Occupancy Grid Map Generation from OSM Indoor Data for Indoor

Positioning Applications

Thomas Graichen

, Rebecca Schmidt

, Julia Richter

and Ulrich Heinkel

Professorship Circuit and System Design, Chemnitz University of Technology, Reichenhainer Straße 70, Chemnitz, Germany

Keywords:

Occupancy Grid Maps, OpenStreetMap, Indoor Maps, Indoor Localisation.

Abstract:

In recent years, there is a growing interest in indoor positioning due to the increasing amount of applications

that employ position data. Current approaches determining the location of objects in indoor environments

are facing problems with the accuracy of the sensor data used for positioning. A solution to compensate

inaccurate and unreliable sensor data is to include further information about the objects to be positioned

and about the environment into the positioning algorithm. For this purpose, occupancy grid maps (OGMs)

can be used to correct such noisy data by modelling the occupancy probability of objects being at a certain

location in a speciﬁc environment. In that way, improbable sensor measurements can be corrected. Previous

approaches, however, have focussed only on OGM generation for outdoor environments or require manual

steps. There remains need for research examining the automatic generation of OGMs from detailed indoor

map data. Therefore, our study proposes an algorithm for automated OGM generation using crowd-sourced

OpenStreetMap indoor data. Our experiments with nine different building map datasets demonstrate that the

proposed method provides reliable OGM outputs. The proposed algorithm now enables the integration of

environmental information into positioning algorithms to ﬁnally increase the accuracy of indoor positioning

applications.

1 INTRODUCTION

Indoor positioning has received much attention in re-

cent years due to the vast amount of applications that

employ position data. Indoor positioning is the pro-

cess of determining the location of objects in indoor

environments. Positioning systems are applied to lo-

calise and track assets in production buildings, to nav-

igate persons through indoor environments or to anal-

yse a person’s trajectory in elderly care applications,

for example. Such systems use different types of

technologies, such as inertial sensors, visual mark-

ers, cameras, time of ﬂight (ToF) sensors, or Wi-Fi-

based technologies. All these localisation techniques

have different disadvantages in indoor environments,

which lead to inaccurate localisation results. Wi-Fi

signals, for instance, can be interfered by metallic ob-

jects, ToF-based approaches require a line of sight

and inertial sensor data is prone to error accumula-

tion. Even technologies such as ultra-wideband sys-

https://orcid.org/0000-0003-1861-6033

https://orcid.org/0000-0002-9249-3812

https://orcid.org/0000-0001-7313-3013

https://orcid.org/0000-0002-0729-6030

tems with a theoretically achievable accuracy of 10

cm can be inﬂuenced by the environment, so that the

positioning error reaches values of up to 3 m. As a

consequence, the acquired sensor data can be inaccu-

rate and unreliable, which results in invalid localisa-

tions, such as persons detected within a wall.

In order to compensate localisation errors, it is

necessary to include further information about the ob-

ject to be positioned and about the environment into

the positioning algorithm. One possibility to improve

the localisation accuracy is the integration of indoor

map data: the given structure of buildings with its spe-

ciﬁc spacial dimensions, such as corridors, stairways

and doors, allows an elimination of invalid positions.

Moreover, by considering the structure of indoor en-

vironments, both probable and improbable object oc-

currences can be deﬁned. A person walking through a

building will probably not walk directly near the wall

and will deﬁnitely not pass through a wall. An incor-

rectly acquired position within a wall could then be

adjusted to a valid position next to the wall. Meth-

ods that are implementing such position corrections

are described in Section 2. The described example is

visualised in Figure 1.

168

Graichen, T., Schmidt, R., Richter, J. and Heinkel, U.

Occupancy Grid Map Generation from OSM Indoor Data for Indoor Positioning Applications.

DOI: 10.5220/0009348501680174

In Proceedings of the 6th International Conference on Geographical Information Systems Theory, Applications and Management (GISTAM 2020), pages 168-174

ISBN: 978-989-758-425-1

Corridor

Room

Figure 1: Localisation Example: Correction of Incorrect

Sensor Data by including Information about Indoor Envi-

ronment. The Blue and Grey Dots Indicate Sensor Measure-

ments Whereas Grey Dots Correspond to Invalid Positions,

E. G. within a Wall, That Can Be Adjusted to a Correct Po-

sition Indicated by Green.

