Researchers are combining AI-based satellite image recognition with drift prediction models to improve collection of floating space debris in the ocean.
Identifying and tracking floating space debris is essential to ocean cleanup efforts. However, despite the abundance of satellite imagery and weather data currently available, effective systems for doing so remain elusive.
The AI for Detecting Ocean Plastic Pollution with Tracking (ADOPT) project aims to change this with two types of systems.
“One is to analyze satellite images to identify the garbage patch, and the other is to predict where the garbage patch will be floating by the time the cleaning team arrives at the patch, usually within 24 hours,” explained ECEO scientist Emanuele DalSasso.
The idea is to fulfill a simple need. Governments and NGOs cannot respond immediately when debris is discovered, as it takes time to organize and deploy cleanup efforts.
High-resolution images make it easier to track floating space junk
The ADOPT team initially worked with open-access data collected by the Sentinel-2 satellites, a series of optical imaging satellites launched by the European Space Agency.
But tracking floating debris is difficult because these instruments only pass a particular point in the ocean every six days, and the image resolution is low, at just 10 meters per pixel.
To address these two shortcomings, the team designed an AI system that can also be trained on PlanetScope data. PlanetScope data is a constellation of hundreds of microsatellites that collect images daily at a resolution of 3 to 5 meters per pixel.
The result is an AI-driven detector that captures data from both sources and is updated daily with high-resolution images without the need for data annotation.
Predict where debris will drift
Once floating debris has been identified, the next step is to predict where it will drift by the time the cleanup team arrives.
The second system, developed by Christian Donner of SDSC, is designed to make such short-term predictions. “Models often have biases, so I take widely used models for predicting wind and currents and apply machine learning to correct them,” he said.
“The machine learning program compiles data from different sources and adjusts for these biases to more accurately predict the trajectory of floating debris.”
Only a small amount of field data on the debris patch was available, so he trained the program using data from GPS-equipped drifters as a proxy. These drifters were deployed under the Global Drifter Program and have been used to collect measurements since the 1990s.
overcome the problem of bad weather
However, there’s one big problem. The system doesn’t work well in bad weather, and the optical sensor doesn’t work when you’re above clouds.
“Incorporating radar imagery from Sentinel 1 could be an option. Radar signals can travel through clouds and work day and night,” Dalasso said.
“However, they only provide information about the texture and shape of floating space debris, which means that important spectral signatures detected by optical sensors and essential for debris patch detection are missing.”
For now, the ADOPT team is not considering combination options for radar optics. This may be undertaken by other research groups in the future, as the project will officially end this fall when the two-year funding program ends.
The team plans to leave behind a solid proof of concept, two publications currently being completed, and code for both systems (debris detection and drift prediction).
Going forward, the Dutch NGO Ocean Cleanup will continue to compare algorithms, and the university’s scientists will continue to collaborate to further advance the research.
Source link
