Gathering GPR Inspections and UAV Survey in Cultural Heritage

Documentation Context

Alessandro Arato

, Flora Garofalo

, Giulia Sammartano

and Antonia Spanò

(DAD) Department of Architecture and Design, Politecnico di Torino, Torino, Italy

(DIATI) Department of Environment, Land and Infrastructure Engineering, Politecnico di Torino, Torino, Italy

Keywords: UAV Photogrammetry, Ground Penetrating Radar (GPR), Historical Maps, GIS, Cultural Heritage.

Abstract: The archaeological researches and more generally the Cultural Heritage (CH) documentation and

conservation activities have been favourably disposed to the use of new technologies, with renewed and

increasing interest in the use of integrated techniques.

In the field of Geomatics the advent of advanced technologies has allowed and facilitated multidisciplinary

studies as well as combined approaches to the documentation in various contexts. The production of spatially

located data (e.g. from active or passive sensors placed in different system segments, from terrestrial to aerial

to satellite position) and their interoperability from different source, with the help of Geographic Information

Systems (GIS), were then made easier.

The work has the aim of investigating the integration of multiple data derived from aerial photogrammetry

products through Unmanned Aerial Vehicles (UAV) survey, from geophysical Ground Penetrating Radar

(GPR) prospection technique and analysis of historical maps. An archaeological area in the south of Piedmont

(Italy), next to the ancient Roman settlement of Pollentia, has been the test case. The present fulfilled test was

objected to exclude ancient presences, although this type of workflow is generally aimed to analyse and

compare results in order to formulate some hypothesis about the potential presence of submerged elements or

built substructures in the investigated area.

1 INTRODUCTION

Adding to the widening of cultural assets typologies

subject to the safeguard aims, ranging from ancient to

modern heritage, from architectural to landscape, and

from mobile assets to intangible cultural heritage, the

increasing challenges of CH conservation has given

rise to an incessant specialization of methods

involved in the protection projects. This trend has

become visible in data collecting and survey, data

archiving and structuring, and finally in

communication and dissemination processes. Some

features are increasingly requested to acquisition

techniques: mainly 3D nature of data, quick acquiring

and low cost solutions, keeping high accuracies and

overall quality. On these bases, the present paper wish

to investigate the integration of strengthened

techniques as Ground Penetrating Radar purposed to

search possible submerged archaeological remains,

with emergent methods for quick large-scale

mapping, generated from UAV photogrammetry.

Moreover the examination of historical maps,

generally highly significant in this kind of project,

have been compared and commented by GIS tools in

final results.

2 METHODS

2.1 UAV Photogrammetry

The UAV photogrammetry has developed very

rapidly whereas it is able to fulfil successfully the

large-scale maps updating in many fields such as

environmental, CH applications. (Haala et al., 2011)

One relevant strong point of UAV systems for

photogrammetry is the nadiral point of view

connected with the ability to choose the height of

flight depending on the requested scale of survey.

The low cost, the high handily with contained

playloads, and the ability to perform autonomous

flights starting from a predefined flight plan, are other

decisive advantages. Navigation sensors such as

GNSS receivers (Global Navigation Satellite System)

Arato, A., Garofalo, F., Sammartano, G. and Spanò, A.

Gathering GPR Inspections and UAV Survey in Cultural Heritage Documentation Context.

In Proceedings of the 2nd International Conference on Geographical Information Systems Theory, Applications and Management (GISTAM 2016), pages 85-91

ISBN: 978-989-758-188-5

and inertial sensors (IMU, Inertial Measurement

Unit) are conveniently integrated; the effectiveness of

the whole system depends from the balance of their

accuracy and reliability with the Flight control

device. (Rinaudo et al., 2012)

The product of UAV photogrammetry consists in

a set of images from which it is possible to extract

point clouds, 3D textured models, Digital Surface

Model (DSM) and Orthophoto the processes are

mainly automated by an image matching technique,

so it’s clear the reasons of increase.

In the next experience the only ortophoto will be

used, which is absolutely the most automated product

that is possible to derive, and this is one of the main

reason of the method choice.

Structure-from-Motion (SFM) systems allow

estimating the 3D position of points represented in

multiple images, reconstructing the geometry of the

represented object (structure) and the acquisitions

position (motion), even when are not available the

defined camera calibration parameters. The

characteristic elements (tie-features) of images are

automatically preliminarily extracted; in particular

they are extracted objects (points) recognizable by the

software thanks to the radiometric contrast in

different directions, which identifies unambiguously

the pixel. One of the most common algorithms for

features identification and description for the

association and the consequent orientation of images

in photogrammetric software and computer vision is

SIFT (Scale-Invariant Feature Transform). (Lowe,

2004; Szeliski,

2011)

In the next step of feature matching, points with

more similar descriptors (recognized by the software

as homologous) are aligned to make the next step of

triangulation - bundle adjustment.

In SFM processes the orientation parameters are

not necessarily known; so, using the rules of the

epipolar geometry, which regulates relations between

the tie points (TP), the acquisition centers and the

point 3D position, to direct acquisitions and estimate

internal and external orientation.

Figure 1: Two example of unmanned vehicles, the ones

used in the further test: hexacopter by Mikrokopter and

Ebee by Sensfly.

2.2 GPR (Ground Penetrating Radar)

Survey

GPR method is based on the generation of an

electromagnetic (EM) impulse through a signal

generator, connected to a transmitting antenna (Tx)

and a receiving antenna (Rx). The GPR systems work

in the frequency range between 10 MHz and 3 GHz.

The choice of the system frequency is essential and

has to be made according to the scope of the survey.

In fact, being the frequency inversely proportional to

the wavelength, lower frequencies imply deeper

signal penetration but lower resolution, while higher

frequency allow the increase the resolution but

reducing the signal penetration.

In archaeological researches, GPR survey is used

in order to correlate electromagnetic impedance

contrasts to possible submerged remains.

The electromagnetic discontinuities (or dielectric

interfaces), act as signal reflectors, refractors and

diffractors, via multiple mechanisms (e.g. Davis and

Annan, 1989; Reynolds, 1997). Soil specific

characteristics, like structure and texture, mineral

composition, pore-fluid distribution and chemistry,

influence the signal penetration as well (e.g. Sen et

al., 1981; Friedman, 1998; Cosenza et al., 2003).

Each reflection implies an energy loss, and the

received signal is strongly attenuated with respect to

the constant transmitted signal. Common GPR

surveys are carried out along profiles, with the aim of

obtaining 2-dimensional (2-D) sections of the subsoil.

A radar profile is a composition of all the recorded

signals, called traces, and allows to understand the

subsoil inhomogeneity, or dielectric interfaces. In

archaeological context, the analysis of a single profile

is not sufficient, and a 3-D survey must be done to

map the lateral discontinuities. The results are usually

presented in the form of time-slices, horizontal maps

at constant time (or depth).

2.3 Historical Maps Contribution and

Their Spatial Reference

The success of the GPR survey is highly affected by

the prevision of searched remains. The shape of

possible submerged building elements related with

the period of construction and especially the depth at

whom they are expected to be found are very relevant

in order to correctly plan the parameters of GPR

survey. Then a panoramic examination of historical

cartography is usually highly advisable.

The geometric and projective characteristics and

properties of early cartographic representations have

been largely studied when digital technologies have

GISTAM 2016 - 2nd International Conference on Geographical Information Systems Theory, Applications and Management

spread new perspectives for the visualization in the

same spatial reference of recent maps. (e.g.

Livieratos, 2006; Balletti, 2006).

Many research projects has directed efforts towards

the use of GIS techniques of calibration, georeferencing

and transformation of projection, for enabling the

comparison among historical and actual maps.

In brief, the relation among the points coordinates

of historical maps that have been transformed in image

coordinates and the homologous coordinates in recent

reference system are searched by Ground control

points, derived by GPS survey or digital recent maps.

3 THE EXPERIENCE ON THE

ARCHAEOLOGICAL AREA

The site under investigation is a plot of about 200 x 200 m;

the area is flat, with some irregularities and some obstacles

but the key point is the proximity to the ancient Roman

settlement of Pollentia, the most ancient city of Tanaro

Valley in South Piedmont. Archaeological researches

lightened that the roman village was extended south to the

roman amphitheatre, currently visible because of the village

developed upon it. (Figure 2, 3). (Spanò et al., 2007)

Since the plot is placed close to the road for Alba

Pompeia, some funerary remains or burials were the only

ancient elements that were reasonably expectable to be

found.

Figure 2: Excerpt from the

orthophoto IT 2000 -

Piedmont Region and

highlighting the involved lot.

Figure 3: The Pollentia

large-scale archaeological

map superimposed on the

regional digital map.

The site presents a plantation of walnut trees, arranged

in the 1960s, so the usually desirable maps or aerial

photograms recognition aimed to search crop-marks was

in this case to be discarded. Therefore, the GPR survey,

spatially located by the UAV photogrammetrical

application, was performed according to the next report

and after results some historical maps have been useful to

complete the reading.

3.1 Photogrammetric Survey of Lot

2076 in Pollenzo by Drones

The first purpose of the photogrammetric flight on the

Pollenzo 2076 lot was exactly the fast generation of a

high resolution ortophoto, using aerial photos acquired

contemporary to GPS survey in order to keep easily the

spatial reference of GPR profiles. Actually two flights

have been fulfilled by two different systems: a fix wing

drone performing a 120 m height flight (Ebee SenseFly)

and a hexacopter drone (by Mikrokopter) for a lower

flight ensuring higher resolution. Provided that the two

aircrafts are featured by different maximum altitude,

maximum flight time, radio link range etc. they share

important abilities for mapping purposes. Mainly it is the

automated plan flight that is based on waypoints: that

enable to foresee all parameters in order to acquire

perfectly planned photogrammetric strips, including the

desired overlapping and ground sampling distance.

The setting up of a topographic network and the

measure of control points by GPS/GNSS techniques

preceded the flights, guaranteed the processing of the

orthophoto in the reference system of regional maps

(WGS84-ETRF2000) and ensured the required

accuracies.

The digital camera Sony Nex 5 acquired the large

amount of aerial photos, reduced to 104 through a

automated selection procedure; those photos with an

overlap between consecutive frames of more than

90% are very suitable for a SfM approache.

Photogrammetric orientation and ortophoto

generation results:

 104 photograms

 ground resolution: 0.03 m/pix,

 Camera stations: 104,

 Tie-points: 21122,

 Medium error on GCP: 0.06 m,

 Medium error on CP: 0.13 m

3.2 GPR Data Acquisition and Signal

Processing on Lot 2076

The survey was conducted along parallel lines,

keeping at least 1 m of distance from the trees. The

tree roots have a significant influence on signal

penetration, as they act as diffractors and tend to

scatter the signal in several directions. The site was

almost plain, with obstacles (trunks, branches, soil

accumulations) and some irregularities. A muddy

upper soil layer (5 cm) was present over all the area.

GPR data were acquired along 84 parallel profiles,

covering a total area of 2000 m

(Figure 4). 50

profiles, having 66.5 m length, were acquired along

NW-SE direction, while the remaining 34 profiles,

Gathering GPR Inspections and UAV Survey in Cultural Heritage Documentation Context

along SW-NE direction, were 40 m long.

The inter-profile distance was 0.6 m (except 3 m

gaps in correspondence of the tree rows), in order to

guarantee a spatial coverage, adequate to expected

size of the remains.

Figure 4: Acquisition profiles on orthophoto: blue and red

lines indicate, respectively, NW-SE and SW-NE profiles.

GPR survey was conducted by using IDS K2-

MCH radar system, with 70 MHz and 200 MHz

antennae, and an integrated survey wheel encoder.

Each trace was sampled in-time, over a sampling

window of 120 ns, with 1024 samples. A single radar

trace was acquired every 3.5 cm along the profiles.

An example rawdata profile (acquired along NW-

SE direction, y=0 m), is shown in Figure 5.

Figure 5: Example of raw GPR profile, (NW-SE direction,

y=0 m).

Following a preliminary rawdata analysis, we

decided to process only the data acquired with the 200

MHz, which were more focused on the target depths

(down to 1.4 m) and had the necessary resolution to

discriminate possible targets.

The example profile shows the main bang (main

reflection at ground-antenna interface), several

diffraction hyperbolae (10-20 ns), and some

continuous reflections (25.30 ns, at distance between

0-20 m and 30-50 m). At higher times, ringing signals

mask the presence of other possible reflectors.

Figure 6: Example of resulting GPR profile, (NW-SE

direction, y=0 m).

Each profile has been processed for reducing the

noise and extracting the significant signal features.

Low-frequency noise filter, move start-time, trace

migration and band-pass filter have been applied to

extract the resulting profiles. An example is shown in

Figure 6. White features correspond to the stronger

amplitudes of reflections, which can be associated to

the anomalies present at the site.

Time-to-depth conversion, for each profile, has

been done by imposing a wave velocity of 0.08 m/ns.

This value is a reasonable approximation of GPR

wave velocity in sedimentary soils, (fine sands with

high water content, silty sands, silts, etc.). The Italian

Geological Survey identifies the soil at Pollenzo site

as clayey-sandy alluvial sediments (Carta Geologica

d’Italia, Foglio n° 68, Carmagnola, 1:10000).