The occurrence probability of a speciﬁc object within

a building can be modelled by means of so-called

occupancy grid maps (OGMs), which represent the

occupancy probability of an object on the ﬂoor plan

of a building. The OGM is modelled by a cell ma-

trix whereas each cell is a square area of the indoor

environment holding the probability of being occu-

pied. Hence, generating OGMs requires ﬂoor plans

and consequently indoor data of buildings. Since in-

door data about building had either been generally not

available or is only provided in form of Computer

Aided Design (CAD) formats, an alternative data

source was created by the OpenStreetMap (OSM)

community: indoor map data has been collected by

volunteers and now provides detailed crowd-sourced

information about the structure of buildings. These

mapping activities have been being increased in re-

cent years and have led to a wider availability of in-

door data. An overview of mapped data is listed in the

OSM Wiki (OSM-Community, 2020).

To date, little attention has been paid to the in-

volvement of OSM indoor data in OGM generation.

This paper therefore examines OSM indoor map data

as a data source for the generation of OGMs and in-

troduces a procedure to create such OGMs as an input

for indoor positioning algorithms.

The paper is structured as follows: Section 2

presents state-of-the-art methods for OGM generation

and outlines the research gap. Thereupon, Section 3

introduces the proposed procedure to generate OGMs

from OSM indoor data by illustrating the system con-

cept overview and subsequently describing the real-

isation of the single system modules. The obtained

results are presented and discussed in Section 4. Fi-

nally, Section 5 concludes the paper and gives an out-

look on future work.

2 RELATED WORK

Occupancy grid mapping was initially introduced

by Moravec and Elfes in 1985 (Moravec and Elfes,

1985). Originally, this mapping procedure was de-

veloped for noisy sonars and called “mapping with

known poses”. In literature, especially in the ﬁeld

of probabilistic robotics, occupancy grid mapping is

often referred to as the process of generating maps

from noisy and uncertain sensor data while the po-

sition of the robot with the attached sensors such as

cameras, laser range scanners and LIDAR is known

(Matthies and Elfes, 1988), (Konolige, 1997), (Thrun,

2001). In this mapping problem, the aim is to build an

occupancy map of the environment, in which the oc-

currence of obstacles is stored.

For positioning/localisation, the opposite prob-

lem has to be solved: Based on an existing map, the

position of objects shall be derived, also in the pres-

ence of noisy sensor data. In our case, the existing

map is an OSM indoor map that has to be transformed

in an occupancy grid map ﬁrst. In this context, OGM

generation is the transform of a ﬂoor plan into in-

dependent discrete cells. Each cell stores a variable

estimating the grade of its occupancy. The variable

can either be binary or continuous, stating whether the

cell is occupied or not or indicating the grade of oc-

cupancy, i. e. the occupancy probability of the object

to be localised.

Extant literature gives insight on how OSM maps

are transferred to OGMs and thereafter used for local-

isation purposes.

In their publications, Kurdej et al. present a local-

isation system for intelligent vehicles that uses OSM

outdoor map data as a-priori information (Kurdej,

2015), (Kurdej et al., 2012). This systems generates

OGMs based on OSM road and building information

and matches sensor data from optical sensors against

these OGMs.

Herrera et al. are the ﬁrst to generate OGMs from

OSM indoor maps (Herrera et al., 2013), (Herrera

et al., 2014). Their algorithm derives the OGMs from

a manually deﬁned graph network that overlays the

indoor map data. This graph consists of nodes, which

were deﬁned by empirical studies and denote proba-

ble indoor positions. However, these nodes have to be

manually added to the graph.

Naik et al. proposed OSM-based indoor data for

robot navigation and generated a primitive OGM

for that purpose (Naik et al., 2019). This genera-

tion methodology involves only a limited set of ob-

jects, namely information about rooms and corridors.

Moreover, the OGM distinguishes between only two

occupancy states.

Occupancy Grid Map Generation from OSM Indoor Data for Indoor Positioning Applications

169

Step 1:

Coordinate

System

Transformation

OSM

Indoor Map Data

in WGS 84

Indoor Map Data

in OGM

Coordinates

Step 2:

Map Data

Sorting

Ordered Indoor

Map Data

Step 3:

OGM

Rendering

OGM

OGM Resolution

Local Coordinate

System (Origin

and Orientation)

Figure 2: Concept Overview: The Algorithm for OGM Generation Is Divided into Three Main Steps.

