A Toolset to Detect and Classify Active Deformation Areas using

Interferometric SAR Data

Jos

e A. Navarro

, Mar

ıa Cuevas

, Roberto Tom

, Anna Barra

and Michele Crosetto

Centre Tecnol

ogic de Telecomunicacions de Catalunya (CTTC/CERCA),

Av. Carl Friedrich Gauss 7, 08860 Castelldefels, Spain

Dpto. de Ingenier

ıa Civil, Escuela Polit

ecnica Superior de Alicante, Universidad de Alicante,

P.O. Box 99, E-03080 Alicante, Spain

Keywords:

Ground Deformation Analysis, Ground Deformation Classiﬁcation, Process Automation.

Abstract:

The MOMIT project is targeted at showing how remote sensing techniques may help to monitor and then

maintain railway infrastructures. This project has deﬁned several demostrators to fulﬁll such goal. The au-

thors have been involved in the design and development of several sofware tools needed to implement the

ﬁrst demonstrator, dealing with ground movements nearby the railway infrastructures. Among these tools,

ADAﬁnder, ADAclassiﬁer and los2hv have been developed at the CTTC. The aim of ADAﬁnder is to detect

—and update— areas where active deformation processes are taking place; ADAclassiﬁer is targeted at iden-

tifying the kind of processes undergone by such areas; ﬁnally, los2hv is a tool computing the horizontal and

vertical components of the movement measured along a line of sight. This paper takes care of describing these

three application, in the context of the MOMIT project, not forgetting the beneﬁts derived from the automation

of the methodologies they rely on. The formal testing process of the tools as well as some results obtained

with real datasets are also presented.

1 INTRODUCTION

The MOMIT project—(Multi-scale Observation and

Monitoring of railway Infrastructure Threats), see

(MOMIT, 2017) for details—aims at developing and

demonstrating a new use of remote sensing technolo-

gies for railway infrastructures monitoring. MOMIT

solutions will mainly aim at supporting the mainte-

nance and prevention processes within the infrastruc-

ture management lifecycle. The overall concept un-

derpinning MOMIT project is the demonstration of

the beneﬁts brought by Earth Observation and Re-

mote Sensing to the monitoring of railways networks

both in terms of the infrastructure and of the surround-

ing environment, where activities and phenomena im-

pacting the infrastructure could be present. MOMIT

will leverage on state of the art technologies in the

ﬁelds of space-based remote sensing and RPAS (Re-

motely Piloted Aircraft Systems) based to perform

different kind of analysis thanks to the wide variety

of sensors they could be equipped with.

To achieve its goals, six demonstrators showing

how these data and technologies may contribute to

such objectives are being built, namely:

1. Ground movements nearby the infrastructure.

2. Hydraulic activities nearby the track.

3. Global supervision for natural hazards.

4. Electrical system monitoring.

5. Civil engineering structures monitoring.

6. Safety monitoring.

The Division of Geomatics of the CTTC (Cen-

tre Tecnol

ogic de Telecomunicacions de Catalunya)

takes care of building some of the components inte-

grating the ﬁrst demonstrator, whose objectives are

detailed in (MOMIT, 2017). The goal most relevant

to the work presented here is to use the PSI (Persis-

tent Scatterer Interferometry) method for monitoring

the sites of interest, allowing the measurement of dif-

ferential deformation of the ground and each single

structure, with millimetric / centimetric precision.

Several software applications are needed in the

context of this demonstrator. This paper describes

those developed at CTTC, namely, ADAﬁnder, ADA-

classiﬁer and los2hv. The ﬁrst one, ADAﬁnder, is

a tool to detect and update ADAs (Active Deforma-

tion Areas) using Sentinel-1 imagery and PSI. The

Navarro, J., Cuevas, M., Tomás, R., Barra, A. and Crosetto, M.

A Toolset to Detect and Classify Active Deformation Areas using Interferometric SAR Data.

DOI: 10.5220/0007617701670174

In Proceedings of the 5th International Conference on Geographical Information Systems Theory, Applications and Management (GISTAM 2019), pages 167-174

ISBN: 978-989-758-371-1

167

main goal of such application is to update and as-

sess the geohazard activity (volcanic activity, land-

slides or ground subsidence among other phenomena)

of a given area. ADAclassiﬁer goes a step beyond

ADAﬁnder, trying to classify the kind of deforma-

tion process undergone by the ADAs — that is, is the

detected deformation process a landlisde, a sinkhole,

or what else? Finally, los2hv computes the horizon-

tal and vertical components of the movement mea-

sured along the LOS (Line Of Sight). The horizon-

tal component of such movement is one of the inputs

of ADAclassiﬁer; therefore, a tool to compute it was

needed.

