Dokumentacja najważniejszych petroglifów za pomocą skanowania światłem strukturalnym oraz analiza najbardziej zniszczonych petroglifów metodami vPTM i vRTI

2020

2(62)

DOI: 10.37190/arc200207

Introduction

Digital image processing techniques have been de-

veloping since the beginning of this century, resulting in

more and more uses for them. For documenting and ana-

lysing rock art heritage, two groups of techniques deserve

special attention. Both of them help to identify items that

are not visible to the naked eye.

The rst group are applications based on the transfor-

mation of a single digital image. Among them, ImageJ [1]

deserves special attention. This is an open-source Java ap-

plication commonly used by many rock art researchers. It

allows simple geometric transformations to be made (for

example, cropping, scaling, resizing and rotating, ipping

vertically or horizontally) as well as much more advanced

image enhancement (for example, smoothing, sharpening,

edge detection, median ltering, thresholding on grayscale

Jacek Kościuk*, Małgorzata Telesińska**, Maciej Nisztuk***,

Marta Pakowska****

Documentation of the most important petroglyphs

by structured light scanning and analysis

of the most damaged petroglyphs by vPTM and vRTI methods

Dokumentacja najważniejszych petroglifów

za pomocą skanowania światłem strukturalnym

oraz analiza najbardziej zniszczonych petroglifów metodami vPTM i vRTI

* ORCID: 0000-0003-0623-8071. Faculty of Architecture,

Wro cław University of Science and Technology, e-mail: jacek.kosciuk@

pwr.edu.pl

** ORCID: 0000-0002-1100-642X. Faculty of Architecture,

Wrocław University of Science and Technology.

*** ORCID: 0000-0001-6520-5128. Faculty of Architecture,

Wrocław University of Science and Technology.

**** ORCID: 0000-0003-0857-2700. Faculty of Architecture,

Wroc ław University of Science and Technology.

and RGB images, and adjusting the brightness and con-

trast of 8, 16 and 32-bit images) and colour processing (for

example, splitting 32-bit colour images into RGB or HSV

components, merging 8-bit components into a colour im-

age, converting RGB images to 8-bit indexed colour, and

applying pseudo-colour palettes to grayscale images) [2].

From the more than 500 more advanced plugins avail-

able for ImageJ, DStretch [3] stands out. It is particularly

useful for documenting and analysing rock art. To rein-

force hue dierences of neighbouring pixels, Dstretch

algorithms use decorrelation stretch in the image colour

space and manipulate the hue and saturation of the image

[4]. ImageJ and the DStretch plugin are used extensively

in the documentation and analysis of rock paintings (par-

ticularly pictographs) and shallow rock engravings.

The second group of applications process series of im-

ages of the same object taken in dierent lighting condi-

tions. The rst algorithm of this type dedicated to docu ment

ing and analysing cultural heritage was surface reectance

transformation, developed in 2000 at Hew lett-Packard

Laboratories [5]. One year later, the same team from

Hewlett-Packard Laboratories developed another method

for picture enhancement – polynomial texture map ping

(PTM) [6]. Since then, surface reectance transformation

was further developed by Cultural Heritage Imaging and

today is commonly known as reectance transformation

imaging (RTI) [7]. In the traditional approach, as an

input,

44 Jacek Kościuk, Małgorzata Telesińska, Maciej Nisztuk, Marta Pakowska

RTI needs photos taken on the eld with the help of a tri-

pod, strong portable light source, two black balls that lat-

er on help to calculate angles of illumination, measuring

tape, and string. Usually, 40–90 images are collected, each

with dierent vertical and horizontal angles of illumina-

tion. This method is mainly used when the distinguish-

ing feature of an artefact is not the colour itself but rather

its spatial shape, or both. In this last case, an interesting

implementation of the method described here is the use

of RTI on images enhanced with DStretch algorithms [8].

Purpose of research

The presented research is part of the larger project

“Architectural examination and complex documentation

of Samaipata (Fuerte de Samaipata/Bolivia) site from the

World Heritage List”. One of the main aims of this pro-

ject was to document and analyse vanishing examples of

pre-Inca and Inca petroglyphs

Due to the rapid erosion of the rock at Samaipata, some

petroglyphs, still well recognisable two hundred years ago

[9], are now entirely unidentiable, and in some cases,

it is dicult to nd any traces of them. Thus, documen-

tation of the current condition of the rock itself and the

most important petroglyphs was an urgent priority. The

second goal of the research was to recognise and interpret

what is depicted on each petroglyph (as far as their current

Cf. J. Kościuk, M. Ziółkowski, B. Ćmielewski, D. Ulloa Vi-

daurre, Samaipata project – aim of the research, methodology, and meth-

ods of documentation, in the same issue of “Architectus”.

Fig. 2. Screenshot from the Artec Studio user interface

(elaborated by M. Pakowska)

Fig. 1. Schematic idea of structured light scanning

(d – the distance between the camera and the structured light source;

β – the angle between the axis of the camera and the axis

of the structured light source) (elaborated by M. Pakowska)

stage of preservation permitted). We were also aware that

for some of the most damaged petroglyphs, this could be

the last opportunity to document their quickly disappear-

ing traces. Therefore, the proper selection of equipment,

software, and analytical methods was fundamental to the

success of the entire project.

Methods, equipment, software,

and general workflow

Of the most advanced documentation methods availa-

ble to us, structured light scanning and short-range digital

photogrammetry were chosen as the most appropriate for

detailed documentation of the petroglyphs.

Structured light scanning

One of the methods selected to collect data on the shape

of the petroglyphs is structured light 3D scanning (Fig. 1).

Due to the portability of the equipment, as well as its

technical parameters, an Artec Eva scanner was used for

work. With an Artec Eva, the whole process of scanning

is controlled by a computer, or more precisely, the Artec

Studio software. It not only supports the scanner, but also

aids the post-production of data. During the work, the

Artec Eva scanner emits white strips of light that are de-

formed on the surface of the object according to its shape.

The camera records them on the matrix. Knowing the mu-

tual position of the light emitter and the camera and map-

ping the strips on the matrix makes it possible to calculate

the distance of the examined point from the scanner. This

results in getting the x, y, and z coordinates of each of

the points. This is therefore a triangulation method using

white light.

According to the manufacturer’s technical specica-

tions, the Artec Eva scanner oers 0.5 mm resolution of

scanning with 0.1 mm accuracy [10]. The recommended

optimal depth range for this scanner of between 40 and

100 cm was observed while registering to the common co-

ordinate system. The registration algorithm based on com-

paring both, the geometry of the surface and the texture.

It must be noted that this scanner, like virtually all

structured light scanners, is not particularly suitable for

 Documentationofthemostimportantpetroglyphs / Dokumentacjanajważniejszychpetroglifów 45

Preliminary observations on 3D models

The red-brown colour of the Samaipata rock makes it

dicult to observe the delicate shadows cast by shallow,

strongly eroded petroglyphs. This applies to observations

in the eld as well as on textured virtual 3D models. The

numerous lichens, mosses, and algae covering the rock

cause additional diculty. They create multi-coloured

mosaics against which faint traces of petroglyphs and

even inscriptions carved by modern tourists are practical-

ly invisible. Observation of virtual 3D models in grayscale

with the texture preview turned o dramatically improved

the readability of all these traces (Fig. 3).

It is worth noting that similar results were obtained

from the observation of a much less detailed virtual 3D

model of the whole rock delivered from 3D terrestrial

scanning

. Additionally, in this case, turning o the tex-

ture preview showed many traces that were not visible

either in the eld or on the textured 3D model (Fig. 4).

The use of hill shading algorithms known from many

practical applications in remote sensing [12]–[14] turned

out to be particularly helpful in both cases. Further, previ-

ously invisible details were revealed when the lighting an-

gle changed. This gave us the idea of using PTM and RTI

techniques for detailed analysis of the most eroded petro-

glyphs. In our case, however, these techniques were used

on virtual models and not on real objects. Therefore, un-

like traditional applications, we will refer to them as vir-

tual PTM (vPTM) and virtual RTI (vRTI).

Using vPTM and vRTI

While traditional PTM and RTI techniques are rela-

tively easy to use in laboratory conditions where special

rigging provides lighting at predened angles, their use

outdoors for large objects poses many problems (Fig. 5).

One of the basic issues is taking a sucient number

of photos with the camera in a xed position and with

Cf. J. Kościuk, B. Ćmielewski, M. Telesińska, A. Kubicka, 3D

terrestrial laser scanning of El Fuerte de Samaipata, in the same issue

of “Architectus”.

outdoor use. Direct sunlight falling on the scanned object

competes too much with the light strips emitted by the

scanner. Scanning in such conditions is practically impos-

sible. This problem was overcome with the help of a big

tarpaulin sheet that shaded each petroglyph from the sun.

It also protected the scanner against gusts of wind par-

ticularly troublesome at noon. We noticed that gusts of

wind in the afternoon signicantly hindered scanning. Al-

though this nding has been challenged by the manufac-

turer’s representatives, it seems to us that diering density

streams of hot and cold air disturb the strips of light pro-

jected by the scanner.

The scan data collected in the eld were further pro-

cessed with dedicated software – Artec Studio Professional

(version 11, build 11.2.2.16) (Fig. 2). The nal results

were 3D mesh models with an average density of 7 trian-

gles/cm

. Documentation of four petroglyphs was created

this way. The size of the models when stored as *.obj les

ranged from 1500 to 3500 MB. The maximum resolution

of the accompanying texture les was 16 384 × 16 384

pixels.

Close-range photogrammetry

Close-range digital photogrammetry was used for the

largest petroglyphs. Digital photos were collected with

a Sony Alpha STL-A65 camera with an APS-C sensor

matrix of 24 megapixels. The Sony 18-55mm f/3.5-5.6

standard lens typically sold with a Sony Alpha camera

was attached. The camera was set to 400 ASA in aper-

ture priority mode, while the aperture itself was perma-

nently xed to f = 11. The focal length of the lens was set

to 18 mm and locked. We determined that these settings

were optimal when using this camera for digital photo-

grammetry. The problem of scaling individual photogram-

metric projects was solved by placing a few measuring

sticks (each 2 m long) around petroglyphs. Depending on

the size of the petroglyph, 100 to 500 photos were taken,

which were then processed by the AgiSoft PhotoScan Pro-

fessional software (version 1.4.4, build 6848) [11]. As in

the case of structured light scanning, the nal results were

3D mesh models stored together with the corresponding

texture as *.obj les.

Fig. 3. Impact of multi-coloured

texture on surface detail visibility:

A – textured 3D model;

B – same fragment

after turning texture off

(elaborated by J. Kościuk)

46 Jacek Kościuk, Małgorzata Telesińska, Maciej Nisztuk, Marta Pakowska

dierent angles of lighting. The horizontal and vertical

angles at which the object is photographed, as well as

the angles of illuminating light, must be determined quite

precisely. Hence, when using the traditional PTM or RTI

method in the eld, sophisticated procedures (Fig. 5) are

used, with inclinometers, strings, and black, shiny spheres

allowing the angle of incidence of light to be determined

at the post-processing phase. In addition, canvas sheets

are used to eliminate direct sunlight competing with arti-

cial directional lighting. The work is tedious, time-con-

suming, and requires a lot of equipment to be brought to

the site. However, since we have virtual 3D models, why

not use them for this purpose instead of taking pictures in

the eld? Computer visualisations of virtual 3D models

in pre-set lighting conditions and viewing angles evenly

covering the entire hemisphere around the object may be

equivalent to the lighting arrays used in laboratory condi-

tions (Figs. 6, 7).

Fig. 4. Impact of multi-coloured texture on surface detail visibility: A – textured 3D model; B – same fragment after turning texture off

(elaborated by J. Kościuk)

Fig. 5. Examples of using

the RTI technique

to document larger artefacts

(sources:

A – https://ipch.yale.edu/news/

rti-training-ipch-digitization-lab;

B, C – https://raan.hypotheses.

org/1326)

[accessed: 17.10.2018]

Virtual rig and computer visualisation of 3D models

The laboratory rules of creating photographs for PTM

were applied to a computer program, and the photos were

substituted with renderings prepared in Cinema 4D [15]

software (version R15.008). The scanned and compiled

model was moved to a virtual rig studio with 54 prede-

ned positions of light equally spread over the virtual

hemisphere at 15, 30, 45, 60, and 75 vertical degrees. On

top of the dome, a virtual camera was placed (Fig. 7).

For particularly big and detailed 3D models, the mesh

grid was decimated to 1 million triangles and 600 thou-

sand points. The mesh le size was optimised for the ren-

dering time of one frame as well as for sucient resolu-

tion of the resulting le. The physical dimensions of the

model had no relevance in the virtual environment.

The software provides the possibility to align the

lights to the normal plane of the object. The selected

 Documentationofthemostimportantpetroglyphs / Dokumentacjanajważniejszychpetroglifów 47

type of lighting was a surface emitting parallel rays of

light. This eliminated the problem of a direct spotlight,

which can disturb the shape of the shading. Additional-

ly, a second diuse light was also used to illuminate the

scene evenly from all directions, assuring that details in

the deepest shadows were lit up. This issue was also dis-

cussed and solved during consultations with Piotr Wit-

kowski, who had pre viously used the traditional PTM and

RTI methods [16].

Adequate material parameters were also selected to

simulate light absorption and reection properties of the

sandstone rock. The chosen settings were neutral grey

(RGB: 232, 232, 232, 232), 80% diuse, and rough sur-

face. The environment lighting was switched o, and the

Phong rendering model was chosen. The aperture size was

set to f = 8, and exposure time for the virtual camera was

constant and equal to 1/30 sec. Thanks to these settings,

the problem of overexposure was eliminated. For ex-

tremely eroded petroglyphs, to highlight small dierences

on almost at surfaces, the 3D models were scaled twice

according to the vertical axis.