To summarise, the presented related work either fo-

cusses on OSM outdoor data or is lacking a full au-

tomation of the OGM generation and is only cover-

ing a small part of building features. Therefore, we

propose a methodology for a highly automated gen-

eration of OGMs based on OSM indoor maps that is

involving as much as possible information about the

interior of a building.

3 METHODOLOGY

This section ﬁrst provides an overview about the al-

gorithm with its single steps and thereafter explains

the realisation of every step as well as the input and

the output data in detail.

3.1 Concept Overview

Our proposed algorithm for OGM generation is di-

vided into three steps: coordinate transformation, data

sorting and OGM rendering, as can be seen in Fig-

ure 2.

In the ﬁrst step, the input OSM map data, which

is represented in WGS-84 (World Geodetic System

1984) format has to be transformed into OGM co-

ordinates. Therefore, the coordinates of the indoor

map data are ﬁrstly converted to a metric representa-

tion and to a local coordinate system (LCS) that, to-

gether with its origin and orientation, has to be de-

ﬁned depending on the speciﬁc positioning applica-

tion that will use the OGMs. This transformation is

necessary to represent all objects, i. e. both objects

from the OSM map as well as objects to be located,

in the same local and metric coordinate system. A

transformation of indoor data into local metric format

brings the beneﬁt of compatibility to other devices.

Other devices might be industrial robots also work-

ing with a local metric coordinate system or devices

processing a given grid map. Afterwards, a transfor-

mation from the metric LCS in OGM coordinates is

applied based on a manually deﬁned OGM resolution.

Thereby, the OGM coordinates represent the cell in-

dices of the OGM.

During map data sorting, all indoor map objects

are assigned to a speciﬁc priority level that corre-

sponds to a layer in the rendering process, where the

OGM is rendered layer by layer in a ﬁxed order. This

assignment is based on the OSM tags of each geo ob-

ject. For the OGM rendering, we deﬁned a certain

order in a lookup table to achieve sensible OGM out-

puts. This is necessary, because geo objects within

OSM data sets may be overlapping. For instance,

room areas and their walls are described by two sepa-

rate geo objects, whereas the mapped area of the room

may be overlapping with its mapped walls. At this

point, the OSM mapping scheme considers walls as

a second layer over room areas, so that the mapped

wall boundaries deﬁne the real-world physical walls

above underlying room areas. Now consider an ex-

emplary rendering output of unsorted data: Rendering

the walls of a room before rendering its area would

lead to a loss of wall position information due to the

”over-rendering” of walls by the room area. This ex-

ample illustrates the necessity of a sensible data sort-

ing in accordance with OSM mapping schemes. Next

to the chronological order objects will be rendered,

the layer speciﬁes the occupancy probability for ob-

jects in that level, and the shape the object will be

rendered with.

Finally, the OGM is built by rendering the map

objects that were assigned to the speciﬁc layers with

their according probabilities.

GISTAM 2020 - 6th International Conference on Geographical Information Systems Theory, Applications and Management

170

3.2 Realisation

3.2.1 Input Data

The algorithm requires three sets of input informa-

tion: The OSM indoor map data, the origin of the lo-

cal coordinate system and the desired grid map reso-

lution. The OSM indoor map data is stored within an

XML (Extensible Markup Language) ﬁle including

indoor geo objects, such as rooms, walls, doors and

corridors, characterised by a set of nodes with longi-

tude and latitude coordinates as well as by OSM tags,

which are describing the meaning of each object.

The second required information are the position

and orientation of the LCS. The position consists of

a WGS-84 coordinate (latitude and longitude) and

the orientation is deﬁned by the rotation angle be-

tween the ordinate of the WGS-84 coordinate sys-

tem and the ordinate of the LCS. The origin of the

LCS as well as the rotation angle can be set by us-

ing the JOSM (Java-OpenStreetMap)-Editor (OSM-

Community, 2019a) with measurement functionali-

ties provided by plug-ins (OSM-Community, 2019b).

The grid maps resolution is the third input data of

this algorithm and it is speciﬁed in pixels per meter

(px/m). This resolution is used for the conversion of

geo object positions to the OGM coordinate system,

which is using pixels as units.

The described input data is parsed at the initiali-

sation of the algorithm and stored in an internal data

structure for further processing in the following com-

puting steps.