2 AUTOMATING

METHODOLOGIES

The three tools presented in this paper rely in preex-

isting methodologies, that is, these applications have

automated a set of procedures that already existed.

In the case of ADAﬁnder, a methodology to per-

form the identiﬁcation and assessment of ADAs was

presented in (Barra et al., 2017) back in 2017. In that

work, the authors explain in detail the procedure to

identify the active deformation areas and also to as-

sess the certainty of such ﬁndings; the set of points

making each ADA as well as its area of inﬂuence and

the quality index stating the goodness of the assess-

ment are the most relevant outputs. The input, the set

of PSs (Persistent Scatterers, aka the points) covering

the area to analyze.

ADAclassiﬁer relies on a methodology that is un-

der active reseach—although mature enough as to be

automated, thus easing the experimentation in this

area and helping to improve the methodology itself.

For more details, see (Tom

as et al., 2018).

Finally, the work described in (Notti et al., 2014)

is the base of the implementation of los2hv.

The aforementioned procedures usually rely on

the heavy use of a GIS (Geographic Information Sys-

tem) tool and the expertise of its operator; there-

fore, these are time-consuming, error-prone pro-

cesses, which require qualiﬁed human resources.

This paper presents an implementation of the three

methodologies and algorithms above, incarnated as

the above-mentioned ADAﬁnder, ADAclassiﬁer and

los2hv applications. The goals of such implementa-

tions are (1) to automate the respective procedures

to avoid unnecessary human errors, (2) to reduce the

time needed to identify and preclassify the ADAs,

thus opening the door to more frequent updates and

analysis, and (3) to reduce the expertise required to

obtain such results, being possible to integrate the

process in an semi-automated production workﬂow,

if necessary and possible.

3 THE APPLICATIONS

3.1 ADAﬁnder

The ﬁrst tool in the set is ADAﬁnder. This application

is used to identify those areas that are undergoing a

deformation process, whatever it is. Additionally, an

assessment of the quality of the detection process is

performed.

As inputs, it takes:

• An ESRI shapeﬁle containing the PSs that will be

used by the detection algorithm. ADAﬁnder needs

the attributes related to these PSs to be, at least,

the coordinates of the points, their velocity, and

the deformation time series measuring the move-

ments undergone by the PSs at various times in

the past.

• Optionally, another shapeﬁle deﬁning a polygon

delimiting the area of interest to be analyzed. All

PSs in the input shapeﬁle are taken into account

when such polygon is not provided.

• The set of parameters deﬁned by the methodology

described in (Barra et al., 2017).

The output is made of two ESRI shapeﬁles, the

ﬁrst one containing the polygons deﬁning the bound-

aries of the ADAs; the other contains the set of PSs

passing the set of ﬁlters set by the methodology.

Figure 1 depicts the user interface of the GUI

(Graphical User Interface) version of the ADAﬁnder

tool.

The output ADA shapeﬁle includes among its at-

tributes the results of the quality assessment of the de-

formation detection process. Such asessment is repre-

sented by a four-level variable whose values measure

the certainty (”very sure”, ”sure”, ”not so sure” and

”not sure at all”) of the deformation detection pro-

cess. Figure 2 shows some ADAs; their colours cor-

respond to just two of the certainty levels described

(”very sure”, green, ”sure”, blue).

For more details on the ADAﬁnder tool, please

refer to (Navarro et al., 2018a) and (Navarro and

Cuevas, 2018b).

3.2 ADAclassiﬁer

The second application in the toolset is ADAclassi-

ﬁer. This tool takes the output of ADAﬁnder, that is,

GISTAM 2019 - 5th International Conference on Geographical Information Systems Theory, Applications and Management

168

Figure 1: The GUI version of the ADAﬁnder tool.

Figure 2: An example of the output of ADAﬁnder. Colours

are used to show the quality assessment.

a set of ADAs, and tries to determine the kind(s) of

deformation process(es) undergone by each of them.