Near each rendered petroglyph 3D model, at least one

shiny black ball was placed. Its position did not change

for each of 54 rendered images, and later, one black ball

was used in the RTIBuilder software for computing the

horizontal and vertical angles of the light source. The user

marks the location of the shiny black balls, and the soft-

ware calculates light angles for each respective image.

All the visualisation calculations were executed on

a desktop computer equipped with an Intel

Core™ i7-

4790 CPU and 16 GB RAM, running a 64-bit operating

system. The graphics card was 2x NVIDIA GeForce

GTX 970 4 GB. The time to calculate rendering for each

scene depended on the lighting orientation and was vari-

able. The maximum timing for one frame was 30 sec for

an image 2000 × 2000 pixels at 200 dpi resolution. The total

time to generate 54 renderings of one 3D model was ap-

proximately 20 min. The rendering process was automated

Fig. 6. RTI lighting array used in the laboratory

(source: http://culturalheritageimaging.org/

What_We_Offer/Gear/Lighting_Array/) [accessed: 17.10.2018]

Fig. 7. Schematic idea of a “virtual rig”

(elaborated by M. Budlewska)

by applying the Winparrot (version 2.1.3) macro script,

which took care of automatically adjusting the light po-

sition for each of the 54 images, queuing all tasks, and

saving nal results.

Processing of images from the “virtual rig”

The series of renderings delivered from our “virtual rig”

were processed with RTIBuilder [17] (version 2.0.2). Based

on the input images from our virtual rig, the software cal-

culated data for light angles and stored it internally.

All the renderings were processed and combined into

a single le containing the mathematical model of the in-

spected surface. The RTIViewer (version 1.1) attached in

the software bundle [18] allows the object to be interac-

tively re-lit from any angle, permitting accurate examina-

tion of all its parts. In addition, it allows for various dis-

play enhancement (for PTM les: Diuse Gain, Specular

Enhancement, Normal Unsharp Masking, Image Unsharp

Masking, Luminance Unsharp Masking, Coecient Un-

sharp Masking, Static Multi Light, Dynamic Multi Light,

Normals Visualisation; for RTI les: Specular Enhance-

ment, Normals Visualisation) enabling details invisible in

the standard viewing mode to be revealed (Fig. 8). Thanks

to these various lighting and preview options, it was pos-

sible to examine the object in detail, especially with the

PTM method.

In order to use all the possibilities oered by both vis-

ualisation techniques, an RTIBuilder-compatible PTM

Fitter [19] plug-in from Hewlett-Packard was also used.

The PTM Fitter extends the functionality of the RTI soft-

ware and is operated via the host-software interface.

When the CGI input data are ready, they can be pro-

cessed into RTI or PTM images via the tting action.

Fitting is the process of matching together the lighting

data from input photos in order to produce a nal RTI/

PTM image. Each image type has a particular tting meth-

od. Workow with the RTIBuilder is straightforward.

48 Jacek Kościuk, Małgorzata Telesińska, Maciej Nisztuk, Marta Pakowska

Using a wizard-like interface, the software guides the

user through the following steps:

– Selecting input images from the folder data source.

The les should be named sequentially using only alphanu-

meric text without spaces and should be organised accord-

ing to the RTIBuilder software manual [20] suggestions;

– Choosing the RTI or PTM tting method;

– Pre-processing the input data set. The user selects

the black glossy ball model location on the images;

– When the computing of light source directions is

completed, the software is ready to process the series of

images and produce the nal RTI or PTM le.

Although the workow is relatively simple, it is neces-

sary to observe some basic rules regarding the preparation

of the les and the structure of the folders. The prepara-

tion process is explained in the software user manual [20].

The RTIBuilder provides two computational methods

of images processing: RTI and PTM. In order to create

PTM images, we used the PTM Fitter method from Hew-

lett-Packard. For RTI images, the Hemispheric Harmonics

(HSH) Fitter method was chosen. Each method has sev-

eral settings responsible for the accuracy of the computed

mathematical model. For PTM, two options are available:

LRGB (medium accuracy) and RGB (high accuracy). For

the RTI method, lighting mapping accuracy should be

set up – a higher numerical value of lighting mapping ac-

curacy creates a more accurate model at the cost of big-

ger les.

Based on experience gathered during this project, the

PTM method gives more options for displaying the nal

model le. According to RTIBuilder software developers

[21], RTI produces a more accurate model of the exam-

ined object at the cost of fewer model display options.

In the case of eroded bas-relief petroglyphs, we did not

nd any evidence that the quality of the produced RTI

images in corresponding display modes was higher than

in PTM images. It is possible that better results from the

RTI method could be revealed when using a higher quality

input, for example, images with higher colour depth or

with non-compressed le formats. As the input, we used

renderings saved in JPEG format with 24-bit colour depth

and compression level set to 95%.

During the project, we encountered some software

limitations. Apparently, tters are single-threaded and are

limited to a 2 GB RAM usage limit. A software crash can

occur when RAM usage exceeds 2 GB. This limits the

processing of images with a resolution above 4k pixels.

The RAM usage depends on the software settings and se-

lected method, with the RTI tter generally better han-

dling higher resolution input data. For example, we pro-

cessed RTI tter images with resolutions of 4000 × 4000

pixels and with third-level accuracy in 385 sec with a le

size of 411 MB. The PTM Fitter did not manage to nalise

the process with the 4k resolution images when the RGB

option was enabled.

Results and discussion

In total, using alternatively short-range digital photo-

grammetry and structured light scanning tools, 11 petro-

glyphs of varying sizes and dierent stages of preser-

vation were documented in detail. For many of them, it

was almost the last moment when their traces could still

be captured. However, in a few cases, despite using the

vPTM technique, it was already too late to unambiguously

reconstruct their original shape.

However, it should be borne in mind that only a small

group of petroglyphs was selected for this study – those

for which there was a theoretical chance that the applica-

tion of the methods proposed by us could give satisfac-

tory results. On the Samaipata rock, more than 200 other

places were identied where once there were undoubtedly

other glyphs (zoomorphic or geometrical), today com-

pletely indistinguishable.

Petroglyph from sector W06 – “Jaguar?”

As it is one of the best-preserved petroglyphs, we used

it to test various documentation methods. Structured light

scanning, TLS scanning, and short-range digital photo-

grammetry were used.

The gural representation of a sning feline is easy

to recognise, so there was no need to use PTM or RTI.

Instead, the possibility of using alternative methods to

visualise the results of 3D laser scanning and digital pho-

togrammetry were tested. The documentation made with

a structured light scanner (Fig. 9A), as theoretically the

most accurate, served as comparative material.

The petroglyph and its immediate surroundings were

scanned with a Leica P40 scanner from six positions with

distances from the centre of the gural representation

of 5 to 10 m. The scanning density was set to 6 mm @

10 m. Overlapping 3D point clouds guaranteed coverage

and points were measured every 3 mm in the worst case.

