Plastic Litter Project (PLP) - elena-andreini/TriesteItalyChapter_PlasticDebrisDetection GitHub Wiki

Discripation:

Remote Sensing of Sea Surface Artificial Floating Plastic Targets with Sentinel-2 and Unmanned Aerial Systems (Plastic Litter Project 2019)", focuses on detecting floating marine plastics using remote sensing technologies, specifically Sentinel-2 satellite imagery and Unmanned Aerial Systems (UAS). The study is part of the Plastic Litter Project (PLP), which aims to understand the spectral behavior of floating plastics and develop methods for detecting them in marine environments

Satellite Mission: Sentinel-2

Data Type: Imagery

Imagery Level:

Level-1C (Top-of-Atmosphere reflectance, TOA)

Level-2 (Bottom-of-Atmosphere reflectance, BOA)

Atmospheric Correction:

Applied: Yes

Algorithm: ACOLITE (version 20190326.0)

Bands:

The study primarily used the 10-meter resolution bands (RGB and NIR: B2, B3, B4, B8). The 20-meter and 60-meter bands were not used due to the small size of the targets.

Spatial Resolution:

10 meters for RGB and NIR bands (B2, B3, B4, B8)

20 meters for other bands (not used in this study)

Temporal Coverage:

The study was conducted over a two-month period (May–June 2019), with 13 scheduled Sentinel-2 acquisition dates.

Due to weather conditions (clouds, high winds), only 5 usable Sentinel-2 images were captured during the experiment.

Geographic Coverage:

The study area was located in the coastal waters of Lesvos Island, Greece, near Tsamakia Beach in Mytilene.

The targets were deployed in open water conditions, above seagrass meadows (Posidonia oceanica), to minimize seafloor reflectance.

Labels/Annotations:

Type:

Pixel-wise annotations based on UAS imagery.

The percentage of plastic coverage in each Sentinel-2 pixel was calculated using object-based image analysis in eCognition Developer 9.5.1 software.

Classes:

Plastic debris (PET bottles, plastic bags, and natural debris like reeds).

Water (seawater).

Mixed pixels (combinations of plastic and water).

Debris-Free Scenes:

Yes, the study included debris-free scenes (pure seawater pixels) for comparison with pixels containing plastic targets.

Relevance to Project: This dataset is highly relevant for detecting floating marine plastics because it provides ground-truth data from artificial plastic targets deployed in a real marine environment.

The combination of Sentinel-2 imagery and UAS data allows for the validation of remote sensing algorithms for plastic detection.

The study demonstrates the potential of using multispectral satellite data to detect plastics, even at low coverage levels (e.g., 25% PET abundance in a pixel).

Access:

The dataset used in this study is available for download from the University of the Aegean's Marine Remote Sensing Group webpage: http://mrsg.aegean.gr/.

It is also available on the ZENODO database: https://doi.org/10.5281/zenodo.3752719.

Dataset Contents:

Sentinel-2 L1C granules catalogue (Word document).

Sentinel-2 L2 subsets (after atmospheric correction with ACOLITE, in netCDF format).

UAS images of the plastic targets for each acquisition date.

Point vectors with percentage plastic presence for each Sentinel-2 pixel (in ESRI vector format).

Additional Notes: Preprocessing: The Sentinel-2 data was atmospherically corrected using ACOLITE, which also performed sun glint removal and cloud masking.

Data Quality Issues:

Sun glint and cloud cover were significant challenges, reducing the number of usable images.

The small size of the plastic targets (5x5 meters) made detection difficult, especially in mixed pixels.

Usage Restrictions: The dataset is openly available for research purposes, but users should acknowledge the source and cite the original paper.

Next Steps: Download Imagery: Access the Sentinel-2 and UAS data from the provided links.

Apply Further Corrections: If needed, apply additional atmospheric or glint corrections using tools like ACOLITE or Sen2COR.

Validate Detection Algorithms: Use the ground-truth data (UAS images and plastic coverage percentages) to validate and refine plastic detection algorithms.

Expand to Larger Areas: Apply the methodology to larger coastal areas or other regions to test its scalability.