The man overboard system from PureTech Systems utilizes a patented approach using geospatial video analytics and pairs of thermal imaging cameras which face each other. The synchronized video clips are then used to accurately detect a fall. At the same time, the system also ignores environmental interference, background imagery from various sea states or shoreline views and avoids false alarms due to normal operating conditions such as blowing debris, crew operations and other deck activity.
In addition to having tremendous potential to cater to the current requirements of the marine industry, man overboard technology is directly applicable to the broader safety category of “long-fall events” and applicable to many additional applications.
Reusing the Technology
Accurately detecting a human falling from a considerable height in an outdoor environment is not an easy problem. An object falling from such heights can reach speeds of 70 mph. A long-fall detection system therefore must be able to detect objects moving at these speeds, while accommodating for the surrounding environment, including varied weather conditions. A reliable detection system must accurately detect human falls in all these situations, while avoiding false alarms and providing actionable data to react to the situation.
Key components of the man overboard detection solution which are applicable to other long-fall situations include the ability to determine an object’s real size, classification, location, speed and trajectory – confirmed via synchronized data from two independent camera views.
Although cruise lines and ferries visiting U.S. ports and rated with passenger loads greater than 250 are the only vessels currently subjected to man overboard detection system mandates, smaller cruise ships and ferries, commercial ships, off shore platforms and even military vessels are all subject to a man overboard event. Although not a mandate, many of these users may have the desire to install MOB systems for added safety or to reduce operational liability. Video detection is highly scalable; to apply such a system design to these additional platforms merely requires a change in the camera mounting locations, and potentially a lens change. For some markets, a lower cost camera may be a more appropriate choice to ensure both safety and affordability.
Bridges / Buildings
The use of long-fall technology on bridges, buildings and commercial sites is also directly applicable to the long-fall technology used for MOB events. In these installations, additional savings may be realized as there may not be the need for higher priced marine-grade cameras, which have extra protection against extreme weather and a corrosive salt environment.
Although cruise ships try to avoid operating into high risk areas, the threat of piracy, hijacking and kidnapping is still a safety issue which must be considered. Ideally, detection and deterrence measures can be taken early to keep pirates at a safe distance from the vessel, but in situations where this cannot occur, or in the case of a covert boarding attempt, the same technology used to detect a person falling from a ship can also be used to help detect an illegal boarding attempt.
When used in this manner, the system is not recognizing a fall, but rather the class of the object (human, boat, etc.), along with location data, speed, trajectory and background modeling to determine if a person is moving upward, onto the ship. If this is the case, the system can provide a video-based anti-boarding alarm for verification and immediate action by the crew.
When we think of long-fall events, we typically think of the unfortunate scenario where the falling object is a human. However, many industrial operations have the need to monitor for dropped objects, including equipment that has been dropped. In some cases, this could relate to a safety action like the sounding of an alarm or confirming that implemented safety processes are working. It may be used to help find the equipment after it has been dropped, saving the time spent finding the lost item and/or the cost of replacement. Other uses are not necessarily tied to the detecting of the drop itself, but rather to the idea that a tool or piece of equipment has been placed somewhere. In such instances, the video analytics can identify this dropped or placed object and provide a notification for future reference should the item be reported as misplaced.
In the case of dropped objects, versus human sized targets, the resolution of the video must be increased to accommodate the appropriate number of pixels to detect and differentiate the object from the background and environmental conditions. Similarly, the distance at which the item may be detected will also decrease due to the smaller object size.
Detection Using Intelligent Video
The use of geospatial video further enhances the detection capabilities of this type of solution. Geospatial video is the understanding of where each video pixel resides in “real” space – meaning latitude, longitude and elevation. This adds another dimension to video analysis by allowing the software to not only understand the physical location of the object, but also the real size, the real speed and the real acceleration. Therefore, although an object may be falling at the same expected speed of a human, a geospatial video solution can understand that this same object is too big or too small to be a human, and suppress the alarm.
The use of opposing cameras provides a means by which the video analytics can further confirm the event is happening on the ship itself, and is not something occurring in the background, such as on the water, in the sky or on the dock. In addition to actually being detected in both images, an event occurring between two camera pairs must pass several additional tests for the duration of the fall – including time stamp, object size, object location, object speed, object type and fall trajectory – on both cameras within the pair before being considered an alarm, thus reducing false alarms.
The use of intelligent video for these long-fall scenarios also increases the overall understanding of the event when the alarm is raised to appropriate safety or security personnel. Specifically:
Image Data – As humans, we inherently rely on vision as one of our primary senses; the human mind will rely on its sense of sight as a means to confirm any other data received. Using video analytics for detection means data preferred for validation by a human – still images and looping video – is instantly available to reduce the time required to confirm the event. To further ensure rapid verification, video analytic systems highlight the potential target with graphical “bounding boxes” on both still images and full motion video clips in order to quickly communicate what event raised the alarm.
Location Data – In addition to logging the GPS location of the vessel ship at the exact time of the event, video solutions leveraging geospatial data also provide insight as to the location where the event took place on the vessel or the structure itself. Including location information with each alarm allows first responders to more quickly reach the incident location or confirm the point of origin.
The cruise line industry has recognized the need for man overboard detection systems which provide accurate detection and notification of falls from sea-going vessels. This same video-based technology, which monitors thermal images, processes them using intelligent video algorithms and combines this data with geospatial techniques, is also readily applicable to other maritime industries, including ferry operators, oil platforms, commercial shipping and military vessels. Other types of industries that have the need to monitor for long falls of humans or dropped equipment, such as bridges, buildings, work sites and other industrial location, can also benefit from this type of video detection technology.
Learn more about other types of intelligent video monitoring.