Surveys of seabirds and marine mammals using digital aerial counts are the basis for many environmental impact assessments as well as population monitoring in German offshore waters. Automation of these surveys through AI-supported image data analysis has been a research activity at BioConsult SH for quite some time.
In comparison to manual processing, use of AI for a semi-automated evaluation of digital aerial images promises considerably faster analysis.
However, can an AI-based approach achieve the same quality of results as trained observers? This is a prerequisite for establishing automated object recognition as a standard, for example in maritime spatial and environmental planning.
A study now published in the Vogelwarte journal showed that the HiDeFIND AI model developed by BioConsult SH reached a very high sensitivity of over 99 % for seabirds and marine mammals in digital video. This included key species of marine communities such as guillemots, red-throated divers and harbour porpoises.
HiDeFIND therefore is a promising alternative to manual object detection for analysing digital aerial data of seabirds and marine mammals.
BioConsult SH already applies the model in maritime spatial and environmental planning, environmental monitoring and research to support the data collection process. Quality assurance is still carried out by highly trained observers, who ensure a consistently very high data quality.
Please refer to Topics and Services on our website for further information.
Here, the article from the Vogelwarte journal as well as a popular science article on the same topic that was published in Der Falke are available for download:
