cais1.gif (153930 bytes)

CAIS Persian Text.gif (34162 bytes)


The Circle of Ancient Iranian Studies

 Persian Section.PNG (9914 bytes)


About CAIS


Daily News

News Archive


CAIS Seminars

Image Library





Contact Us


Facebook-Button.jpg (107165 bytes)


Rag-i bibi: A Sasanian Rock Relief in Afghanistan



Baghlan Map.png (117095 bytes)

rag-i-bibi_sasanian_rock_relief1WM.JPG (89473 bytes)

rag-i-bibi_sasanian_rock_relief2WM.JPG (98147 bytes)

rag-i-bibi_sasanian_rock_relief3WM.JPG (67487 bytes)

rag-i-bibi_sasanian_rock_relief4WM.JPG (89861 bytes)

rag-i-bibi_sasanian_rock_relief5WM.JPG (85174 bytes)

rag-i-bibi_sasanian_rock_relief6WM.JPG (108866 bytes)

rag-i-bibi_sasanian_rock_relief7WM.JPG (86841 bytes)

rag-i-bibi_sasanian_rock_relief8WM.JPG (87369 bytes)

rag-i-bibi_sasanian_rock_relief9WM.JPG (146347 bytes)

  (Click to enlarge)

Abstract: The Rag-i Bibi (lady of the artery), it is a Sasanian rock relief situated one kilometer south of the village Shamarq and 10 kilometres south of Pol-e Chomri, in nowadays Afghanistan. The relief is 4.9 m high and 6.5 meters wide in diameter.


Rag-i Bibi's pre-Islamic name is unknown. However, the modern name is given by Shi’a Muslim locals who believe it is the statue of Fatima, the favourite daughter of Muhammad, the Prophet of Islam, riding a horse.


The relief suffered extensively by the Taliban forces who are Vahabi-Muslims (a Saudi’s sect of Islam) and not only opposed the representation of human and animals forms, but also are anti-Shi’a Muslims. After the invasion of Afghanistan by US-led forces and the fall of the Taliban, the locals asked the authorities to save the relief from further destruction. The knowledge of its existence came to the attention of the archaeologists for the first time in 2002.


The relief shows a kingly figure riding a horse at a gallop in pursuit of a rhinoceros. His figure is about 2.4 meter high, if he would stand. Beside the king, there are three further figures standing behind his horse. The upper end of the relief also shows a series of ill-preserved architecture.


Although the Taliban severely damaged the head of the rider, the crown allows a reliable identification. Certain stylistic details suggest that the rider is the Sasanian Emperor Shapur I. The style of the relief is Sasanian combined with Partho-Gandharan features.




A methodological exercise of a 3D scanning


By: Philippe Martinez



The importance of this relief was recognized as soon as the first images were received at our laboratory. Not only was the rock cut relief unknown, but it was also apparently difficult to reach and an obvious target for the antiquities traffickers. Exceptional measures were therefore required to enable us to preserve the most complete record of the sculpture and its surroundings. Because the relief was so high, it was not possible to be on the same level and to take photographs without distortion : this made it impossible to choose either photogrammetric or to draw from photographs. However, our laboratory had been experimenting with optical triangulation 3D digitizing. This seemed an excellent opportunity to test its relevance in a difficult context where no other means seemed to be appropriate.

However, this approach had technical limits, as the 3D scanners available on the market are not universal tools. They can be divided into groups according to the technology on which they are based. These groups differ in terms of range and resolution. In the case of the Rag i Bibi relief, we had to consider the cliff on which the relief was carved. Such a site would in itself require long range scanning, but we were also looking at an intricate sculpture with high relief and precise details, calling for short range scanning. Because Grenet considered that the mission was urgent and that only a limited task force could be employed, we opted for a Minolta Vivid 700 short range scanner, graciously lent to us by Minolta America, thanks to our active collaboration with the San Francisco based Insight group, or Institute for the Implementation of Digital Techniques. Although already slightly outdated, the Minolta Vivid camera would deliver crisp and detailed scans that should make possible an exact digital copy of the sculpture.

However, this choice necessarily caused problems. The Minolta camera is a short range scanner and that really means short. The best measures are taken at around 0.8-1.2 m. : the actual measure can reach an object at around 3.0 m., but the measurements at that range are of a lesser quality and lower density. That meant that the further from the object the camera was sited, the lesser the density, a serious problem in areas with much sculpted detail, which might be difficult to reach. Another problem was that, since it is based on a red laser, the scanner cannot be used in bright daylight, as the sensor is blinded by the wavelengths of solar light. Work could only, therefore, be undertaken when the cliff was in shadow.