3.2.2 Step 1: Coordinate System Transformation

Because indoor positioning systems are comprised of

several system components with their own single co-

ordinate systems, it is necessary that all the different

components share the same local coordinate system

in the overall indoor positioning application. Such

components can be different kind of sensors, robots

or algorithms that further process positioning data.

For OGM generation, the input data needs to be

transformed into OGM coordinates, which then rep-

resent the originally metric dimensions of an indoor

environment in OGM pixel coordinates, as already

outlined in Figure 2. By means of a geographic li-

brary (Karney, 2013), (Karney, 2019), the indoor map

coordinates will be transferred from WGS-84 format

into a metric coordinate system by solving the in-

verse geodesic problem, which determines the short-

est route between two points on the surface of the

Earth. Thereby, the results are the metric distance

components on the latitude and longitude arc ∆x

and

∆y

between the origin of the LCS as well as a node

of a geo object at the position (x

). Afterwards

the rotation angle α of the LCS is applied to ﬁnally

transform ∆x

and ∆y

into local x and y coordinates.

Figure 3 visualises this coordinate system transforma-

tion.

Geo Object

LCS

Origin

Figure 3: Transformation of a Geo Object at Position

) given in WGS-84 Coordinates to a Metric LCS.

∆x

and ∆y

Represent the Distance Components on the

Latitude and Longitude Arc between the Origin of the LCS

and a Geo Object. α Is the Rotation Angle between the

WGS-84 Coordinate System and the LCS.

Finally, the map data in LCS coordinates is trans-

formed in OGM coordinates. This is achieved by

manually deﬁning the resolution of the OGM in pix-

els per meter (px/m) and afterwards multiplying each

LCS coordinate with this resolution value to obtain

rounded OGM coordinates in Pixels (px). The choice

of the resolution depends on the accuracy of the posi-

tioning system, whose results shall be improved. For

instance, an UWB system with an accuracy of 20 cm

can use an OGM with a resolution of 5 px/m.

3.2.3 Step 2: Map Data Sorting

As already described in Section 3.1, an ordered map

data set is necessary for generating an OGM with a

layer-wise rendering methodology that also involves

over-rendering. Though, due to the structure of the

OSM data deﬁnition, the geo objects within the in-

door map data are unsorted and might be overlapping.

Therefore, a lookup table that assigns relevant tags of

the geo objects to a speciﬁc rendering layer was cre-

ated. This lookup table is shown in Figure 4.

The ﬁrst rendering layer (L1) holds basic indoor

geo objects with the lowest limitations for position-

able objects, such as rooms, corridors, steps, stair-

ways or elevators. In that sense, lowest limitations

refers to the highest probability for a valid indoor po-

Occupancy Grid Map Generation from OSM Indoor Data for Indoor Positioning Applications

171

LayerLookup-Table

L1    Rooms,Corridors,Steps,Elevators, etc.

L2   Walls

L3   Openings: Doors, Entrances, etc.

Figure 4: Lookup Table to Assign Geo Objects by Means

of Their Application-Relevant Tags to a Speciﬁc Rendering

Layer.

sition. The second layer (L2) contains all kinds of

walls that deﬁnitely restrict the freedom of movement

of positionable objects. Openings, such as doors and

entrances, are considered separately to be placed as a

third layer (L3) atop of walls and to enable the over-

writing of limitations set by the walls.

3.2.4 Step 3: OGM Rendering

As a ﬁnal step of the OGM generation, the rendering

is performed.

Thereby, every ﬂoor of a building with its speciﬁc

geometry results in a separate OGM. Consequently, a

canvas is created for every ﬂoor and the dimensions of

these canvases are deﬁned by the lowest and highest

OGM coordinates of every ﬂoor, which designate the

canvas boundaries.

The rendering itself handles 8 bit grey scale val-

ues, which encode positioning probabilities in a range

from 0 % (0.0) to 100% (1.0). When rendering in-

door areas of Layer L1, i. e. rooms, corridors or steps,

these areas are ﬁlled with a grey scale value of 0.75

as it is shown in Figure 5b. Using a probability value

of 75 % instead of 100 % allows to subsequently add

popular paths with even higher probabilities, so the

grid map can be optimized in retrospect in case such

frequently used paths are known.

