GIS technical African Cultural Heritage Sites and Landscapes Project by Heinz Rüther, Roshan Bhurtha, Christoph Held, Ralph SchrÖder and Stephen Wessels, University of

In 2005 PositionIT published an article on the African Cultural Heritage Sites and Landscapes Project. This report is a follow-up on the previous publication, it describes the evolution of the project and some of the many challenges which had and still have to be faced and overcome.

t is the objective of the African Cultural Heritage Sites and ILandscapes Project to preserve ’s rich architecture and cultural landscapes in digital form through detailed and accurate spatial documentation using state-of-the- art spatial data acquisition methods and instruments (see Fig. 2 for list of documented sites). The project is executed by the Zamani Research group at the University of Cape Town.

Since the start of the project in 2004 over 20 important African heritage sites have been digitally documented by means of laser-scanning, photogrammetry, photography, GIS, GPS and conventional survey techniques. The Fig. 1: Team picture, 2009, Musawwarat-es-Sufra , from left to right: Ralph SchrÖder, generates a wide variety of 2D and Christoph Held, Roshan Bhurtha, Heinz Rüther and Stephen Wessels. 3D data formats to ensure a holistic documentation of African heritage. These include a GIS of each site and architects and historians, as well as team is scanning 50 scans in half a its surrounds, highly detailed 3D for conservation, restoration, and as a day and a "team record" of 140 scans models at various levels of resolution, record for future generations. in one day with a single scanner was as well as line-models of important achieved during a recent project in Data structures, ground and floor plans, in . Scanner sections and elevations, full-dome 3D models are an essential component technology and methodology has HDR panoramas for interactive tours, of the project, and their creation is the changed significantly and currently videos, photogrammetric datasets and most time-consuming and demanding week-long field campaigns typically contextual imagery. of all data types generated in the produce 500 – 1000 scans per site project. Five years ago, a dataset of using two scanners. The US partner of Zamani, the New York 10- to 20-million laser-scanned surface based JStor/ database enriches points for a single building seemed like The dramatic increase in scan rates is the spatial data with a wide range of a number out of an astronomy book. due to the significantly faster scanning site related digitised documents such Now, the advent of fast phase-based of phase based scanners in comparison 1 as books, travel reports, excavation and laser scanners makes this figure sound to time-of-flight instruments. For condition reports, and publishes the data meagre and point clouds produced for example, a scan which took two to via the www.JStor.org and some sites approach data volumes of three hours with the time-of-flight www.aluka.org websites. The initiative’s several billions. In the 2005 article, scanner used in early fieldwork, can aim is to create a digital record a scanning rate of about 50 scans now be completed with a phase-based of structures and landscapes for over five days was reported for a instrument in two minutes for the education, research by archaeologists, typical structure. Now the Zamani same resolution. The trade-in for the

1 Recent developments in time-of-flight scanners show a significant improvement in scanning speed which now approaches that of phase-based scanners.

PositionIT – July 2010 33 GIS technical

increased scan speed is a reduced scan range. This increase in scanning efficiency has had an impact on the team’s approach. Being able to complete more scanner setups not only made it possible to scan entire sites where previously only selected structures could be covered, but it is also possible to scan in much greater detail, as fill-in scans for small areas can be completed in a few minutes. Thus the number of scan holes in the final model could be reduced significantly. Previously, such additional setups would have been considered too resource-intensive. That there is a price to pay for this improvement becomes obvious when facing the exponentially increased processing time required when registering the significantly larger number of individual scans.

The team’s policy is to register the scans by surface matching rather than via targets. Placing of targets would make the fieldwork too complicated and slow. Also, the religious nature of many of the documented sites prohibits the placing of targets while the fragile structures of historical sites, their complexity and their extent often makes it difficult to reach positions where targets would best be positioned. An additional unexpected problem can arise from often present curious children who find targets an attractive Fig. 2: List of documented sites. decoration and much more appealing when worn around the neck or glued to a forehead than attached to a wall.

Data processing without targets is undoubtedly slower than a target- supported approach but advances in registration software together with the experience gained with more than 5000 scans have enabled the team to speed up the registration process. While the registration of about 40 scans could take one to two weeks in the beginning of the project, the alignment of the scans for the documentation of Lalibela in (see Figs. 4 and 5), consisting of about 1200 scans, was completed in less than 12 weeks. Recording the position of each scanner set-up with RTK-GPS and adopting a consistent orientation of the scanner helped to support and speed up the manual alignment process. Fig. 3: Scanning the Lion Temple, Musawwarat-es-Sufra, Sudan. Within the past five years, more than 6000 laser-scans have been recorded, tens of thousands of times the shutters and existing topographic maps. During terabytes, resulting in data storage, of several SLRs and compact cameras this time, the five members of the back-up and presentation problems. have been released and some square Zamani team, with the assistance of It is a most frustrating experience for kilometres of terrain surrounding more than 25 international interns the Zamani group to produce high sites were surveyed or digitised from on temporary assignment to the resolution 3D models in the knowledge satellite images, aerial photographs project, have processed data of several that present desktop computers will

34 PositionIT – July 2010 GIS technical

only allow the display of ten or less percent of the available dataset. While this is disappointing, there is a consolation in the expectation of future availability of significantly faster and more powerful computers and software, thus fulfilling one of the objectives of the principal investigator of the Zamani project who initiated the project with a view to creating a record of African heritage for future generations.