Up to six different kinds of deformation phenom-

ena are detected. These are: landslides, sinkholes,

land subsidences, constructive or consolidation set-

tlements, expansive soils and temperature effects. A

different algorithm (or sub-methodology) has been

devised for each of these. For instance, Figure 3

roughtly depicts how landslides are detected.

Being able to detect so many kind of deformation

processes implies that ADAclassiﬁer needs a substan-

tial number of inputs. These are:

• The ADAs and PSs ﬁles created by ADAﬁnder

(see section 3.1). Obviously, these ﬁles need

not to be created by ADAﬁnder, but the set

of attributes must match those required by

ADAclassiﬁer—and that ADAﬁnder does include

Figure 3: Algorithm to detect landslides.

in its output.

• A DTM (Digital Terrain Model), to compute

slopes.

• A series of polygon vector maps (inventories from

now on,) in the form of ESRI shapeﬁles, to check

whether an ADA has already being catalogued as

belonging to any of the six aforementioned de-

formation processes. The required inventories are

those for landslides, sinkholes, land subsidences,

infrastructures. A geologic map (another polygon

vector map) is also needed.

• Again, an ESRI (polygon) shapeﬁle storing the

horizontal component of the movement for the

study area. This is, typically, the output of los2hv

(see section 3.3 for details).

• Finally, the set of parameters—typically

thresholds—needed by the different algo-

rithms in charge of the classiﬁcation processes

must be supplied. Examples of such parameters

are slopes, Pearson coefﬁcients to state whether

some statistical check is positive, or the minimum

percentage of overlap of an ADA and the poly-

gons in some inventory to consider that they do

intersect.

The output of ADAclassiﬁer is another ﬁle with

ADAS, where the set of attributes characterizing each

of them has been extended to include six more. Each

of them state the possibility that the ADA belongs

to the corresponding deformation process. This is

so because all the detection algorithms are applied

to each ADA. Therefore, and although incongruous,

some ADAs might be classiﬁed as positives in more

than one deformation process.

Four levels of certainty are deﬁned by the method-

ology: ”it is an X”, ”it could be an X”, it is not an X”

and ”X has not been checked”, where ”X” stands for

A Toolset to Detect and Classify Active Deformation Areas using Interferometric SAR Data

169

each of the six deformation detection processes exe-

cuted by ADAclassiﬁer—for instance, when talking

about landslides, ”it is a landslide” would be one of

the values of the corresponding attribute.

Note that one of the four values is ”not checked”.

This is so because ADAclassiﬁer may decide not to

apply one or more of the six detection processes be-

cause of the lack of data. As stated above, a notice-

able number of inputs is required. Just the inventory

ﬁles already amount to ﬁve. Taking also into account

the DTM and the horizontal components of the move-

ment, it is easy to realize that many will be the situa-

tions when the full set of ﬁles will not be available.

To alleviate this usual problem, ADAclassiﬁer

makes optional almost every input ﬁle (this includes

the DTM, the inventories and the horizontal compo-

nents). Each time the application is run, it analyzes

the dependencies of each sub-algorithm and decides

which of these may be executed as a function of the

inputs provided by the user. Consequently, it is nec-

essary to add the ”not checked” value as one of the

possible outputs of each classiﬁcation process. Note

that this behaviour makes the tool much more ﬂexible:

while concentrating the detection of six deformation

processes in a single application, it may be used to

check only one of these, just providing the available

set of data.

The ADAclassiﬁer tool is described in more detail

in (Navarro et al., 2018b).

3.3 los2hv

The los2hv tool is targeted at the computation of the

separate horizontal and vertical components of the

ground displacement measured with PSI technolo-

gies along the satellite’s LOS. Both ascending and

descending datasets—that is, ﬁles with PSs derived

from SAR (Synthetic Aperture Radar) images cap-

tured when the satellite is traveling from south to

north and from north to south respectively—are re-

quired. los2hv, in its current version, is not able

to compute such horizontal and vertical components

when only one dataset (ascending or descending) is

available.

The tool accepts (input) and produces (output)

ESRI shapeﬁles for compatibility reasons. On out-

put, los2hv creates two ﬁles, to store, respectively, the

horizontal and vertical components of the movement

as observed along the LOS.

los2hv performs a tessellation of the whole area

covered by the PSs included in the two input ﬁles.

