Man Overboard

detect persons going overboard

The system is designed to detect and alert when a passenger is in danger of falling or going overboard from a cruise or passenger ship

The system is designed to detect persons going overboard from passenger ships and cruise ships.
The method of detection does not require the passengers to wear a device to trigger an alarm. The challenge is to achieve high rates of detection of a human falling overboard, while keeping the false alarms limited to no more than one, daily.

The unique offering of this project, consists in the development of a data fusion middleware that builds on an innovative set of image recognition and motion detection algorithms, aimed at lowering false alarms and achieving accurate detection of real (human-shaped) targets.

The system will operate with very good performance at a wide range of sea condition which affect the movement of the ship. Feature reduction algorithms are guaranteed to discriminate actual events even in the most crowded scenes (when people gather in one of the ship’s observation decks). The sensors’ location and coverage is such that all areas around the ship and to an extent (typically 5m) from the ship’s hull are covered.

It is very important to mention that the system is not based on a single-factor authentication method, i.e. the process does not rely on one method to identify a user. Instead the system relies on Image Classification, Trajectory and Thermal Footprint analysis.

Following the alarm, the safety responsible will use a control station to review the footage of the event from the available sensors.

The solution has received funding from the General Secretariat for Research and Technology (GSRT) under the project name “METIS – Multimodal machine learning and signal processing technologies from multiple heterogeneous information sources for safety applications and surveillance of critical infrastructures“.