Man overboard events continue to be a common occurrence in the cruise industry. Such events can occur at all hours of day and night, all weathers and sea states. If a fall is from a great height, such as the top of a cruise vessel, the fall can injure, knock unconscious, or even kill the victim. If the air or water temperature is cold, a more than temporary exposure can lead to hypothermia and death.
The incidence of man overboard events on ocean-going passenger cruise vessels has been of such concern, that the US Congress recently passed the Cruise Vessel Safety and Security Act which requires that such vessels now integrate technology to detect passengers who have fallen overboard.
To meet the needs of the new legislation, engineers at Arizona-based PureTech Systems have developed a new patented man overboard detection software to do just that. When combined with the company’s PureActiv geospatial video management software, the system can alert a ship’s crew that such an incident has occurred and provide them with complete visual logs, enabling them to dispatch search and rescue teams in a timely fashion.
According to Dr. Fereydoun Maali, Chief Imaging Scientist at PureTech Systems, the system uses a number of opposing video camera pairs mounted at strategic locations on a vessel where they are oriented so that they can capture images of any objects falling from the ship. Although many different types of cameras could potentially be used, at present, the systems that have currently been deployed employ thermal video cameras that can image objects at all times during the day and night.
More specifically, each pair of thermal video cameras image an overlapping volume between them with coverage extending from the lowest passenger deck to the waterline. Mounting the cameras in such a manner ensures that the system can capture images of all objects that fall from the vessel, because any object that does fall will necessarily pass the lowest deck on the ship irrespective of the deck from which it has fallen.
Time stamped video images from each pair of cameras are delivered to a video analytics server in a networked IT center in the electronics room of the ship. Custom image processing software running on the server then analyzes the data to determine the characteristics of the falling object and predict the likelihood that a man overboard event has occurred.
Software pairs and tracks each of the blobs in the pair of images...
To do so, the pair of images is first preprocessed to classify the images into background and foreground regions. A segmentation algorithm then partitions the foreground pixels in the images into clustered sets of pixels, or blobs, whose characteristics may conform to those of an object falling from the ship. Bounding boxes are then wrapped around each of the blobs, from which the characteristics of the object can be calculated.
In situations such as man overboard events, providing a high level of situational awareness helps first responders gain a full understanding of the event as quickly as possible. This is achieved through geolocation of the alarm location, instant access to detection images, looping videos of the event, as well as audio notification.
So that the system does not overlook any potential fall, the software concurrently pairs and tracks each of the blobs in the pair of images. By tracking the centroids of the blobs with respect to one another, the software can create a track segment which represents the trajectory of the fall of the object in image space. For each paired object, the software then calculates the location of the object in real world space with respect to the vessel, as well as the trajectory of the falling object and its actual speed and acceleration.
“By analyzing the characteristics of the images from both cameras, the software determines how well the size and temporal properties of the objects represented by the blobs conform to one another. If the objects do not demonstrate conformance, the objects are either dismissed as not being due to the same actual object, or are negatively scored. On the other hand, if the properties do conform, then the width and height of the object is calculated and compared to the expected width and height of a falling person. In doing so, the software can accurately classify a fall event while suppressing any events that could lead to a false alarm,” said Dr. Maali.
When deployed on a cruise liner, the modular image processing software is fully integrated into PureActiv Systems’ geospatial video management system, which uses the results from the image analysis operations to highlight to ship’s personnel where the event has occurred.
To do so, a visual indicator, which is mapped onto a plan view of the ship, is transferred in real time from the central server to a PC-based display system on the bridge together with looping videos and still images of the event, as well as an audio alarm indicating that such an event has occurred. In addition, ship personnel can immediately view live feeds from cameras at the very location on the ship where the event has occurred.
Prior to rolling out the system, PureTech Systems trialed the system extensively on several passenger vessels, determining its accuracy by detecting falls from lifelike dummies of both adults and children from various decks. The system has under gone testing along both warm water and cold water routes aboard super-liner class ships, successfully detecting such events under a variety of conditions.
It is currently awaiting formal industry testing requirements from the recently formed International Standards Organization (ISO) Man Overboard Detection working group to facilitate a formal industry certification status.
Written by Dave Wilson, Senior Editor, Novus Light Technologies Today