The size of the tesserae (grid spacing) is decided by

the user. As a consequence, each PS belongs to one

of the resulting tesserae.

Figure 4: los2v: tesserae and ascending and descending

PSs.

There may be tesserae where (1) there are no PSs,

(2) there are only PSs from the ascending input ﬁle,

(3) there are only PSs from the descending input ﬁle

and, ﬁnally (4) there are PSs from both input ﬁles (see

Figure 4). For those tesserae of type (4), the ground

movement is averaged. The resulting amount, mea-

sured along the LOS, is then converted to horizon-

tal and vertical components according to the formulae

described in (Notti et al., 2014). The value of the hor-

izontal and vertical components of the ground move-

ment corresponds now to the whole area covered by

the tessera where the points involved in the computa-

tion where located.

These values are saved to the corresponding out-

put ﬁles. Note that the user may select to store

points (centroids of the tesserae with data) or squares

(boundaries of the tesserae). The second kind of out-

put (squares) is the one used by ADAclassiﬁer. The

centroid output is provided for whatever other pur-

poses.

For more information about los2hv, please refer

to (Navarro and Cuevas, 2018a).

4 IMPLEMENTATION AND

INTEGRATION

4.1 The Language of Choice

The toolset has been implemented in C++ to boost

performance. Other popular languages, as Python,

have been avoided precisely for that reason. Although

developed using Microsoft’s Visual Studio, special

precautions have been taken to make the source code

portable, particularly for the most popular C++ com-

piler used in the Linux operating system, i.e. gcc.

Choosing C++ opened the door to using several

GISTAM 2019 - 5th International Conference on Geographical Information Systems Theory, Applications and Management

170

open source libraries on which the toolset relies. The

most relevant ones are:

• Qt. Although it has been used for several pur-

poses, the main target was to provide a portable

GUI for the three applications.

• Shapelib. This library is a very convenient tool to

read and write ESRI shapeﬁles.

• Clipper. A library for clipping and offsetting lines

and polygons.

The selection of the programming language has

direct implications on how the logic of these appli-

cations may be embedded in other software compo-

nents. See section 4.2 for further details.

4.2 Three Flavours

Each of the applications described in section 3 are

available in three different incarnations or ﬂavours:

• As a C++ class in a library (one for each applica-

tion). Third party (C++) sofware willing to embed

the logic of ADAﬁnder, ADAclassiﬁer or los2hv

as a black box, only needs to instantiate the cor-

responding class. Thus, embedding the necessary

logic to be able to identify or classify ADAs or to

compute the horizontal components of the move-

ment is just one procedure call away. Obviously,

only sofware components developed in C++ will

be able to integrate the logic in the library.

• As a command-line utility. This makes possible to

integrate these tools in batch workﬂows, since no

human intervention is required to run them. See

section 4.3 for details on options ﬁles, the mech-

anism used to obtain the information controlling

the behaviour of the applications.

• As an application featuring a GUI. This ﬂavour is

the best one for experimenting because of its ease

of use. GUI-based applications, however, cannot

be integrated in batch workﬂows.

Note that the command-line as well as the GUI

versions of the applications are just interfaces calling

the classes in the library that actually implement the

logic of the processes.

4.3 Option Files

Both the command-line and GUI versions of the ap-

plications in the toolset rely on option ﬁles to retrieve

the information deﬁning how to proceeed—input or

output ﬁles, thresholds controlling some condition,

etc. Note that this is so for the GUI-based tools too;

in fact, the GUI is just a mechanism to ﬁll the gaps in

an option ﬁle template. This simpliﬁes the design of

the classes implementing the logic, since only one in-

terface (the option ﬁle) needs to be taken care of. The

commmand-line incarnations of the three tools have a

single parameter: the name of the options ﬁle with the

program’s parameters.

The option ﬁles used by the toolset are uncom-

plicated, plain text ﬁles including pairs of labels and

values:

POSITION_X = 5

4.4 Real-life Shapeﬁles

ESRI shapeﬁles, although standardized, may include

variable sets of attributes. Even when a shapeﬁle in-

cludes the full set of attributes needed by a tool (for

instance, the x and y coordinates as well as the ve-

locity or the deformation time series in the case of

ADAﬁnder), they may appear in different columns of

the attribute (.dbf) ﬁle. This usually depends on the

provider of the ﬁles.

This variability might become a serious problem,

since the input module should be adapted for each

kind of shapeﬁle to process.