Texturing of 3D models is a complex Fig. 4: Rendered 3D models of Biet Emmanuel, Lalibela, Ethiopia and the of Shela, and, as yet, not fully resolved aspect on Lamu Island, derived from laser-scan point clouds. of laser scanning (see Fig. 6). For a number of reasons which cannot be discussed here, the project adheres to a philosophy which favours post scan colouring as opposed to using colour information acquired by the scanner. Landscapes, rock shelters and some minor structures have already been draped with photography, but, although extensive photography with calibrated cameras is taken at each site, not all of the documented buildings have been textured yet. Covering an entire structure with high resolution quality photographs while keeping the same lighting conditions throughout is very difficult, if not practically impossible. Unlike 2D-orthophoto software, 3D-texturing software does not yet cope well with the problem of blending diverse images within a scene, nor can large image numbers be accommodated Fig. 5: A collage of 3D models of the rock-hewn churches in Lalibela, Ethiopia, and the entire documented site. in a single run. It is unrealistic to expect that every single surface and aspect of detail on a complex structure can be covered with photography and that it is therefore highly unlikely that perfect, hole-free texturing of a structure can ever be achieved. There is clearly room for improvement in the area of texturing and researchers and academics worldwide are presently involved in developments in this field. The Zamani group has developed its own in-house texturing software which is in a testing stage.

Changes in the documentation approach

Based on five years of experience with data acquisition and processing and as a result of technological advances, the team has modified field methods, changed equipment, developed an optimised production pipeline and experimented with different output formats. This process is ongoing and unlikely to be finalised for some time, Fig. 6: Textured 3D model of a rockshelter in the Cederberg. if ever.

36 PositionIT – July 2010 GIS technical

Examples of changes in field methods are the above mentioned uniform orientation of the scanner at all setups and the on-the-fly determination of scanner positions and height with RTK or conventional survey, both of which significantly improve the registration process.

In the beginning of the project only buildings were scanned. Now, if possible, entire landscapes have been recorded, such as the Valley of the Queens in (see Fig. 7) or the Rock churches of Lalibela and Mapungupwe hill.

Wherever more than one scanner is available (and here the team relies heavily on the support of suppliers of laser scanners) a combination of scanners is employed, where long and medium range scanners are used to produce a framework for a first rough registration followed by large numbers of short range scans to fill in detail and Fig. 7: Drainage lines, Valley of the Queens, Luxor, , created from terrestrial laser scanning of the entire valley. to refine the registration. While the framework approach meant a significant improvement to the process, the registration of hundreds of scans into a single model remains complex and time consuming; it requires experience and lots of patience and is by no means trivial.

A modification to the GIS procedure was the digitising of the 3D surface models of buildings and sites as generated by terrestrial laser scanning. The project has no access to airborne scanners and the limited project funds do not allow commercially contracted scanning of the site environments. It was therefore decided to create GIS layers by levelling the 3D model in an orthogonal projection display mode and digitise Fig. 8: Leica HDS3000 scanning the San Sebastian Fortress, Island, Mozambique. features on screen off the model.

There were also changes in the photographic activities in the field. First Documented sites and other manmade structures, using medium and long range scanners photographic panoramas were captured To date, over 20 sites with some 70 in order to create a 3D terrain and with handheld cameras and 28 mm structures in ten African countries have drainage model to be used for the lenses, as extreme wide angle or fisheye been documented. Although the principal prediction of rain water flow for the lenses were at that stage considered objectives of the database are the design of mitigation measures to protect unsuitable due to their considerable creation of information for education and tombs, older than three millennia from image distortion. This approach research and to create a record for the irreversible flood damage (see Fig. 7). required 64 images for an all round future, the acquired data also serve as panorama. Today, a 10 mm lens used in an important component of conservation, One of the most extensive field projects conjunction with off-the-shelf panorama restoration and monitoring projects. of the Zamani team are the rock hewn software makes it possible to generate An example for such an application is the Ethiopian Orthodox churches of Lalibela full dome panoramas of excellent quality terrestrial laser scanning of the Valley (see Figs. 4 and 5). High resolution 3D out of only seven images. To overcome of the Queens in Egypt completed by models generated from the scans of the the problems of often extreme the Zamani team in 2007 for the Getty churches are being used for restoration differences in light conditions in a scene, Conservation Institute, Los Angeles. This in a UNESCO/ World Monuments HDR (high dynamic range photography) mission entailed scanning 1 km2 of initiative for condition reports and is employed for most panoramas. valley terrain, including tomb entrances planning of restoration works.

