Method and apparatus for microplastic identification with polarized digital holography
AssigneeThe University of Hong Kong
InventorsYanmin Zhu, Yuxing Li, Yin Mun Edmund Lam
Life of this patent
Estimating the term…
Abstract
A method for aquatic microplastic identification is provided. The method includes steps as follows: emitting a laser beam from a laser source, such that the laser beam passes through a liquid sample with MP samples in a sample channel and then a polarizer and is received by polarization camera; capturing a sample image of the MP samples by the polarization camera, wherein the sample image comprises interference patterns resulting from superposition of object and reference waves; and feeding the interference patterns into a morphology analyzing module for real-time tracking and analyzing by using a lightweight convolutional neural network (CNN) model, so as to classify the MP samples.
Claims20
An apparatus for aquatic microplastic identification with polarized digital holography, comprising:
- a laser source configured to provide a laser beam;
- a polarization camera positioned under the laser source such that the laser source aligns with the polarization camera to optically couple with the polarization camera for emitting the laser beam to the polarization camera, wherein the polarization camera comprises a polarizer mounted in front of the polarization camera so as to optically couple with a camera lens of the polarization camera and to enable the camera lens to obtain images with different polarization states simultaneously;
- a sampling channel positioned between the laser source and the polarization camera;
- a recorder electrically coupled with the polarization camera, wherein the polarization camera is further configured to capture sampling images of microplastic (MP) samples in the sampling channel and transfer them to the recorder;
- and
- a morphology analyzing module electrically coupled with the recorder and configured to access the recorder for real-time tracking and analyzing, so as to classify the MP samples, wherein the morphology analyzing module classifies the MP samples based on interference patterns resulting from superposition of object and reference waves shown from the sampling images.
An apparatus for aquatic microplastic identification, comprising:
- a housing having a first room and a second room which are separated from each other to generate a sample channel therebetween;
- a laser source disposed within the first room and configured to emit at least one light beam toward the sample channel;
- a polarization camera disposed within the second room, wherein the laser source aligns with the polarization camera to optically couple with the polarization camera for emitting the laser beam to the polarization camera, and wherein the polarization camera comprises a polarizer mounted in front of the polarization camera so as to optically couple with a camera lens of the polarization camera and to enable the camera lens to obtain images with different polarization states simultaneously;
- and
- a microprocessor electrically coupled with the polarization camera and applying a morphology analyzing module for real-time tracking and analyzing, so as to classify MP samples within the sample channel, wherein the morphology analyzing module classifies the MP samples based on interference patterns resulting from superposition of object and reference waves shown from sampling images captured by the polarization camera.
A method for aquatic microplastic identification, comprising:
- emitting a laser beam from a laser source, such that the laser beam passes through a liquid sample with MP samples in a sample channel and then a polarizer and is received by polarization camera;
- capturing a sample image of the MP samples by the polarization camera, wherein the sample image comprises interference patterns resulting from superposition of object and reference waves;
- and
- feeding the interference patterns into a morphology analyzing module for real-time tracking and analyzing by using a lightweight convolutional neural network (CNN) model, so as to classify the MP samples.
Dependencies are inferred from the claim text.