The solution to avoid this problem are the ”read-

map ﬁles” deﬁning how the relevant attributes in a

shapeﬁle are organized. Of course, the attributes

needed by the tools to work properly must always

be present in the ﬁles, but thanks to these read-map

ﬁles it is possible to deal with changes in the positions

where these appear.

Thus, a read-map ﬁle is just an extra options ﬁle

(see section 4.3) where, by means of couples of labels

and values, the positions of the attributes on which

each tool rely are speciﬁed.

Such a simple mechanism makes the toolset much

more ﬂexible and able to deal with real-life datasets.

5 QUALITY ASSURANCE

Prior to the delivery of the tools to the MOMIT con-

sortium, these went through a rigorous testing pro-

cess to guarantee that they worked correctly. Note

that both a test plan and a test report concerning the

whole toolset are described, respectively, in (Navarro

et al., 2018c) and (Navarro et al., 2018d). There, the

full details on how testing was devised and took place

are provided. However, and for the sake of complete-

ness, the next sections brieﬂy presents the most rele-

vant steps taken to guarantee the quality of the code.

In the case of ADAﬁnder, the manual methodol-

ogy had been used for some time in GIS environments

when this tool was developed, so datasets including

A Toolset to Detect and Classify Active Deformation Areas using Interferometric SAR Data

171

both inputs and outputs were available. Therefore, the

tests consisted essentially in comparing the results of

the manual procedure with those created by the tool.

A mechanism to quickly compare the results pro-

duced by the manual and automated solutions was

devised; basically, it consisted on exporting the val-

ues of the attributes to check for both outputs (man-

ual and automated), sorting these to easily match the

attributes in each ﬁle and then computing the differ-

ences of their values, which, in all cases, were un-

der the threshold set by the precision of the typical

8-byte IEEE 754 double (around the 15th decimal po-

sition). This means, from the numerical standpoint,

that the results (for instance, the coordinates of the

output points) were equivalent. The values of the at-

tributes standing the level of certainty were checked

for absolute equality since these were represented by

integer (non-ﬂoating point) magnitudes.

The situation of ADAclassiﬁer and los2hv was

different; no previous results existed, so no reliable

dataset to compare their outputs was at hand. The

approach, therefore, was to create synthetic datasets

for both applications. The rationale behind the way

these datasets were created was to organize the infor-

mation in easily identiﬁable geometric patterns, so,

when combined, the area(s) where positive results

lie is (are) also predictable. For instance, all ADAS

were identical square-shaped polygons distributed in

a 24×24 ADAs checkerboard pattern (see Figure 5a);

the horizontal components of the moment were dis-

tributed in 4 adjacent vertical stripes covering 6×24

ADAs each and setting a known value to each of them

(0, 10, 20 and 30 mm/year respectively) as shown in

Figure 5b. Other easily identiﬁable patterns were used

for the rest of input ﬁles.

Then, for example, assuming that some classiﬁca-

tion process needs the value of the horizontal compo-

nent of the movement to be higher than 25 mm/year,

then only the rightmost column made of 6×24 ADAs

will meet the requirement. This may be seen in Fig-

ure 5c where the ADAs and the horizontal compo-

nents of the movement (ﬁgures 5a and 5b) have been

overlapped.

6 PERFORMANCE EVALUATION

Only one of the applications developed and tested,

ADAﬁnder, relies on a procedure for which previous

results already existed. Therefore, it was the only one

for which a performance reference was available to

compare with. Since such procedure was executed

manually (a series of steps performed by an opera-

tor using the tools offered by a GIS) a noticeable im-

Figure 5: ADAclassiﬁer: synthetic dataset samples.

Table 1: Performance of the several tools.

Tool Dataset Time (s)

ADAﬁnder 20,351 PSs. 2

ADAﬁnder 926,916 PSs. 179

ADAclassiﬁer 144 ADAs, 3600

PSs, between 4 - 8

polygons per

inventory, DTM with

1441×1441 z values.

125

los2hv 2 × 135 PSs. Grid:

7×7 tesserae.

provement of performance was expected due to au-

tomation.

There were no performance references for the two

other tools (los2hv and ADAclassiﬁer) so no spe-

ciﬁc expectations about performance improvements

existed. All tests took place using a computer with

the following characteristics: Windows 10 64-bit, In-

tel Core i5-5300U @ 2.3 GHz, 2 cores, 4 threads, 8

Gb RAM, 500 Gb magnetic (non-SSD) hard disk. Ta-

ble 1 shows the performance of the three tools.

