Conclusion 
Small object detection is the basic intention of the performance standard for the night vision equip- 
ment for high-speed crafts. The current state presents an opportunity for advanced and inexpensive 
alectro-optical sensor technology combined with image processing using neural networks. This could 
eaxceed performance speculations made at the turn of the millennium. Even though classification is not 
yet reliable, possible indications from the system could improve situational awareness and close the 
zap between radar and direct view. It is strongly recommended to include an interface to the ship gyro 
Compass system in order to enhance the performance. Based on this conclusion, it should be reviewed 
now to incorporate optional daylight cameras and optional Al-based object detection into the existing 
ISO 16273 standard for maritime night vision systems. This will enable a sensible application in the 
maritime industry without parallel standards created by different technical expert groups. The initia- 
tive to harmonise situation awareness systems should be closely linked to this approach. 
in order to enhance maritime Al-based object detection, it is in the interest of the industry to record 
additional maritime datasets as it moves towards a higher automation of nautical tasks. Maritime da- 
tasets should be made accessible to all interested parties. Nevertheless, approving authorities need to 
maintain a non-public dataset for validation purposes. It is advisable to standardise a minimum set of 
object classes used by all parties in order to ensure the universal and easy adoption of any equipment 
5y all seafarers. Examples for such classes could be “buoy”, “sail boat”, “vessel”, “obstacle”, “harbour 
infrastructure”, “energy harvesting”, and “shoreline”. The interaction between EO sensors and situa- 
tion awareness functions should be reviewed and defined. The market overview of the Al systems 
oresented in excerpts in this report highlights that situation awareness should not be limited to active 
or passive EO systems but rather should include AIS and RADAR as well as other potential maritime 
data sources. It would be beneficial to have performance indicators for object detection and semantic 
segmentation for both the IMO and ISO. The performance metric mAP is currently established and 
shall be considered. 
Acknowledgment 
We are grateful for the support of crew of the research vessel ATAIR and crew of the rescue vessel 
NIS RANDERS as well the employees from German Maritime Search and Rescue Service (DGzRS) sup- 
oorted the field test. The technical and scientific work presented here, was funded by the Federal 
Ministry for Digital and Transport.