The results were presented in the visualisation mode that

Fig. 8. Examples of RTIViewer display modes of PTM files:

A – default; B – normal visualisation; C – specular enhancement;

D – static multi-light (elaborated by J. Kościuk)

 Documentationofthemostimportantpetroglyphs / Dokumentacjanajważniejszychpetroglifów 49

highlighted all slopes (Fig. 9B), referred to in Leica Cy-

clone as “Cloud Silhouette” (CS). The results, although

not as detailed as in structured light scanning due to scan-

ning resolution, proved to be satisfactory. This method of

visualising the 3D point cloud was used to enhance the

legibility of subsequent analyses. The combination of the

CS with the standard visualisation mode of the 3D point

cloud (greyscale reection intensity) oered in Leica Cy-

clone turned out to be a better representation of all sur-

face details than the “Cloud Shaded” option implemented

by Leica (Fig. 9C). An analogous eect was obtained by

combining the CS with an RGB-coloured 3D point cloud

image (Fig. 9D).

Photogrammetric documentation consisting of 48 pho-

tos with resolutions of 24 million pixels each was com-

piled with the AgiSoft PhotoScan Professional software.

By applying a CS layer, a at orthoimage that was devoid

of spatial depth (Fig. 9E) obtained the third dimension

(Fig. 9F). This method proved to be particularly useful

when the high accuracy and sampling density guaranteed

by structured light scanners were not needed, and at the

same time, texture details and the high resolution of an

RGB image was important.

A detailed analysis of the obtained images indicated an

alternation of the earlier original petroglyph. The channel

surrounding the gure might have been a later addition

or at least have been recut. More clear images (especially

Figs. 9B, D), in which not only outlines but also details

are visible much better than in nature, also allowed to pro-

pose the identication of the feline as possibly a jaguar.

Petroglyph from sector W05 – “Puma(?)”

Only very weather-beaten traces of this gural petro-

glyph were extant. Earlier research described it as a rep-

resentation of a puma or jaguar [22], but today it is di-

cult to determine the shape of the gure, not to mention

the particular species.

To document the last traces that will soon be gone for-

ever, the petroglyph and its closest vicinity were scanned

with a structured light Artec Eva handheld scanner. Both

the analysis of the 3D model in the textured version (Fig.

10A) and the one without colour information (Fig. 10B)

did not reveal new details enabling the identication of

the petroglyph shape.

In this situation, the vPTM and vRTI techniques pro-

posed above were used. By putting together 54 renderings

from our virtual rig, each with a dierent lighting direc-

tion (Fig. 10C), it was possible to recognise a shape which

might suggest the representation of puma (Fig. 10D). How-

ever, this reconstruction (Fig. 10D) must be treated with

great caution since the surface is highly eroded and addi-

tionally disturbed by mosses and lichens. Further doubts

can be raised when comparing this reconstruction with

the drawing by Alcide d’Orbigny and Ponce from 1832

[23]. There, the schematic representation of the feline is

turned 180 degrees. Additionally, the map of the rock of

Samaipata drawn by Leo Pucher in 1936 [22] shows this

petroglyph in an inverted orientation. It is dicult to un-

equivocally determine the reliability of either drawing,

especially since, for example, rhea representations are

presented in a dierent orientation by both authors. The

presented reconstruction (Fig. 10E) should therefore be

treated only as one of the possible interpretations of the

traces that have survived to this day.

Fragments of the rough rock surface, especially around

the north-western edge of the oval, were detected on the

images obtained with the vPTM technique. This suggests

that the gure has been intentionally recut (and perhaps

also erased) by the latest occupants of the site. This obser-

vation may point to the pre-Inca origin of this petroglyph.

PetroglyphfromsectorW06–“CrawlingSnake”

This petroglyph, or more precisely, the place where it

once was, is one of the worst preserved among the ex-

amples analysed. It is even dicult to identify exactly

where to look for traces of it. Analysis of the drawings

by d’Orbigny and Pucher [22], [23] only indicates the

area north of the petroglyph probably depicting the jaguar

– more precisely, between it and the steep rock fault in

the north-west (Fig. 11A).

Fig. 9. Petroglyph from sector W06 – “Jaguar(?)”.

Comparison of documentation and methods of visualising results:

A – structured light scanning (image without texture);

B – TLS (slope gradient); C – TLS (reflection intensity and slope gradient);

D – TLS (RGB image and slope gradient); E – digital photogrammetry

(RGB image); F – digital photogrammetry (RGB image)

and TLS (slope gradient) (elaborated by J. Kościuk)

50 Jacek Kościuk, Małgorzata Telesińska, Maciej Nisztuk, Marta Pakowska

As the area where one should look for traces of the lost

petroglyph was much larger (approx. 5 × 10 m), structured

light laser scanning was replaced in this case by short-

range digital photogrammetry. Based on 74 photos, each

with a resolution of 24 million pixels, a 3D model was

built consisting of over 15 million triangles (Fig. 11B).

Using vPTM and vRTI techniques (Fig. 11D), we were

able again to identify at most a vague outline of a snake

crawling towards the north-west (Fig. 11D). However, ac-

cording to some earlier investigators, this petroglyph was

interpreted a coiled snake representation – at least, that is

how Leo Pucher depicts it on his map of Samaipata rock

from 1936 [22]. We did not nd any archival photographs

that indisputably conrm that there was a representation of

a coiled snake, rather the opposite – Alcide d’Orbigny on

the drawing from 1834 [23] shows, although schematical-

ly, the snake crawling towards the north-west. This might

be perhaps a random coincidence that does not prejudge

the correctness of our reconstruction. The preserved traces

are so indistinguishable that they also allow for several

other interpretations, but in the light of collected data, this

one seems to be the most reliable.

Fig. 10. The “Puma(?)” petroglyph from sector W05:

A – RGB image (photo by J. Kościuk); B – greyscale image without texture information (elaborated by M. Pakowska);

C – samples of renderings with different lighting directions (elaborated by M. Budlewska);

D – multidirectional light vPTM image (elaborated by M. Nisztuk); E – tentative interpretation (elaborated by J. Kościuk)

Fig. 11. Petroglyph from sector W06 interpreted as “Crawling Snake”:

A – RGB image (photo by J. Kościuk); B – greyscale image without texture information (elaborated by M. Pakowska);

C – samples of renderings with different lighting directions (elaborated by M. Budlewska);

D – multidirectional light vPTM image (elaborated by M. Nisztuk); E – tentative interpretation (elaborated by J. Kościuk)

 Documentationofthemostimportantpetroglyphs / Dokumentacjanajważniejszychpetroglifów 51

Petroglyph from sector W09 – “Feline (Puma?)”