3D scanning by optical triangulation is based on a straightforward technological approach already known in Total Stations. A laser beam or ray emitted by the sensor hits the object which reflects it. This image, returned to the camera, is seen and captured by a CCD device. The distance between the laser and the CCD being fixed and known, the distance to the object is automatically deduced and turned into an actual physical x, y, z spatial measurement corresponding to every pixel of the capturing CCD camera. The camera is also used for capturing images : the colour of the zone being scanned is registered to the measured 3D points, each receiving a specific colour close to the actual colour of the surveyed area under a specific light. Because of the size of the CCD, the images are not of a high precision and quality. However, with a relief that did not preserve much of its original polychromy, this seemed to be of secondary importance.

Advantages of the chosen camera included its compact size, user-friendly operation and its robustness, all of which were central to the success of the mission. It could be carried on board different aircraft, and its robustness was essential when being driven on Afghan roads or carried up cliff paths.

The camera, set on a regular video tripod, chosen for its stability, is manageable by a single operator. It is powered by straight ac-dc electric power or can be operated from a car battery or using a generator. The operator directs the lens towards the surface to be scanned, helped by a small colour monitor set in the back of the camera. He chooses an acceptable level of zoom, considering the distance, before pushing a button to launch a pre-scan to provide an idea of the quality of the measurement thus acquired. The sensor has a limited depth of field of 200 mm., and the user has to define the exact distance at which this depth of field has to be situated from the camera, helped by an autofocus function. If the surface is complex, with different levels and depths, it might be necessary to do many shots of the same area, with different settings. The data were stored directly on a 512 mb Compact Flash card, which enabled us to come on site without a control computer, the data set being saved on a self-powered, portable hard drive.

However, the practical limitations of access to the sculpture almost jeopardized the whole operation. The relief is accessible only via a narrow path, traced by local goatherds and villagers. Workmen carried the gear to the working site, its compactness being welcome. A petrol powered generator was rented and left on site under guard. Though noisy and at some point slightly capricious it enabled us to get the necessary power. The camera was protected through the use of a power regulator. Even with the equipment on site, the topographical limitations initially appeared to be impossible to overcome. The rocky ledge in front of the relief was uneven and only part of the relief was accessible to scanning, when the tripod was almost fully extended, leaving most of the upper parts out of reach to the laser. Furthermore, the cliff was only in shadow during the afternoon, when the remaining glare was still disturbing for the sensor.

It was therefore decided to build a platform in front of the sculpture and to scan at the end of the afternoons, in semi-dusk : we also had to respect the curfew observed by foreign missions. Unfortunately, it was not possible to build a wooden scaffold because of the poor quality of the locally-available wood and the slope of the rock under the relief. A search for metal scaffolding was finally successful when we were directed to an abandoned cement plant built by a Czech consortium. Our Afghan carpenters amazed us with their erection of a sturdy platform directly on the native rock that was stable enough to welcome the visit of the provincial governor. This impressive device enabled us to be level with the rock cut platform and established a base for the whole niche and for the necessary distance to scan most of the reachable surfaces.

The platform was also useful for the more classical inspection of the sculpture and for its photographic survey. Considering the ambient light, it was only possible to scan for about two hours each day before returning to Pul-i Khumri. On the other hand, the time thus freed was used every morning to check the data gathered the previous day, thus validating on a daily basis. During the six remaining days, it was possible to gather more than 1000 scans, in around 15 hours. Conditions were also problematic for the quality of the colour images embedded in the scan files. There was no way to calibrate or control the lighting of the rock face, which changed during each scanning session, from yellowish to dark and blueish. These images were later corrected and homogenized, but it is clear that our final viewing of the model with colours is not faithful enough, although it remains informative.

Our scanning operation resulted in an important archive of more than 1000 scans, although each covered only a limited surface of the monument. The scans had to be reassembled to form a thorough 3D model of the monument. This was done through the use of the GSI Studio software, which we were able to use thanks to our relationship with Insight. The files from the camera are text files containing the information for the three x,y,z measurements of the points and other values such as the reflection of laser beam on this point and the digital colour image linked to the 3D point cloud. When imported in GSI, these data sets are turned into 3D meshes linked to a colour information for each of its vertices. The actual registering of the different scans is a complex operation. It could have been helped by the positioning in the scenes of specific targets, but due to our limited access to the relief we had to choose a more straightforward approach that only considered a limited amount of shared surfaces between each scan. As we could not use targets to help registration, we used the tools offered by GSI Studio, i.e. algorithms based on an approach called ICP (Iterative Closest Point), which considers the topology of each of the surfaces (or point clouds) and by considering each point and its closest neighbours manages to match the corresponding topologies and mix the scans according to these matches.