Note that in the case of ADAclassiﬁer, the pro-

cess includes the identiﬁcation of six different classes

of deformation phenomena (that is, the test covers all

the possible classiﬁcation processes available). Test-

ing for fewer options will reduce the time needed to

process data.

GISTAM 2019 - 5th International Conference on Geographical Information Systems Theory, Applications and Management

172

Figure 6: ADAﬁnder. Results for the Tenerife dataset

(source: (Barra et al., 2017)).

7 REAL TEST CASES

As stated in section 5, two real datasets were used to

check the correctness of the ADAﬁnder application.

The ﬁrst one corresponds to Tenerife, one of the Ca-

nary Islands; the other one, the Gl

ories dataset covers

one of the neighbourhoods of the city of Barcelona.

The results related to the most representative one,

Tenerife, are brieﬂy presented here. Gl

ories is a very

small dataset, containing only 20351 PSs only and

covering a much smaller area, so it was used primarly

to test the correctness of the application taking almost

no time to obtain results (about 2 s); it will not be

presented here.

The main parameters deﬁning the dataset and how

it was processed follow (please, refer to (Barra et al.,

2017) for a detailed description of their meaning):

• Number of PSs in the input ﬁle: 926,916.

• Standard deviation of the velocity of the whole set

of PSs (σ

): 2.35 mm/year. The factor to apply to

to obtain the stable/ moving threshold: 2.

• Points are considered isolated when having only 1

or no neighbours.

• Mimimum set of a cluster of points to be an ADA

candidate: 5.

• ADA radius: 40 m.

The results are shown in Figures 6a and 6b. The

left side of the ﬁgure depicts the distribution of PSs

and their velocities. On the right side, the 72 ADAs

detected by ADAﬁnder are shown. Note the QI (qual-

ity index) values is measuring the reliability (cer-

tainty) of the detection process. The meangins of

these values are: (1, red): very reliable; (2, orange):

reliable; (3, yellow): not so reliable; (4: green): not

reliable.

It is important to remark that the results obtained

by ADAﬁnder with this (and any other real) dataset

are the same than those obtained when computing the

results in a manual way.

Figure 7: Location of the test sites for the ADAclassiﬁer

and los2hv tools.

The situation of ADAclassiﬁer and los2hv is dif-

ferent. These applications are much newer and no

previous real datasets were at hand, so, at the time

of writing this paper, still no actual results based on

real datasets were available. Nonetheless, the Univer-

sity of Alicante, in cooperation with the University of

Florence, and using datasets kindly provided by this

last institution, is currently and actively testing these

tools. The areas covered are Zeri, Abbadia San Salva-

tore and Campiglia Marittima, all of them located in

Italy (see Figure 7).

Unlike formal tests, where it is possible to predict

the outcomes of the tools because of a controlled set

of input data, real datasets pose a much more difﬁcult

challenge due to the complexity of the information

they contain. Therefore, the expertise of seasoned ex-

perts in the ﬁeld is required; validating the correctness

of the output is a slow task; furthermore, obtaining

data to check the six different deformation processes

is not easy—inventories, for instance, are very hard to

ﬁnd.

However, and although it is not possible to pub-

lish any deﬁnitive conclusions now, the preliminary

results obtained up to now are very promising—

conﬁrming, at least for the time being, the positive

results of the formal tests.

8 CONCLUSIONS

The toolset presented in this paper, in either of its

three incarnations (library, command-line or GUI ap-

plication) is an efﬁcient way to automate the detection

and classiﬁcation of ADAs in very short times, free-

A Toolset to Detect and Classify Active Deformation Areas using Interferometric SAR Data

173

ing their users from the error-prone, manual process

required when a GIS tool is used to put into practice

the underlying methodologies—(Barra et al., 2017;

Tom

as et al., 2018). For instance, the ADAﬁnder

tool, for which former references existed, is at least

10 times faster than a regular GIS operator.

The tools are also ﬂexible, being able to adapt

themselves to variations in the format of the input

shapeﬁles thanks to the use of the so-called ”read-map

ﬁles.”

