Image Credit: sonoware GmbH
Detecting Ghost Nets Using Side-Scan Sonar and Machine Learning: sonoware Receives €110,000 AI Funding Grant
sonoware GmbH and GEOMAR Helmholtz Centre for Ocean Research Kiel have been awarded €221,000 from the AI Fund of the State of Schleswig-Holstein for their collaborative project: GhostNetBusters.
Ghost Nets: The Silent Killers of the Baltic Sea
Lost fishing gear is currently the largest source of pollution in the world’s oceans. According to the NGO The Ocean Cleanup, 92% of the Great Pacific Garbage Patch is composed of fishing gear that has either been lost or illegally disposed of at sea. Data from the World Wildlife Fund for Nature (WWF) indicates that ghost nets account for 30% to 50% of ocean plastics. The Food and Agriculture Organization of the United Nations (FAO) estimates that approximately 12,500 km of ghost nets are currently located within the Baltic Sea alone. These ghost nets either float through the currents, continuously trapping marine wildlife, or sink to the ocean floor, creating deadly blankets over whatever they entangle. In the Baltic Sea, shipwrecks and bridge pillars frequently trap ghost nets, turning them into lethal traps for marine life that use these structures as habitats.
Addressing the Need for Identifying Ghost Nets with Existing Equipment
GhostNetBusters aims to utilize existing side-scan sonar technology to detect the iron and steel lines commonly found in gill nets on the ocean floor. These materials appear as glowing, tangled ropes in sonar images and can be difficult to distinguish from the ocean floor, even to the human eye. Areas with complex sediment shapes or other geological or biological structures often require human verification when a ghost net is suspected. Since these dives require significant effort from volunteer divers and specialized equipment, locating and removing ghost nets is an expensive process. By applying machine learning algorithms to existing side-scan sonar technology, GhostNetBusters aims to significantly increase the efficiency of ghost net removal by reducing the need for unnecessary verification dives.
“There are already numerous efforts underway to remove ghost nets. We aim to support in finding and localizing these,” said Mia Schumacher, Data Scientist at GEOMAR Helmholtz Centre for Ocean Research Kiel.
Maximizing Efficiency for Existing Sonar Technologies
Side-scan sonar technology is commonly used to investigate the ocean floor. However, many conventional object detection algorithms require significant processing power or are incapable to comprehend the challenging data quality of commonly used side-scan sonar technology. Even fewer systems are capable to do so on board and during mission operation. To enable even smaller systems to utilize onboard object detection software, sonar data needs to be preprocessed using noise reduction algorithms and then interpreted on embedded platforms within Autonomous Underwater Vehicles (AUVs) or Remotely Operated Vehicles (ROVs). It’s essential that these processing units do not reduce mission time by draining battery capacity, so the algorithms must be efficient. If successful, the ghost net detection algorithms could be implemented across all commonly used AUVs and ROVs.
“I’m curious to see how classic digital signal processing can be enhanced with artificial intelligence to add value both to this project and to other maritime applications. The atmosphere among the partners is great, and I’m excited to see the results of our joint efforts,” said Christian Lüke, Chief Executive Officer at sonoware GmbH.
Partnerships Among Equals Create a Strong Project Team
The project team consists of GEOMAR Helmholtz Centre for Ocean Research Kiel (lead) and sonoware GmbH. Associated partners include One Earth – One Ocean e.V. and the Scientific Diving Association e.V., both of which have a proven track record of removing large volumes of ghost nets from the Baltic Sea. The project launched in June 2024 and will conclude in December 2025.
The project is funded by the State of Schleswig-Holstein as part of its AI initiative.