This petroglyph is the best preserved in the whole an-

alysed group. However, comparing its current condition

(Fig. 12) with photos published by Hermann Trimborn [22],

it can be seen that the surface has eroded signicantly in

the last 50 years. Numerous inscriptions scratched by tour-

ists on the surface of this petroglyph (Fig. 13), which were

still clearly visible 50 years ago, are now barely noticeable.

Given the rapidly progressing erosion, it was decided,

even though the puma drawing is still distinguishable, to

scan the entire petroglyph using a structured light scanner

in the highest resolution available to us (Figs. 14A, B). In

the future, this may be the only source documenting this

petroglyph.

In addition, vPTM analyses were compiled. The few

inscriptions that are still legible (Fig. 14C) clearly show

how far the surface of the petroglyph has eroded relative

to the situation of the late 1960s.

The representation of the puma is carved in bas-relief,

and a at water channel surrounds it (Fig. 14D). A round-

ed hole (ca. 25 cm in diameter) drilled into the rock in

front of this petroglyph might have been used as an of-

fering hole. It seems that the petroglyph was made at the

same time as the platform arranged around it. Observation

of the sculpting sequence of the surrounding platforms

leads to the conclusion that puma representation could be

attributed to the period of Inca occupation of the site

Petroglyph from sector W09 – “Rhea (Ñandú)”

This petroglyph, known nowadays only from historical

sources, is an entirely dierent case. One can easily iden-

tify the place where this petroglyph was once located, but

the gure is completely indistinguishable.

The size of the area (ca. 6 × 7 m) excluded structured

light scanning, so again digital close-range photogramme-

try was used. More than 300 photos resulted in a dense

cloud of ca. 21 million points, which was reduced to a 3D

model of 4 million faces (triangles). It proved to be fully

sucient to reproduce all the details of the examined sur-

face (Fig. 15A). An orthoimage of 0.5 mm/pixel resolution

also met our requirements (Fig. 15B).

However, attempts to nd any traces of the rhea (ñandú)

ended only with a suggestion about the size and orienta-

J. Kościuk, G. Oreci, M. Ziółkowski, A. Kubicka, R. Muñóz

Risolazo, Description and analysis of El Fuerte de Samaipata in the

lightofnewresearch,and a proposaloftherelative chronologyofits

main elements, in the same issue of “Architectus”.

Fig. 12. The “Puma(?)” petroglyph from sector W09.

Condition as in July 2016

(photo by J. Kościuk)

Fig. 13. The “Puma(?)” petroglyph from sector W09.

Condition as in 1967

(source: [22])

Fig. 14. Petroglyph from sector W09 – “Puma(?)”:

A – RGB image (photo by J. Kościuk); B – greyscale image without texture information (elaborated by M. Pakowska);

C – vPTM image (multidirectional light) (elaborated by M. Nisztuk); D – tentative interpretation (elaborated by J. Kościuk)

52 Jacek Kościuk, Małgorzata Telesińska, Maciej Nisztuk, Marta Pakowska

tion of the gure. Generally, it agrees with the schemat-

ic “icon” on d’Orbigny’s drawing [23]. There, however,

the gure is shown in an entirely dierent manner than

all other petroglyphs, which, as we can see today, are of

bas-relief type. Can this dierence in the way of depicting

the rhea (ñandú) on d’Orbigny’s drawing mean that it was

executed as a sunken relief? This dierence might be one

of the reasons that it has disappeared utterly

, or it might

have been intentionally hammered out.

Using vPTM and vRTI techniques, we ended up with

a problematic reconstruction (Fig. 15E). Even attempts to

scale the virtual 3D model twice alongside the z-axis did

not bring clear results. One must admit at this point, that

the reconstruction proposed there (with head down) has

no analogies in pre-Columbian art in the whole region, so

it can not be treated as a right interpretation. It should be

also emphasized that this reconstruction is based solely

on the technical analysis of preserved traces with no ref-

erence to possible analogies in Inca or pre-Inca iconog-

raphy. Further iconographical studies may still indicate

another direction of interpretation, but this interesting ex-

ample of pre-Columbian rock art is anyhow lost forever

Petroglyph from sector W10

–“Snake-catchingAnimal”

Small bulges on the surface of the rock in the central

part of the W10 sector suggested that this could be another

very blurred petroglyph (Fig. 16A). Structured light scan-

Marta Pakowska brought this possibility to our attention.

For alternative reconstruction cf. J. Kościuk, G. Oreci, M. Ziół-

kowski, A. Kubicka, R. Muñóz Risolazo, Description and analysis of

El Fuerte de Samaipata in the light of new research,and a proposal

of the relative chronology of its main elements, in the same issue of

“

Architectus”.

ning conrmed this assumption, but still, a clear interpre-

tation was not possible (Fig. 16B).

It was only after scaling the virtual 3D model two times

along the z-axis and using vPTM techniques that a more

explicit identication was made – an animal catching

a snake’s head with its snout (Figs. 16D, E). The second

snake is grabbing the tail of the rst one, and the third

snake catches the tail of the second snake. This compli-

cated layout of four interacting animals covers approxi-

mately 1 m sq. It has no other parallels on the Samaipata

rock. Additionally, the representation of the snake-catcher

looks untypical. It does not resemble pumas or jaguars

known from other petroglyphs in Samaipata. Attempts

to identify the species lead towards the white-nosed coati

(lat.: Nasua narica; local names: pizote, antoon, tejón), the

only similar animal from this part of South America, which

are known as snake catchers [24]. Consultations with spe-

cialists from the Wroclaw Zoological Garden have not

ruled out such a possibility. Further studies on possible

analogies in the Inca and pre-Inca rock art of Bolivia may

conrm or exclude this preliminary interpretation.

Thanks to the exact virtual copy of this petroglyph,

even if it erodes completely, additional research and iden-

tication attempts will be still possible.

Petroglyph from sector W10

–“MeanderingSnake”

About 2 m northwest of the above-described petro-

glyph, an elongated twisting shape was observed at the

bottom of the 25 cm wide channel draining rainwater

down the slope (Fig. 17A). Due to the progressing ero-

sion, additionally intensied by the rainwater owing

abundantly here, structured light scanning was used at

the highest resolution available to us for this petroglyph

(Fig. 17B).

Fig. 15. Petroglyph from sector W09 interpreted as “Rhea (Ñandú)”:

A – RGB image (photo by J. Kościuk); B – greyscale image without texture information (elaborated by M. Pakowska);

C – samples of renderings with different lighting directions (elaborated by M. Budlewska);

D – multidirectional light vPTM image (elaborated by M. Nisztuk); E – tentative interpretation (elaborated by J. Kościuk)

 Documentationofthemostimportantpetroglyphs / Dokumentacjanajważniejszychpetroglifów 53

After computing the vPTM analysis (Fig. 17C), a clear

image of a snake meandering up the slope in the middle of

the channel was shown (Figs. 17D, E).

The degree of erosion, proximity to the petroglyph rep-

resenting a snake-catching animal, and the similar way of

carving suggest that both petroglyphs form one panel.

