Visualising the Invisible: Visualising Historic Shipwrecks
In 2001 the School of Media Arts and Imaging established the Centre for Digital Imaging (CDi) with strategic funding from University of Dundee and Scottish Enterprise, Tayside, Scotland. The research remit of the Centre is premised upon the promotion of a sustainable research ethos, regarded as essential to the establishment of a centre of excellence in the developing field of 3D visualisation.
Article Authors: Steve Flack, Chris Rowland
In 2004, Head of School, Steve Flack and Senior Researcher, Chris Rowland, established the “3D Visualisation Research Group” as the research arm of CDi. The group was established to investigate the application of 3D technologies and techniques in medical imaging, forensics, architecture, anatomy, archaeology and long term illness patient communication.
Rowland works as part of a collaboration between St Andrews University and Dundee University working under the name ADUS (Archaeological Diving Unit Surveys), that represents the forefront of innovative, high-definition, multi-beam sonar surveys of archaeological sites. The ADUS research team includes Chris Rowland (3D Visualisation), Martin Dean, (Maritime Archaeologist, St Andrews University), and Mark Lawrence, (Maritime Archaeologist, SOMAP). The team produces unique animated 3D visualisations using point clouds derived from surveys of historic wreck sites in coastal waters. These highly detailed 3D visualisations are being used to inform UK government policy to deal with ‘environmentally sensitive’ wreck sites.
The expertise and experience profiled by ADUS has led to the commission of a number of significant surveys. This includes an ongoing forensic study of the SS Richard Montgomery munitions wreck in the River Thames for the Maritime and Coastguard Agency (UK Government, Department for Transport). The methodologies and techniques developed in this landmark survey led to the commission to survey the wreck of the Royal Oak in Scapa Flow.
Background: The Sinking of HMS Royal Oak
On the night of the 13th October 1939, in a daring mission, U boat commander Gunther Prien managed to manoeuvre his submarine U47 into Scapa Flow through the narrow channel at Kirk Sound that had been thought to be effectively blocked by sunken ships designed to deter the enemy. Shortly after midnight U47 approached the unsuspecting Royal Oak laying at anchor in Scapa Bay. Most of the fleet was out to sea, and HMS Royal Oak was the only capital ship present.
The Royal Oak was thought to be well protected against submarine attack, but after being struck by four well placed torpedoes from U47 the ship listed to starboard and sank within 15 minutes, becoming the first battleship to be lost in the Second War. Gunther Prien returned to Germany to receive a hero’s welcome, but 833 men of a crew of 1, 234 aboard the Royal Oak had died.
The wreck has been leaking oil into Scapa Bay for the past 67 years and still contains a significant cargo of munitions. The wreck is protected by UK government and is listed as an official war grave, which prevents recreational divers from approaching the site. The Ministry of Defence, Salvage and Marine Operations unit are responsible for the wreck and have been working on various methods to extract the remaining oil from the hull.
Surveying the Site
Chris Rowland and ADUS survey the HMS Royal Oak site
In 2006, the Salvage & Marine Operations unit commissioned ADUS to provide highly detailed sonar images of the current state of the Royal Oak, The sonar images created by ADUS are of such accuracy that even small changes in the hull over time, through deterioration, can now be monitored closely year on year.
The survey undertaken by the ADUS team used the ‘multibeam sonar system’, to produce images of the wreck of the Royal Oak. Unlike side scan sonar, multibeam systems can provide bathymetric (depth) data for every part of the seabed that provides an acoustic return, allowing three dimensional digital terrain models (DTMs) to be created.
|ISHAP deployment system|
High-definition side scan sonars can provide excellent images, and were considered, but they lack the positional accuracy of multi-beam equipment. The system, based around the Reson 8125 Seabat, is deployed from a survey boat and is mounted on the end of a deployment system (referred to as ISHAP) developed by ADUS in order to get the sonar head as close to the wreck as possible when in the water.
The multi-beam and positioning system used for this survey by ADUS allowed the production of a Gazetteer of Observations that includes all the main features of the wreck, relatively positioned to centimetric accuracy and absolutely to within c.10cm.
In order for the multi-beam system to produce such detailed images the sonar head itself must be very accurately positioned. This is achieved by using satellite positioning (GPS) with an additional correctional base station set up on the nearby shore. In addition the motion (i.e. roll pitch yaw and heave) of the survey boat has to be effectively compensated for in the data collected by using very sensitive motion reference systems onboard.
The sonar system builds a three dimensional model of the wreck and the seafloor by collecting many millions of accurately positioned ‘spot heights’ derived from the sonar ‘pings’ sent and returned to the sonar head as the survey boat travels back and forth over the wreck. The many millions of ‘XYZ’ points generated from the survey are edited to filter out digital noise and the resulting dataset is imported into animation software.
The Royal Oak shipwreck visualisation project commenced by identifying the main issues with creating a visualisation from multi-beam sonar data. The following issues where considered:
- Millions of points in a single file can be slow to move around in 3D packages
- A need for visual depth cues, (ie. what points are close and which are further away?)
- Point rendering styles are limited in current specialist survey software
- Colour information is absent from XYZ data but its presence helps with the viewer’s perception of space
- Gaps between points can confuse the viewer’s perception of the object
- Details can be difficult to interpret in still images
The point cloud data produced from the sonar survey requires clean-up as it often contains spurious arc shaped points that are artefacts resulting from the sonar range setting. The range is set in metres that correlate with the estimated distance from the sonar head to the target object. An accurate range setting produces a more detailed image. The clean-up process removes the artefacts and also reduces the overall number of points.
The multi-beam process produces similar datasets to those from laser surface scans in that there are masses of points that are patched together from different passes to complete the 3D image. The key difference is that the target object is immobile and the scanning equipment has to be continually repositioned around the object in what are often extremely unstable conditions.
The next step is to align the separate lines of data to piece together a 3D image of the complete shipwreck. Each pass can reveal additional details that add to the overall picture. In the Royal Oak project, over 60 separate survey passes were collected from the immediate vicinity of the wreck. It is important to understand that the tide may rise or fall during the survey process. It is therefore imperative that the global positioning system is as accurate as possible in order to correctly align each survey pass. Our method operates to an accuracy of approx 1cm, ensuring that any problem of alignment is minimised.
The data is exported from the multi-beam system as a point cloud (ASCII) file, then imported into Autodesk Maya software using our proprietory plug-in. The Royal Oak point cloud contains over 1.4 million points which are converted into a particle object by the plug-in. We chose the Maya software because of its open-ended architecture and MEL scripting language as well as its versatile animation tools. We exploited the functionality and flexibility of Maya to render large particle objects to accurately represent the shipwreck and its surrounding seabed. Maya’s camera tools have also proved invaluable, giving us many options to plan, compose and frame animated shots of the wreck.
The application of digital cinematography was essential to communicate the overall condition of the wreck site to the “lay” viewer. It was paramount that the final images could be interpreted by members of the general public unaccustomed to viewing surfaces represented by point clouds. The potential audience includes policy makers in UK government, Ministry of Defence officials and not least, members of the Royal Oak Survivors Association and their families.
In previous surveys, still image screen grabs were used to illustrate written reports that described the state of the wreck. Unfortunately, still images of point clouds can be difficult to visually decode. Often, points further away from the viewer appear the same size as those that are actually closer, resulting in a confusing visual experience and misinterpretation.
Further confusion is experienced when distant point can be seen between the gaps in the foreground points.
To combat some of these characteristic anomalies, we applied a number of depth cueing techniques to the dataset. Particles were rendered in separate passes determined by their distance from camera, ( ie. particles further away are rendered smaller than those lying closer to the camera). In addition, an occlusion object was modelled with polygons to emulate the shape of the ship’s hull. This was placed inside the particle object and served to prevent particles on the far side of the hull from being rendered between the particles on the nearside.
The sonar sidescan data does not provide any colour information so we added a ramp utility to the Maya plug-in to allow user controlled placement of colour or greyscale ramps to the particle objects. The ramp utility can be rotated to orient the ramp direction allowing the animator to adjust ramps and match any angle of the wreck’s position on the seabed. This proved particularly useful on the Royal Oak as the wreck is upturned with the port side raised above the seabed.
An obvious aid to understanding the 3D image was the use of animated cameras in Maya. The cameras are tracked and panned across the site in a controlled and repeatable manner allowing specific details to be highlighted on the wreck. This movement over the 3D image further reinforces the positional relationships between individual particles, (ie.those particles closer to the camera appear to move more quickly than those further away therefore resolving issues associated with parallax).
Is it Useful?
The Ministry of Defence were particularly interested in the distribution of hot tapping valves on the upturned hull that are used to extract oil from the hull as it slowly works its way towards the surface. The 3D images of the wreck clearly show where these valves are situated and can now be used to assist the planning of future oil reclamation exercises.
Details visible around the bow section and the starboard hull clearly show the torpedo damage inflicted on the wreck. Forensic analysis by Martin Dean (St Andrews University) concludes that further bow damage has been inflicted since the sinking and most likely was caused by the wreck being hit by another vessel. The caps of the main gun turrets can be clearly seen to have dislodged under the pressure of the ship’s weight bearing down on the gun barrels as they dig into the seabed. Many other details have been identified by the 3D visualisation and the resulting High Definition animations are being used to communicate the state of the wreck to those responsible for its management.
The public response to the work has been very positive. Members of the Survivors Association and the Royal British Legion have used the images to commemorate the tragic sinking and loss of life associated with this historic shipwreck.
There have been significant amount of progress made during this project, however there are still an number of problems to be resolved. These include:
- Refining the use of occlusion objects
- The development of efficient depth sorting of the particles during rendering
- Automating point size based on distance to camera
- Applying opacity maps based on distance to camera
The ADUS team have subsequently been approached by the Maritime and Coastguard Agency (MCA) to survey and visualise other environmentally sensitive shipwrecks in UK waters to further inform Government policy on their management. One example is the US liberty ship, SS Richard Montgomery which sank in the River Thames in 1944 laden with munitions.
Although the work so far has been directly linked to highlighting potential environmental issues associated with these sites, there is significant potential in applying the same principles to less sensitive sites. The recreational diving community often relies on written descriptions of the condition of shipwrecks and hand-drawn illustrations to plan expeditions. The ability to access accurate 3D models could help to improve the safety margins for wreck diving enthusiasts. ADUS is currently working on 3D visualisation of the remaining seven shipwrecks of the WWII German High Seas Fleet in Scapa Flow with a view to providing up to date, accurate images of the sites for the recreational diving community. We are also developing a 3D viewer to implement the work in real-time for a more interactive experience.
The 3D visualisation of the Royal Oak was carried out by Chris Rowland at the School of Media Arts and Imaging, University of Dundee.
About the authors:
Steve Flack is Head of Media Arts & Imaging that supports seven communities of practice within a broad grouping of media-focused researchers. Steve is currently engaged in a number of innovative visualisation projects that focus on the convergence of communication and audio-visual technologies.
Contact info: firstname.lastname@example.org
Chris Rowland is a senior lecturer and researcher in the School of Media Arts and Imaging, University of Dundee, Scotland. Chris is a fine art graduate with many years of industrial expereince in animation. His current research interests are centred around the application of 3D animation and visualisation to fields beyond entertainment. Chris is a member of the ADUS research team which specialises in high resolution multibeam sonar surveys and visualisation of environmentally sensitive, historic shipwrecks. He is also an enthusiastic recreational diver.