3.3 Results and Comparison

The processed profiles have been combined in 3D

volumes of amplitude of reflection of the GPR

signals, and several horizontal time-slices have been

extracted. From the analysis of data, the following

considerations can be made:

 Incoherent reflections and diffractions are present

the depth range between 0 and 0.8 m, caused by

coarse gravel, blocks and the roots;

 In the depth range between 0.9-1.3 m, a high-

amplitude area is present; marked with white

lines. It is referred to as “Anomaly 1”.

 Two continuous reflections, located along WNW-

ESE direction (black line), and SW-NE direction

(red line) can be identified. These two anomalies

are referred to, respectively, as “Anomaly 2” and

“Anomaly 3”.

 At higher depths, below 1.5 m from the surface,

no significant anomalies can be identified. This is

caused by the limited signal penetration in this

particular context.

The spatial reference of time-slice was related during

the analysis phase to a local coordinate system, with

the origin is located at the top left corner of the area.

Since the time-slice can be displayed ad a tiff image

format, the cartographic reference has been assigned

by a simple rototraslation function, easily manageable

and imposed in tfw file. (Figure 7).

GISTAM 2016 - 2nd International Conference on Geographical Information Systems Theory, Applications and Management

Figure 7: Superimposition of time-slice at 1.2 m of depth

on orthophoto, with main anomalies derived from

geophisical prospections highlighted.

3.3.1 Historical Cadastral Map

The more significant historical maps of new era for

Piedmont territory is the “Carta Generale de’ Stati di

S.A.R.” di G. T. Borgonio (1680). This map represent

the beginning of a systematic knowledge course of

estates by means of cartography. It correspond to a

renewed will of administrative and military control of

the land (Ricci, Carassi, 1986), but certainly the map

scale is not proper for parcels positioning. In the

political framework of strengthening of Savoy State,

starting from the beginning of the XVIIIth century,

the territorial fiscal control has been launched by the

cadastral registry and maps.

The “Catasto antico sabaudo” achieved during

the XVIIIth century underwent a precise

enhancement with the French real estate Registry of

1802-1814. This new map had to be harmonized to

the new French map of Cassini, having the same

scale.

Figure 8: French particles calastral map (1810), Section Xx

de la Comune de Brà, portion A.

It’s not possible to know many information on

map accuracy; we know that during the XVIIIth

century the Snellius triangulation has been largely

adopted and near the 60’s of the XVIIIth century the

studies aimed to the ellipsoid determination started

within the astronomical observatory of Turin in order

to calculate the degree (Gradus taurinensis Augustæ

Taurinorum, cited in (Frasca, 2004)). Moreover, as

fundamental basis of the whole Piedmont

topographical triangulation, the straight royal road

connecting Torino with Rivoli has been chosen.

The interesting aspect represented in the French

map is a different parcels partitioning as compared to

the current situation.

A seven parameters Helmert transformation (1)

has been used in order to provide a spatial reference

to the French map and compare the position of GPR

anomalies with the position of the no longer existed

parcels.

= T lR (a

, a

) u

(1)

Figure 9: Time-slice and related anomalies referenced to

UAV ortophoto, superimposing regional orthomap.

Figure 10: (upward) Transformation of French map using

points selected from regional medium scale map. (center)

Directions of residual vectors, in red, and (bottom) least

square residuals on points.

Gathering GPR Inspections and UAV Survey in Cultural Heritage Documentation Context

The best dozen of points selected on historical

map and most probably still existing, have been used

to perform the transformation. The least square

residuals calculation provide an uncertainty of points

coordinates near to 20 meters, which is exactly the

distance between anomalies and historical parcels

division. (Figure 10 and graph)

4 CONCLUSIONS

Spatial location and GIS tools have accomplished the

complete comparison among geophysical and

historical data.

The geophysical survey permitted to locate some

GPR anomalies that could have possibly been attributed

to archaeological remains. The joint geomatic-

geophysical survey, with the help of historical maps

recognition and analysis, permitted to locate and geo-refer

the geophysical anomalies, and to describe their nature

more reliably. A successive excavation activity confirmed

that the different anomalies are attributable to geological

and sedimentological discontinuities. Anomalies at point

2 and 3 are then to be linked to a paleo-channel of Tanaro

River, which had a high flow velocity and could transport

gravel particles and blocks. Anomalies at point 1, 4, 5 and

6 are attributable to sand and fine gravel accumulations

due to natural flow patterns or to anthropic channeled

flows within small irrigation channels, which served at

that time to partialize the different soil plots. This last

seem to be perfectly confirmed by historical map.