The iconography of the crawling snake is often found

in both Inca and pre-Inca rock art. On the Samaipata rock,

this type of glyph occurs in three other places: in sector

W05, along a long ramp in sector S12 (this petroglyph is

over 20 m long), and in the channel draining water from

the pool in sector N10.

Petroglyph from sector C05

–“ChoirofthePriests”or“Altar”

On the central part of sector C05, one of the most im-

portant sculpted parts of the Samaipata rock can be iden-

tied: it is what was called the “Choir of the Priests” or

the “Altar”.

Due to the large size of this sculpture (roughly 9 × 13 m

when measured together with the leading north-east water

channel), digital short-range photogrammetry was used

as the documentation method. Based on 83 photos, each

with a resolution of 24 million pixels, a 3D point cloud of

Fig. 16. Petroglyph from sector W10 – “Snake-catching Animal”:

A – RGB image (photo by J. Kościuk); B – greyscale image without texture information (elaborated by M. Pakowska);

C – samples of renderings with different lighting directions (elaborated by M. Budlewska);

D – multidirectional light vPTM image (elaborated by M. Nisztuk); E – tentative interpretation (elaborated by J. Kościuk)

Fig. 17. Petroglyph from sector W10 – “Meandering Snake”:

A – RGB image (photo by J. Kościuk); B – greyscale image without texture information (elaborated by M. Pakowska);

C – samples of renderings with different lighting directions (elaborated by M. Budlewska);

D – multidirectional light vPTM image (elaborated by M. Nisztuk); E – tentative interpretation (elaborated by J. Kościuk)

54 Jacek Kościuk, Małgorzata Telesińska, Maciej Nisztuk, Marta Pakowska

over 15 million instances was built. Due to the size of the

le, when converting to the mesh model, it was decimat-

ed to 3.5 million triangles (Figs. 18A, B). This resolution

proved to be sucient for interpretation of its main parts.

Special attention was devoted to the central part sur-

rounded by a nearly 1-meter wide water-channel. Beside

nine rectangular “seats”, most of them very eroded, trac-

es of small (ca. 14 × 14 cm) double-recessed niches were

found on the vertical surfaces between sitting places.

Further studies included vPTM analysis (Fig. 18C) in the

hope that the top of the central part would reveal traces

of another petroglyph (Fig. 18D). Unfortunately, no such

traces were observed. Even if such a petroglyph exist-

ed here once, it has been completely eroded or has been

intentionally damaged. Additionally, no indication was

found for Giuseppe Oreci’s suggestion of a representa-

tion of a feline preceding a circular channel with radiating

triangular and rectangular niches. This does not exclude

the hypothesis made by Oreci based on his observations

in the eld. At most, it can be said that the use of vPTM

did not conrm his direct observations. However, vPTM

analyses yielded data that allow for a more detailed recon-

struction of all rectangular and triangular niches.

Petroglyph from sector S14

– “Double Puma(?)”

On the southern slope of the Samaipata rock, in a very

prominent position, two images of felines (possibly pu-

mas?) attract the attention of many visitors. The animals

are facing each other, and a narrow gap is left between

their heads. Only the animal on the western side is pre-

served in full. Of the other one, only the general outline

remains. The main part of its torso has vanished and traces

of its head are no longer visible.

Since the continued existence of this petroglyph is

endangered by both the expanding lithic surface and the

strongly inclined slope, digital photogrammetry and struc-

tured light scanning were used to document its current

state. Due to the very dicult access, only 34 photos were

collected. They were, however, sucient to produce ortho-

images of 1 mm/pixel resolution (Fig. 19A) and a detailed

3D mesh model consisting of 8 million triangles. The last,

when inspected with textural information switched o

(Fig. 19B), revealed interesting details. A small rectangu-

lar depression (ca. 45 × 70 cm) above the heads of both

animals might have been meant as a reservoir from which

water owed between the two pumas below.

More detailed structured light scanning shows two

circular protrusions inside this depression (Fig. 19D).

Small cavities in the centre of each protrusion were also

detected. The whole, together with the water drain locat-

ed below, looks a bit like a stylised mask of a mythical

creature.

Since no traces have been preserved to undoubtedly re-

construct the head of the eastern animal, our interpretation

reproduces the outline of the feline preserved next to it

(Fig. 19E).

PetroglyphfromsectorN10–“TripleSnake”

This petroglyph is a very elaborate, nearly 4 m-long

channel through which water ows to the east from a set

of three cascading reservoirs located up the hill on the

western side

Cf. J. Kościuk, G. Oreci, M. Ziółkowski, A. Kubicka, R. Muñóz

Risolazo, Description and analysis of El Fuerte de Samaipata in the

lightofnewresearch,and a proposaloftherelative chronologyofits

main elements, in the same issue of “Architectus”.

Fig. 18. The “Altar” petroglyph from sector C05:

A – RGB image (photo by J. Kościuk); B – greyscale image without texture information (elaborated by M. Pakowska);

C – samples of renderings with different lighting directions (elaborated by M. Budlewska);

D – multidirectional light vPTM image (elaborated by M. Nisztuk); E – tentative interpretation (elaborated by J. Kościuk)

 Documentationofthemostimportantpetroglyphs / Dokumentacjanajważniejszychpetroglifów 55

The bottom of the channel has the form of a meander-

ing snake, although some details of its easternmost part

cannot be clearly seen. Two parallel shelves are also vis-

ible on both sides of this meandering channel, but their

edges are not easily identiable. The colourful mosaic of

lichen and algae covering the rock surface makes it very

dicult to observe this petroglyph directly in the eld

(Fig. 20A).

Since this site was lacking small details as well as high-

ly eroded areas, only close-range digital photogrammetry

was used as the documenting technique. More than 160

photos were shot to cover the area extending for a further

7 m to the east in the hope of nding traces of the contin-

uation of this canal. This resulted in a 3D cloud of 20 mil-

lion points, which was converted to a 30 million-triangle

mesh model (Fig. 20B).

The 3D model did not demonstrate the existence of

a continuation of the meandering channel (Fig. 21D). In-

stead, it clearly dened the northern and southern edges

of the shelves on both sides of the channel. Further details

were observed on the vPTM image (Figs. 21C, D). On

both 25 cm-wide shelves, a similar meandering pattern

has been carved, but this time as a bas-relief (Fig. 21E).

This “Triple Snake” glyph with sunken relief and bas-

reliefs used together has no parallel on the Samaipata

rock.

