Abstract:
The detection of plastic using remote sensing images is a current research demand
due to piling heaps of plastic in the ocean. Plastic is ubiquitous and exerts a
negative influence on marine biota and humans. The plastic cycle consists of
emissions, transportation, weathering, and accumulation. Plastic processing
algorithms are expected to understand these various stages which are critical in
eliminating them. This study introduced a novel index to identify floating macro
plastic in Sentinel 2 satellite images. The atmospherically corrected images using
Acolite and Sen2Cor algorithms were used in testing this index. The index
protected the plastic information and maximized the separation from surrounding
objects. A convolution high pass filter (3x3) was applied after the index to enhance
the plastic objects. The categorization of plastic and nonplastic information was
done by using the scatter plots. These scatter plots were made by placing the Index
applied convolution high pass filtered Acolite/ Sen2Cor image as the “X” axis and
Sentinel 2 bands 5,8, and 9 as the “Y” axis. The index and the scatter plot analysis
identified plastic pixels with more than or equal to 14% plastic bottle percentage.
The other plastic types, such as fishing nets and plastic bags, required pixel
percentages above 50% to be detected. The pixels with high plastic percentage and
100% coverage were located as a separate cluster in the scatter plot analysis.
Therefore, the pixel plastic percentage and the pixel plastic coverage are important
contributors to the accurate detection of plastic. The plastic detection was not
successful for the dates with aerosol, clouds, and smooth sea surface conditions.
The Acolite and Sen2Cor images are not suitable for plastic detection when the
plastic signal is weak.