Many times the use of GPS/GNNS survey and

polynomial transformations enable to reduce the

effects of deformation of sheet maps and those due to

the different systems of representation and

measurement (Baiocchi and Lelo, 2010). Furthermore,

on these basis many project of accessibility of

historical cartographic heritage through the web have

been fulfilled enabling the visualization from Web

services. (e.g. Brovelli et al., 2012).

The enhancement of points residuals after

historical maps transformation is surely achievable by

ground measures, but this is profitable if the historical

maps are recent and drawn by systems which are

comparable with modern methods.

Some considerations upon methods can be

underlined: now the image standard formats are highly

interoperable, since geophysical, historical maps, and

UAV products can be visualized in the same reference

system. UAVs related to SfM technique provide 3D

spatial data remarkably improved in the large scale

environment and their products fit a very large set of

needs.

ACKNOWLEDGEMENTS

Diego Franco from ENGEL-Environmental-

Engineering Geophysics Laboratory took part to the

field GPR survey. The UAV flights have been

performed by the DiRecT Team (Disaster Recovery

Team), involving Aicardi, I. Boccardo, P. Chiabrando,

F. Donadio, E. Lingua, A. Maschio, P. Noardo, G.

Sammartano, F. Spanò.

REFERENCES

Balletti, C., 2006. Georeference in the analysis of the

geometric content of early maps. e-Perimetron 1(1), pp.

32-42.

Baiocchi ,V., Lelo, K., Accuracy of 1908 high to medium

scale cartography of rome and its surroundings and

related georeferencing problems,Acta Geodaetica et

Geophysica Hungarica, 2010, 45,1, 97-104.

Brovelli, M. A. Minghini, M. Zamboni, G. 2012.

Valorisation of Como historical cadastral maps

through modern web geoservices. ISPRS Annals, I-4,

pp. 287-292

Frasca F., 2004, Le attività degli ingegneri geografi

francesi nei territori italiani in età napoleonica,

Informazioni della Difesa, I.

Haala, N., Cramer, M., Weimer, F., Trittler, M., 2011.

Performance test on UAV-based photogrammetric data

collection. ISPRS Archives, Vol. XXXVIII-1/C22.

Livieratos, E., 2006. On the study of the geometric

properties of historical cartographic representation.

Cartographica 41(2), pp. 165-175.

Lowe, D., 2004. Distinctive Image Features from Scale-

Invariant Keypoints, Int. Journal of Computer Vision, Vol.

60, pp. 91-110.

Cosenza, P., Camerlynck, P., Tabbagh, A., 2003.

Differential effective medium schemes for investigating

the relationship between HF relative dielectric

permittivity and water content of soils. Water

Resources Research 39, 1230.

Davis, J.L., Annan, A.P., 1989. Ground-penetrating radar

for high resolution mapping of soil and rock

stratigraphy. Geophysical Prospecting 37, 531–551.

Friedman, S.P., 1998. A saturation degree-dependent

composite spheres model for describing the effective

dielectric constant of unsaturated porous media. Water

Resources Research 34, 2949–2961.

Reynolds, J.M., 1997. An Introduction to Applied and

Environmental Geophysics. John Wiley & Sons, NY.

Ricci Massabò, I, Carassi M., 1986, Amministrazione dello

spazio statale e cartografia nello stato sabaudo, in

MIBAC, Uff. Beni Archivistici, (ed), Cartografie e

Istituzioni di età moderna, Genova.

F. Rinaudo, F. Chiabrando, A. Lingua, A.T. Spanò. 2012.

Archaeological site monitoring: UAV photogrammetry

can be an answer. In: ISPRS, vol. XXXIX n. B5, pp.

583-588.

GISTAM 2016 - 2nd International Conference on Geographical Information Systems Theory, Applications and Management

Sen, P.N., Scala, C., Cohen, M.H., 1981. Relation of certain

geometrical features to the dielectric anomaly of rocks.

Geophysics 46 (12), 1714–1720.

Spano' A.; Bonora V.; M.C. Preacco, 2007, Geomatic

contributions to archaeological investigations.The case

of Torrione of Pollenzo (Piedmont – Italy), ISPRS

Archives, Vol. XXXVI-5/C53.

Szeliski R., 2011. Computer Vision: Algorithms and

Applications, Springer, London. pp. 181-207, 303-332.

Gathering GPR Inspections and UAV Survey in Cultural Heritage Documentation Context