Fig. 19. The “Double Puma(?)” petroglyph from sector S14:

A – RGB image (photo by J. Kościuk); B – greyscale image without texture information (elaborated by M. Pakowska);

C – samples of renderings with different lighting directions (elaborated by M. Budlewska);

D – multidirectional light vPTM image (elaborated by M. Nisztuk); E – tentative interpretation (elaborated by J. Kościuk)

Fig. 20. Petroglyph from sector N10 – “Triple Snake”:

A – RGB image (photo by J. Kościuk); B – greyscale image without texture information (elaborated by M. Pakowska);

C – samples of renderings with different lighting directions (elaborated by M. Budlewska);

D – multidirectional light vPTM image (elaborated by M. Nisztuk); E – tentative interpretation (elaborated by J. Kościuk)

56 Jacek Kościuk, Małgorzata Telesińska, Maciej Nisztuk, Marta Pakowska

Zoomorphic petroglyph from sector N10

Traces of this petroglyph were identied by Rolando

Marulanda [25]. Some irregularities of the surface of the

central terrace of this sector may suggest another zoomor-

phic representation, unfortunately in an advanced state of

erosion.

Unfortunately, this particular place was not recorded

by us with either detailed photogrammetry or structured

light scanning. The only data available for further inter-

pretation were those from the general photogrammetry

and TLS projects covering the entire site.

The last proved to be sucient to render interpretation

possible. Generally, TLS data are not particularly suitable

as a source for detailed 3D models reproducing all the nu-

ances of the surface. However, the mere observation of the

3D point cloud using dierent imaging methods proved

sucient for this task. Three dierent 3D point cloud vis-

ualisation modes were used in this case: RGB coloured,

greyscale representation of reection intensity, and an el-

evation map with hill shading (Figs. 21A–D). All of them

are available in Leica Cyclone as basic colour schemes for

point cloud representation.

Although our tentative interpretation (Fig. 21D) is

slightly dierent from the one proposed by Rolando Maru-

landa [25], it has conrmed that a zoomorphic petroglyph

could once have existed in this place. Its exact shape can

no longer be reproduced. At most, one can venture to say

that the petroglyph could represent an animal belonging to

the felidae family. However, a shadow of doubt remains

as to whether this is not the result of natural erosion of

heterogeneously stratied sandstone.

Conclusions and project limitations

The obtained results conrmed the correctness of the

assumption that with an accurate 3D artefact model, the

process of collecting photos at the dierent lighting an-

gles necessary for PTM and RTI techniques can be suc-

cessfully transferred from the physical environment to the

virtual space. The project also resulted in many indica-

tions as to the correct and eective application of all the

technologies used. The following discussion is divided

into three areas: documentation techniques, rendering im-

ages in a virtual environment, and the use of existing PTM

and RTI software.

In general, somewhat predictably, the outdoor use of

structured light scanners presents many technical prob-

lems. The main one is poor visibility of structured light

strips emitted by the scanner in daylight, and even more so

in full sunlight. Scanning after sunset would be a natural

solution, but for archaeological sites, this is usually not

authorised or involves obtaining special permits and em-

ploying additional personnel (inspectors, guards, techni-

cal employees). Additional problems arise when the scan-

ner and the associated laptop cannot be powered directly

from the network. If powered by batteries, both work with

much lower eciency, which signicantly hinders the

correct completion of the scan of the entire object.

As the experience gained during this project shows, in

most cases, structured light scanning can be replaced with

short-range digital photogrammetry. The key to success

is the use of a camera (full-frame cameras being highly

recommended) with a resolution above 16 million pixels

and a professional, xed, focal length lens that complies

with the basic principles of photography for digital pho-

togrammetry purposes. The latter include the following in

particular:

– The images should be as much as possible taken per-

pendicularly to the documented surface;

– The images must be sharp and devoid of reections;

– The camera should work in aperture priority mode

while the lens aperture should be permanently xed at the

value that will guarantee an optimal depth of eld (usually

f = 8÷11 – depending on lens model and dimensions of the

object);

Fig. 21. Zoomorphic petroglyph from sector N10.

3D cloud from TLS representations:

A – RGB image; B – reflection intensity in greyscale;

C – elevation map with hill shading;

D – tentative interpretation (elaborated by J. Kościuk)

 Documentationofthemostimportantpetroglyphs / Dokumentacjanajważniejszychpetroglifów 57

– The focus should be permanently xed at a distance

that will guarantee sharpness for all the images at the giv-

en aperture size;

– To avoid the eect of vibrations, the shutter speed

should be high, so good lighting conditions are necessary

since raising the ASA value to high will result in grainy

images;

– There should be a minimum of 75% vertical and

horizontal overlap between adjacent images;

– Too many photos may result in a blurry and noisy 3D

model, so good planning of camera positions is advisable.

The additional benet of using the photogrammetric

method is exibility during the acquisition phase, which

enables data collection in places not directly accessible to

hand-held scanners.

When the 3D model of the inspected artefact is ready

(at this point the scanner or photogrammetry obtained

do not matter), the phase of rendering images in a vir-

tual lighting-array will start. The lesson we learned from

our virtual rig prototype includes the following observa-

tions:

– In some cases, the 54 predened lighting positions

of our virtual rig were not sucient, so an additional set

of rigs should be dened;

– In order to assure good reproduction of all details,

the script automating successive renderings should in-

clude adjusting the size of the frame according to the real

size and proportions of the 3D model (the “Triple Snake”

petroglyph is a good example where the rendered object

occupied less than 50% of the image);

– More studies are necessary to establish light de-

nitions and material properties optimal for the proposed

vRTI and vPTM techniques.

Further studies will show if the proposed workow is

also adequate for documenting and analysing artefacts

other than bas-relief petroglyphs.

References /Bibliografia

[1] http://rsb.info.nih.gov/ij/index.html [accessed: 15.07.2019].

[2] https://imagej.nih.gov/ij/features.html [accessed: 19.07.2019].

[3] www.dstretch.com [accessed: 11.07.2019].

[4] Harman J., DStretchAlgorithm Description, http://www.dstretch.

com/Algorithm Description.html [accessed: 3.07.2019].

[5] Malzbender T., Gelb D., Wolters H., Zuckerman B., Enhancement of

ShapePerceptionbySurfaceReectanceTransformation, “Hewlett-

Packard Technical Report 2000-38R1”, https://www.hpl.hp.com/

techreports/2000/HPL-2000-38R1.pdf [accessed: 17.08.2019].

[6] Malzbender T., Gelb D., Wolters H., Polynomial Texture Maps,

“Hewlett-Packard Technical Report 2001-33 (R.1)”, 16.05.2001,

http://www.hpl.hp.com/ptm [accessed: 17.08.2019].

[7] ReectanceTransformationImaging(RTI), http://culturalheritage-

imaging.org/Technologies/RTI/index.html [accessed: 17.08.2019].

[8] Witkowski P., Chyla J.M., Ejsmond W., Combination of RTIand

decorrelation – an approach to the examination of badly preserved

rockinscriptionsandrockartatGebelein(Egypt), [in:] S. Campa-

na, R. Scopigno, G. Carpentiero, M. Cirillo (eds.), CAA2015:Keep

the revolution going. Proceedings of the 43rd annual Conference

onComputerApplicationsandQuantitativeMethodsinArchaeol-

ogy, Archaeopress Archaeology, Oxford 2016, 939–944.

[9] D’Orbigny A., ViajealaAméricameridional(realizadode1826

a1833), Vol. 4, 1466, Buenos Aires 1945.