Speed, ﬂexibility and adequate GUI interfaces

make the toolset a very apt tool for research activ-

ities, where different scenarios—usually deﬁned by

variations on the input parameters—may change the

results. On the other side, the availability of the

command-line versions of the tools makes possible to

integrate them in more production-oriented environ-

ments, thanks to the use of option ﬁles. Furthermore,

it is possible to use the toolset as a black box to in-

clude it in higher level software components, since

C++ classes have been included in the library imple-

menting the logic of the three applications. Last, but

not least, the tools, in their GUI versions, may be

integrated quite easily in either ArcGIS or Quantum

GIS—there is no need to modify the source code; this

may be done using the mechanisms provided by these

GIS tools. In this way, the applications become an

extra set of tools available in the regular work envi-

ronment of a GIS operator.

The toolset has undergone a very strict testing pro-

cess, using either real datasets and the comparison of

manual and automated results to check their validity

or synthetic ones when these were not available—that

is, for ADAclassiﬁer and los2hv. In this case, how-

ever, the software is being actively tested nowadays,

and, although no results are available yet, the partial

outcomes are promising, thus conﬁrming, at the time

being, the positive results of the formal validation pro-

cess.

ACKNOWLEDGEMENTS

This work has received funding from the Shift2Rail

Joint Undertaking under the European Union’s Hori-

zon 2020 research and innovation programme, with

grant agreement No 777630, project MOMIT, “Multi-

scale Observation and Monitoring of railway Infras-

tructure Threats”.

REFERENCES

Barra, A., Solari, L., B

ejar-Pizarro, M., Montserrat, O.,

Bianchini, S., Herrera, G., Crosetto, M., Sarro, R.,

Gonz

alez-Alonso, E., Mateos, R. M., Lig

erzana, S.,

opez, C., and Moretti, S. (2017). A methodology to

detect and update Active Deformation Areas based on

Sentinel-1 SAR images. Remote Sensing, 9(10):1002.

MOMIT (2017). Multi-scale observation and monitor-

ing of railway infrastructure threats. Home | MOMIT

project consortium. http://www.momit-project.eu/.

Web page; accessed 4-December-2018.

Navarro, J. A. and Cuevas, M. (2018a). los2hv -

Compute horizontal and vertical components of the

movement. http://www.momit-project.eu/results-and-

publications. MOMIT project technical note. Online;

accessed 4-December-2018.

Navarro, J. A. and Cuevas, M. (2018b). ADAﬁnder - A

tool to automate the detection of Active Deformation

Areas. http://www.momit-project.eu/results-and-

publications. MOMIT project technical note. Online;

accessed 4-December-2018.

Navarro, J. A., Cuevas, M., Barra, A., and Crosetto, M.

(2018a). Detection of Active Deformation Areas

based on Sentinel-1 imagery: an efﬁcient, fast and

ﬂexible implementation. In Proceedings of 18th

International Scientiﬁc and Technical Conference

(RACURS 2018), Crete, Greece.

Navarro, J. A., Cuevas, M., and Crosetto, M. (2018b).

ADAclassiﬁer - Classify ADAs automati-

cally. http://www.momit-project.eu/results-and-

publications. MOMIT project technical note. Online;

accessed 4-December-2018.

Navarro, J. A., Cuevas, M., and Crosetto, M.

(2018c). SW data processing components test

plan. http://www.momit-project.eu/results-and-

publications. MOMIT project deliverable. Online;

accessed 4-December-2018.

Navarro, J. A., Tom

as, R., Cuevas, M., and Crosetto,

M. (2018d). SW data processing components

test report. http://www.momit-project.eu/results-and-

publications. MOMIT project deliverable. Online; ac-

cessed 4-December-2018.

Notti, D., Herrera, G., Bianchini, S., Meisina, S., Garc

ıa-

Davalillo, J. C., and Zucca, F. (2014). A methodol-

ogy for improving landslide PSI data analysis. Inter-

national Journal of Remote Sensing, (35):2186–2214.

doi: 10.1080/01431161.2014.889864.

Tom

as, R., Cano, M., Pastor, J. L., and Riquelme, A. (2018).

Automatic classiﬁcation of Active Deformation

Areas. http://www.momit-project.eu/results-and-

publications. MOMIT project technical note. Online;

accessed 4-December-2018.

GISTAM 2019 - 5th International Conference on Geographical Information Systems Theory, Applications and Management

174