Optical Detection of Marine Macroplastics using Deep Knockoff
Abstract
This article investigates visible to near infrared spectral signatures of marine macroplastic, using the deep knockoff as a modern statistical variable selection tool. To that end, a public dataset obtained from ground-based labeling of the observations by the Multispectral Instrument (MSI) on-board SENTINEL-2 satellite are used. The deep knockoff enables to identify the most important MSI spectral bands that respond to the occurrence of plastic debris through controlling the false discovery rate (FDR), without any Gaussian assumption or linear relationship between the spectral response and occurrence of marine debris. Consistent with the known physical causalities, the results indicate that the near infrared (band 8, 833 nm) and red (band 4, 665 nm) are the most important bands, respectively, that can confine the FDR, when it comes to discrimination of marine plastics from the background water. In the presence of dense Sargassum macroalgae, two other bands are essential to detect marine debris using MSI data including the green (band 3, 560 nm) and the narrow NIR (band 8a, 864.7 nm)
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFMOS32C1026O