[10] Technical specication, Eva, https://www.artec3d.com/portable-

3d-scanners/artec-eva-v2#specication [accessed: 19.08.2019].

[11] Agisoft Photoscan professional edition, https://www.agisoft.com/

features/professional-edition/ [accessed: 31.02.2019].

[12]

Horn B.K., Hillshadingandthereectancemap, “Proceedings of the

IEEE” 1981, Vol. 69, Iss. 1, 14–47, doi: 10.1109/PROC. 1981.11918.

[13] Kennelly P.J., Hill-Shading Techniques to Enhance Terrain Maps,

2009, https://pdfs.semanticscholar.org/26fc/7b5fbad0e3692d2040b

59530add6d87215d3.pdf?_ga=2.140942640.329251785.1567283612-

1142592307.1532500819 [accessed: 18.08.2019].

[14] Kennelly P.J., Terrain maps displaying hill-shading with curva-

ture, “Geomorphology” 2008, Vol. 102, Iss. 3–4, 567–577, doi:

10.1016/j.geomorph.2008.05.046.

[15] https://www.maxon.net/en/products/cinema-4d/overview/ [ac-

cessed: 11.07.2019].

[16] Witkowski P., Polkowski P., Reectance transformation imaging

about a photographic method for documentation and analysis of

historical artefacts, “Muzealnictwo” 2018, Vol. 59, 54–62.

[17] CulturalHeritageImaging,ReectanceTransformationImaging.

Guide to Highlight Image Processing, Version 1.4., 2013, http://

CulturalHeritageImaging.org/downloads/ [accessed: 8.08.2019].

[18] CulturalHeritageImaging,ReectanceTransformationImaging.

GuidetoRTIViewer,Version1.1., 2013, http://CulturalHeritageIm-

aging.org/downloads/ [accessed: 8.08.2019].

[19] HP Labs, PTM Fitter, 2001, https://www.dropbox.com/sh/jfsy0l-

hxu6zv4i4/AADJpq6E_GJmNw_s5C8r94CVa?dl=0 [accessed:

8.08.2019].

[20] CulturalHeritageImaging,ReectanceTransformationImaging.

Guide to Highlight Image Processing, Version 1.4., 2013, http://

CulturalHeritageImaging.org/downloads/ [accessed: 8.08.2019].

[21] Graichen T., RTI–accuracytests, 2019, http://www.tgraichen.de/

?page_id=542 [accessed: 16.08.2019].

[22] Trimborn H., Der skulptierte Berg von Samaipata, “Archäolo gi-

sche Studien in den Kordilleren Boliviens”, Bd. 3, “Baessler Archiv,

Beiträge zur Völkerkunde” 1967, NF 5.

[23] Ponce Sanginés C., Alcidesd’Orbignyysuviajea Samaipataen

1832, [in:] R.D. Arze Aguirre (dir.), El naturalista francés, Al-

cide Dessaline d’Orbigny en la visión de los bolivianos, Institut

Français d’Études Andines, Plural editores, Lima–La Paz 2002

[1975], 307–315.

[24] https://en.m.wikipedia.org/wiki/White-nosed_coati [accessed:

15.03.2019].

[25] Marulanda R., LarocaesculpidadelFuertedeSamaipata:elemen-

tosdediscusiónparaunenfoqueinterpretativo, [in:] A. Meyers,

I. Combès (comp.), ElFuertedeSamaipata.Estudiosarqueológi-

cos, Biblioteca del Museo de Historia, Universidad Autónoma Ga-

briel René Moreno, Santa Cruz de la Sierra 2015, 34–51.

Acknowledgements /Podziękowania

Thepresentedworkis a partof the research sponsored bythe grant given

totheWrocławUniversityofScienceandTechnologybythePolishNa-

tional Science Centre (grant No. 2014/15/B/HS2/01108). Additionally,

themunicipalityofSamaipata,representedbyMayorFalvioLópesEs-

calera, contributed to this research by providing the accommodation

duringtheeldworkinJuneandJuly2016,aswellasinJuly2017.The

MinistryofCultureandTourismofBoliviakindlygrantedallnecessary

permits(UDAMNo.014/2016;UDAMNo.060/2017).Theresearchwas

conductedinclosecooperationwiththeCentreforPre-ColumbianStud-

iesoftheUniversityofWarsawinCusco.

58 Jacek Kościuk, Małgorzata Telesińska, Maciej Nisztuk, Marta Pakowska

Abstract

The case being studied is one of Bolivia’s most important monuments – El Fuerte de Samaipata, a UNESCO World Heritage Site. This paper pro-

poses a new workow for the two most important aspects of studies on cultural heritage – detailed documentation and analysis. The former includes

the well-known techniques of digital photogrammetry and structured light scanning. The latter comprises polynomial texture mapping (PTM) and

reectance transformation imaging (RTI), both of which have been used since the beginning of this century. The novelty proposed by the authors is

the transfer of part of the data collection process from the physical environment to the virtual space.

Despite some technical problems, by eliminating the tedious and time-consuming process of shooting photos in the eld using specic lighting

angles, this new workow proved to be very ecient, particularly in documenting and interpreting badly preserved examples of bas-relief rock art.

Key words: Samaipata, rock art, UNESCO World Heritage, PTM and RTI, 3D scanning

Streszczenie

El Fuerte de Samaipata to jeden z najważniejszych zabytków Boliwii wpisanych na Listę Światowego Dziedzictwa UNESCO. W artykule przedsta-

wiono dwa sposoby dokumentacji, szczególnie istotne zarówno dla badań, jak i ochrony dziedzictwa kulturowego. Pierwszy obejmuje dobrze znane

techniki cyfrowej fotogrametrii i skanowania światłem strukturalnym. Drugi dotyczy zastosowania technik polynomial texture mapping (PTM) oraz

reectance transformation imaging (RTI). Obie techniki są wprawdzie stosowane już od początku tego wieku, ale nowością zaproponowaną przez

autorów jest przeniesienie części procesu zbierania danych ze środowiska zycznego do przestrzeni wirtualnej.

Mimo pewnych problemów technicznych, dzięki wyeliminowaniu żmudnego i czasochłonnego etapu fotografowania w terenie przy użyciu dokład-

nie zdeniowanych kątów oświetlenia, zaproponowana metoda okazała się bardzo wydajna, szczególnie w dokumentowaniu i interpretacji silnie

zniszczonych przykładów rytów naskalnych.

Słowa kluczowe: Samaipata, sztuka naskalna, Lista Światowego Dziedzictwa UNESCO, PTM i RTI, skanowanie 3D

Dokumentacja najważniejszych petroglifów za pomocą skanowania światłem strukturalnym oraz analiza najbardziej zniszczonych petroglifów metodami vPTM i vRTI

Jacek Kościuk, Małgorzata Telesińska, Maciej Nisztuk, Marta Pakowska

doi:10.37190/arc200207

Streszczenie