Measurement

Overview

Application of Computer Vision

To further improve the efficiency of detecting microplastic particles, we used computer vision and artificial intelligence to detect microplastic particles in images.

Computer vision methods were first used to generate the corresponding labeling information. For image noise, we used mean shift filtering with Gaussian blur mitigation. The noise removed image is grayscaled and adaptively binarized. The results are shown below.

The final binarized image undergoes contours detection and size filtering to get the annotation.

Pre-Experiment - Microplastic Validation in Bottled Water

Numerous studies have reported the presence of microplastics in bottled water. According to reports by Mason et al. and others, PET (polyethylene terephthalate) is the most abundant type of microplastic particles in bottled water, constituting % of all detected microplastics, followed by %, as confirmed by many other articles. However, due to regional variations, it is essential to verify whether these conclusions about the types of microplastics align with our local circumstances before proceeding with relevant engineering.

Therefore, we chose to conduct microplastic testing on commonly available bottled water locally and compared it with commonly researched bottled water types. We utilized the micro-Raman spectroscopy method as employed by Schymanski et al. in their study (Schymanski et al. 2017).

We used the same gold foil PVDF membrane for vacuum filtration on various water samples, with a filtration volume of 100 ml and an effective filtration area of 3.1 square centimeters. The selected water samples included locally prevalent bottled water brands and some brands commonly found in research. These brands consisted of Aquafina 500 ml bottled water, Nongfu Spring 250 ml bottled water, Yi Bao, Baishi Shan, Nestle Pure Life, Kang Shifu iced tea, and Coca-Cola.

After concentrating microplastics from the water samples onto the filter membrane through vacuum filtration, we marked the membrane's surface. Subsequently, we sent it for inspection and qualitative analysis using Raman spectroscopy. Upon completing the analysis, we compared the data to the Raman spectra in the openanalysis, an open-source microplastics database platform (Cowger et al. 2021). The comparison was performed using Baseline correction, as well as Smoothing/Derivative with the specified Preprocessing settings.

The results of this data comparison aligned with previous literature findings. We confirmed that PET (polyethylene terephthalate) was the predominant microplastic particle detected, constituting 70% of our total samples. Based on peak comparisons, we also identified the presence of PC (polycarbonate) besides PET.

After synthesizing the BAM membrane, we needed to conduct BAM membrane effectiveness testing on water samples containing microplastics. The preparation of the microplastics water sample was performed using the following steps:

Materials used:

• PS microplastics 0.1mm microspheres (purchased from Yuan Biotech.)

1.1g of plastic microspheres was extracted using an iron spatula. The plastic microspheres were ground for 1 minute using a 6cm ceramic mortar and pestle. The submerged plastic powder was stirred with 1000ml of ultrapure water using a glass rod for one minute.

Both the pre-test and post-test samples were enriched and sent for inspection using the same procedures as the water sample analysis described earlier. The results of the Raman spectroscopy testing on the pre-filtered samples were similar to the previously presented results, with peak heights resembling the PS spectra in the database. This confirmed that the water samples prepared were essentially free from external microplastic contamination. The results of microplastic content detection showed a 97% reduction in microplastic content.