PositionIT – July 2010 37 GIS technical

The San Sebastian Fortress on Ilha de situations, only a delicately balanced Besides the full dataset being made Mozambique is a further example of combination of polite patience and available to scientists and researchers at an application of data generated by dogged determination will convince www.JStor.org and www.aluka.org, an the Zamani team (see Fig. 8). Here customs officials that "laser scanners" in-house website was created the generated model will be used for are not military equipment and (www.zamaniproject.org). Here "documenting a site" does not refer to restoration work by UNESCO. examples of low resolution models are the production of a lucrative advertising presented together with panorama In , ten important video. Repeatedly the team had to tours, site photography, videos and a shelters in the Cederberg (see Fig. 6) wait for days for the release of the and in the mountains equipment from customs holding areas. travel blogs. have been documented, scanned These waiting periods are filled with The project is not-for-profit and funded and textured. Also scanned was the frequent visits to customs offices for by the Andrew W. Mellon foundation near Kimberly and lengthy negotiations and candid checks until 2011. A public exhibition of the the 3D model has been used to design to see if the valuable laser scanners work is planned for this year to garner walkways and calculate overburden. The are still where they should be. Often awareness and to attract new funding, map and table in Fig. 2 show the sites similar difficulties are encountered when to allow the project to continue. Details documented to date. exporting the equipment again. will be available on our website. And then there is the transfer of delicate Field campaigns equipment over potholed roads and Acknowledgements Field campaigns, in spite of their through long dusty stretches of desert. technical, logistical and personal For one specifically bad dirt road it was This project would not have been challenges, are some of the highlights necessary to purchase mattresses in a possible without the generous grants of of the project. The budget is always small village store as a protective cover the Andrew W. Mellon Foundation and on a shoe-string as every cent saved for the laser scanner. During the past the unwavering support of Dr. Stuart on travelling contributes to the years, the team traveled to sites on Saunders, the foundation's senior documentation of further sites once camels and mules, in traditional dhows advisor for the South Africa programme. funds run out. and motor boats as well as on hired Critical to the publication of 4x4s, local busses and village taxies. the data are JStore and Aluka, New Typically team members become travel Working on heritage sites requires York, and specifically Rahim Rajan, the agents, finding the cheapest flights, sensitivity to local customs and values content development manager of Aluka. organising permits for equipment and makes every field campaign an transport, temporary imports, visas, as The project relies on the advice of three experience beyond the mere struggle well as travel insurance. In view of the experts in archaeology, anthropology with complex survey situations and high cost of excess luggage, personal and African history: Prof. Susan Keech difficult logistics. In discussions with gear is reduced to an absolute minimum Macintosh, Dr. George Abungu, and Imams, Orthodox Church leaders, Prof. Martin Hall. The University of Cape to create space for heavy equipment. Ashanti priest and village chiefs, do’s Town and the Geomatics division host Complications generally arise when and don’ts are explored and local staff checking in the diverse technical is hired to accompany the team. When the project. Francois Stroh from Leica equipment which has already led to working on the Great Mosque in Djenne Geosystems has assisted extensively absurd situations where batteries were special permission had to be obtained through loans of laser scanning not allowed in the hold but permitted on in lengthy negotiations to gain access equipment and Trevor Venter and the same flight in the cabin. And then to the Great Mosque which is strictly off Sean Dane, Optron Geomatics, have there is always the concern that one limits for tourists and non-Muslims and provided Trimble scanners for projects might forget a single cable, an omission in Kumasi a sacrifice had to made at a in Tanzania, Mozambique and South which, if discovered in a remote desert small ceremony to be allowed to work Africa. The Visual Computing Lab of the location, might well put an end to the inside the Ashanti Shrine. ISTI-CNR, Pisa, Italy and InnovMetric entire campaign. Critical inspection by But these difficulties, which are familiar Software Inc, Canada provided software. suspicious border officials of equipment to any geomatician working in remote The Zamani team, consisting of the labelled "laser" can take hours and in areas, fade away in the presence of some cases even days. principal investigator of the project, Africa’s magnificent heritage sites and Prof. Heinz Rüther and the four scientific Traveling to remote locations in Africa landscapes. officers Roshan Bhurtha, Christoph Held, with all its fascinating thrills is not Ralph SchrÖder and Stephen Wessels, without trials and tribulations. Although The future wish to express their heartfelt gratitude for each field campaign every effort Five years of documentation have and appreciation to all the above is made to obtain in advance all enabled the project to complete individuals and institutions. necessary documents, ranging from some 70 important structures. There permissions for temporary equipment are numerous others from different Anybody interested in contributing to import to using a video camera on historical periods, cultures and regions the project in any way can contact the an archaeological site to research to be documented and the project is team via [email protected]. permits to work on protected sites, expected to continue for some time these documents often prove not worth to come. For the near future, sites in Contact Prof. Heinz Rüther, the paper they are written on when , Sudan, Mauretania, Tanzania University of Cape Town, produced at customs or presented as well as the Castle of Good Hope in Tel 021 650-2327, to officials on heritage sites. In such Cape Town are scheduled. [email protected]

38 PositionIT – July 2010