In reality, the process is straightforward. The user loads two scans. On each of them, points that should be the same are pointed and coupled, three pairs being the least required to help the software decide the corresponding orientation of the scans (as the position of the scanner has changed between each scan, and thus the orientation of the scan itself). Once this preliminary set has been made, the software launches its topological correlation and iterates the process until it reaches a level of error that seems acceptable. Each scan remains independent until the user is happy with the matching offered by the software. Every time one match is finished, it is possible to load a new scan and to push the process forward. Once the process is finished for a specific surface, it is possible to launch another matching process that is known as global. This check on the optimization of the topological matching when all the scans present in this 3D surface are considered, instead of the single pairs considered in the ongoing process. Needless to say, this final global approach is complex and greedy in terms of processing power.

Following this approach, each scanning session considered a limited area of the sculpture to keep the topological relationships as close as possible. We ended with six different scan groups sharing part of their topology on certain edges. At this stage, every one of these registered groups was already heavy in terms of RAM access. To move to the final fitting we had to simplify these groups of scans, considering mainly the repeated information in the overlapping areas, and turning these complex clusters into single surfaces that could be joined through a renewed registering process.

Since our portable computer was powerful enough to handle these tasks, it was possible to make a preliminary registration of these groups of scans on a daily basis, thus knowing which areas had not yet been scanned or had to be rescanned when the resulting point cloud overlap or quality was not considered sufficient. We were thus able to leave Pul-i Khumri with a data set that had been checked for its consistency and quality. However we have to note that even with our scaffolding it was not possible to reach the higher parts of the relief in a satisfactory way and that most of the horizontal surfaces set above the actual level of the eye of the scanners could not be scanned. We thus have a final 3D model (Fig. 12, bottom) that is as complete as was possible in our working conditions but with holes that should have been filled by scans done from a higher levels. These holes should be patched in the future by complementary scans or by hand modelling in a 3D commercial package. However in conclusion about 90 % of the surface of the relief was scanned in 3D during six days on site, with a scanning access time of only two full working days. The resulting model is a unique tool to preserve the topological information concerning the surfaces of the sculptured niche. It gave us unprecedented insight into the composition of the sculpture and enabled us to consider aspects that it was not possible to feel when in close contact with the monument. We also have to consider that the complete model can enable us to print in 3D an exact copy of the relief at 1:1 scale for a museum exhibition or at smaller scale for its dissemination and study. The use of other digital formats such as octrees should enable its easy dissemination over the web, using a free viewer running on any computer around the world.

While it has to be stated that the scanning operation was made more complex and difficult by the location of the Rag-i Bibi relief and its three-dimensional reality, it should be noted that this technology could be applied with advantage to other flatter reliefs, such as the monumental rock reliefs of the Sasanian rulers. These might soon find other amazing parallels in the mountains of Afghanistan.



Top of Page



Source/Extracted From: Archéologies d’Orient et d’Occident et textes anciens (AOROC)


Please note: CAIS has the privilege to publish the above article, originating from the above-mentioned source for educational purposes only (Read Only). This article has been published by CAIS in accordance with the author(s) / source's copyright-policy -- therefore, the ownership and copyright of this page-file remains with the author(s) / sourceTo be used for any other purposes, you must directly obtain a  written permission from the copyright owner concerned. 


(For more information, please refer to CAIS Copyright Policy).



<META name="verify-v1" content="Kb4N15t1UVWj7aEXtMAMsR2vpb1WAyOpb5tfwsdcn1w=" />

my_Iran.jpg (13682 bytes)

"History is the Light on the Path to Future"


Persian_NOT_Farsi_by_Shapour_Suren-Pahlav_3D2.gif (177309 bytes)


Encyclopaedia Iranica

BIPS.jpg (15695 bytes)

The British Institute of Persian Studies

"Persepolis Reconstructed"

Persepolis_reconstructed2.jpg (36944 bytes)


The British Museum

The Royal

Asiatic Society

Persian_Gulf_Facebook.jpg (1935028 bytes)

The Persian Gulf

Facebook Page

Please use your "Back" button (top left) to return to the previous page

Copyright © 1998-2015 The Circle of Ancient Iranian Studies (CAIS)