning model and outputs a predicted micro-plastic volume; S4, integrating and outputting the multidimensional characteristics of the micro plastic, wherein the multidimensional characteristics comprise micro plastic particles ID, particle size, shape category, area A, aspect ratio AR, solidity S, circularity C, predicted volume V and material.
  2. 2. The method for high-throughput rapid detection of multi-dimensional features of microplastic according to claim 1, wherein the step S1 specifically comprises the steps of: S11, acquiring a microscopic image of a sample by adopting a low-power microscope, and identifying micro plastic particles through gray threshold segmentation to obtain coordinate positions of the micro plastic particles; s12, automatically positioning a Raman spectrometer to a coordinate position, and collecting Raman spectrum data; And S13, performing spectrum matching based on the characteristic peak position and the intensity similarity threshold value on the obtained Raman spectrum data and a preset spectrum library, and determining the material of the micro plastic.
  3. 3. The method for rapidly detecting the high-throughput characteristics of the micro plastic according to claim 1, wherein the step S2 specifically comprises the steps of inputting the micro image obtained in the step S1 into a pre-trained DeepLabV neural network, carrying out accurate contour segmentation on micro plastic particles in the micro image, outputting a binary mask, filtering impurity interference through morphological operation, traversing the mask contour by using OpenCV, identifying the particles of the micro plastic, calculating and extracting morphological characteristic parameters of each particle, wherein the morphological characteristic parameters comprise equivalent circle diameter of each particle as particle diameter, length-width ratio AR as shape parameter, circularity C as shape factor and number of particles, and automatically classifying according to length-width ratio AR and circularity C threshold value to determine the shape of the micro plastic particles.
  4. 4. The method for high-throughput rapid detection of multi-dimensional features of microplastic of claim 1, wherein in step S3, the machine learning model is a volume prediction model constructed by a random forest regression algorithm, and the training sample set comprises measured volume data of microplastic of different materials and corresponding shape factors, shape parameters and particle areas.
  5. 5. The method for high-throughput rapid detection of microplastic multi-dimensional features according to claim 4, wherein the machine learning model is trained by collecting microplastic samples of known volumes, acquiring microscopic images and raman data thereof, extracting shape factors, particle sizes and particle area features, taking the measured volumes as labels, traversing machine learning regression models in a Pycaret library, and screening optimal models for volume prediction of microplastic particles.
  6. 6. The method for rapidly detecting the high-throughput characteristics of the micro plastic according to claim 3, wherein the principle of determining the shape of the micro plastic particles according to the automatic classification of the aspect ratio AR and the circularity C threshold value is that the shape of the micro plastic particles is fibrous when AR >5, the shape of the micro plastic particles is fragmented when C <0.7 and AR is less than or equal to 5, and the shape of the micro plastic particles is spherical when C <0.7 and AR is less than or equal to 5.
  7. 7. The method for rapidly detecting the high-throughput of the multi-dimensional characteristics of the micro-plastic according to claim 1, wherein the integrated output of the multi-dimensional characteristics of the micro-plastic further comprises the distribution of abundance, material ratio, volume and shape ratio of the micro-plastic.

Description

High-flux rapid detection method for multidimensional features of microplastic Technical Field The invention belongs to the technical field of micro-plastic detection, and particularly relates to a high-throughput rapid detection method for multi-dimensional characteristics of micro-plastic. Background Microplastic (plastic chips with particle size <5 mm) has been widely detected in global aquatic, terrestrial and biological systems as an emerging environmental pollutant, and its ecological risks and health threats are in need of accurate assessment. Traditional microplastic detection methods are highly dependent on manual intervention, and labor-intensive operations result in efficiency and accuracy deficiencies. For example, in the morphological recognition link, the manual microscope is used for observing the particles which are easy to misjudge clay, organic chips and the like into micro plastics, the misjudgment rate is as high as more than 30 percent and the particles are only suitable for particles with the particle size of more than 1 mm, the raman spectrum or FTIR analysis needs to scan the particles point by point, the detection time of a single filter membrane is more than 2 hours, the high-flux analysis requirement of an environmental sample is difficult to meet, and the analysis efficiency and the data reliability are severely restricted. The multi-dimensional characteristic acquisition depends on a plurality of instrument combinations, and synchronous characterization is difficult. The key characteristics of the micro plastic include morphology (particle size, shape and quantity), chemical composition (material) and physical property (volume), and the prior art needs to detect step by step. In the aspect of morphology and component detection, the mainstream method is to acquire morphology parameters through a microscope, then transfer a sample to a spectrometer for component analysis, and the flow is complicated and particle loss or pollution is easy to cause. In addition, volume is a core parameter for evaluating the environmental migration and biological effectiveness of microplastic, but the existing method can only estimate based on spherical assumption, the error of irregular particles (such as fibers and fragments) exceeds 30%, and the time cost of direct measurement means (such as a layered imaging system) on single particles exceeds 5min per sample, so that the efficiency is low. In addition, the prior art has low data integration and standardization, the data generated by multi-instrument combination needs to be manually integrated, the associated characteristics such as morphology, material, volume and the like can not be synchronously output, and a micro-plastic multidimensional characteristic database is difficult to quickly form so as to support subsequent analysis. Disclosure of Invention Aiming at the problems existing in the prior art, the invention aims to provide a high-throughput rapid detection method for the multidimensional features of microplastic. The multi-dimensional feature synchronous detection method realizes that an automatic process replaces manual operation, multi-mode data fusion and synchronous characterization, and standardized data integration and report generation, and automatically outputs a structured report containing parameters such as particle ID, particle size, shape, material, volume and the like, supports cross-research data comparison, and promotes detection standardization. Breaks through the efficiency, precision and dimension limitation of the traditional detection method, and provides a reliable tool for the research of the microplastic pollution mechanism and the environmental monitoring. In order to achieve the above purpose, the present invention adopts the following technical scheme: the invention provides a high-throughput rapid detection method for multidimensional features of microplastic, which comprises the following steps: s1, acquiring a microscopic image and a corresponding coordinate of a micro-plastic sample, and acquiring Raman spectrum data of the coordinate position; S2, carrying out automatic image processing on the microscopic image obtained in the step S1, identifying particles of the microplastic, and extracting morphological characteristics including particle size, shape factor and particle number; Preferably, the step S1 specifically includes the following steps: S11, acquiring a microscopic image of a sample by adopting a low-power microscope, and identifying micro plastic particles through gray threshold segmentation to obtain coordinate positions of the micro plastic particles; And S12, automatically positioning a Raman spectrometer to the coordinate position, collecting Raman spectrum data, and determining the micro plastic material through spectrum matching based on the Raman spectrum data. And S13, performing spectrum matching based on the characteristic peak position and the intensity similarity threshold value on