This step is followed by rendering a gradient tube

adjacent to the inner boundaries of the resulting area

of Layer L1, see Fig. 5c. By means of this gradi-

ent behaviour, a lower probability of positions at the

boundaries of indoor areas is modelled. Hence, the

occupancy probability decreases towards the walls in

the shape of a smoothed circle in our case, while the

grey value the circle is ﬁlled with is linearly reduced

with increasing radius. Other shapes, such as normal

distributions, are sensible as well. The dimensions of

the shape are relative to the size of the indoor area.

In the next step, the walls of Layer L2 are ren-

dered atop of the L1 area. Walls are also modelled

as areas and are ﬁlled with a positioning probability

of 0.0, which ensures that this area is inaccessible for

(a)

(b) (c)

(d) (e)

Figure 5: Illustration of the OGM Rendering Process. 5a:

Example OSM Indoor Map with Windows (Blue), Walls

(Grey), Openings (Red) and the Room Area Itself (Yel-

low). 5b: Resulting Area of L1.with Basic Indoor Objects.

5c: Boundary Modelling with Gradients. 5d: Rendering of

Walls in L2. 5e: Rendering of Openings in L3.

positionable objects and persons. This rendering step

is visualised in Figure 5d.

As shown in Figure 5e, Layer L3 includes open-

ings and their probability behaviour of the area around

them. The openings are rendered atop of the previous

layers and represent a positioning probability of 0.75.

Empirical experiences have shown that people enter

or leave openings in a shape similar to a funnel. In

case of openings that are accessible in both direction,

two funnels are used for rendering, so that the result-

ing shape resembles a hourglass shape. The funnels

are rendered perpendicular to the wall surrounding the

respective opening.

Finally, applying the complete algorithm to OSM

indoor data of a certain building delivers an OGM for

each level of this building, which is the output data of

the OGM generation algorithm.

4 RESULTS AND DISCUSSION

We performed experiments with 9 different building

map data sets, which have qualitatively shown that

the algorithm delivers reasonable OGMs. An example

for one of these generated OGMs, which represents a

complete ﬂoor of a building, is shown in Figure 6.

When comparing the input, i. e. the OSM indoor data,

with the rendered OGM, it can be seen that the dif-

ferent objects of the three layers are correctly repre-

sented in form of probability grey scale values. Our

study therefore proved that the proposed automated

generation of OGMs from crowd-sourced OSM in-

GISTAM 2020 - 6th International Conference on Geographical Information Systems Theory, Applications and Management

172

door data provides reliable results, provided that the

indoor environment was mapped correctly.

Nevertheless, the algorithm still has three limita-

tions, which should be contemplated in future work:

Firstly, the algorithm does not automatically evaluate

the quality and correctness of the input OSM data.

Because this data is mapped by volunteers without

any special training, the data can be very imprecise

and even necessary features such as doors may be

missing and can therefore negatively affect the result-

ing OGM. With the current implementation, no auto-

matic validation of the input data is performed, so that

a manual plausibility check of the generated OGM

had to be performed. Consequently, validating the

input data is still an open subject to be solved. Sec-

ondly, hourglasses, which are rendered with a ﬁxed

pre-deﬁned size at door positions, are not seamlessly

connected to the base probability of the indoor area.

This is because the width of the probability gradient at

the borders of the indoor area depends on the size of

this area (as noted in Section 3.2.4). Accordingly, the

size of the hourglass should be dynamically adapted

to the room size as well. Thirdly, the current place-

ment of the hourglass in narrow corners of a room

lead to an unwanted overwriting of wall information.

5 FUTURE WORK

In addition to the limitations, which have been pre-

sented in the previous section and which will be han-

dled in future, there are two further developments

planned:

As next step, we focus on the integration of the

generated OGMs in an UWB positioning systems in

order to improve positioning accuracy. Therefore, the

OGM shall be used in combination with a particle ﬁl-

ter to eliminate invalid samples. By means of practi-

cal experiments, the impact of OGMs on the accuracy

improvement will be evaluated.

Furthermore, in future versions of this method im-

plementation, points of interest (POIs) and popular

paths in buildings will be integrated in the OGM, as

already pointed out in Section 3.2.4.

The kind of POIs depends on the use case of the

intended positioning system. For instance, an indoor

navigation for museum visitors must consider the area

around paintings as places with high probabilites for

positions. Because paintings are typically mounted at

walls, the occupancy value of the OGM in such areas

must be increased.

For the deﬁnition of popular paths can be applied

both, manual as well as automated approaches: A

manual solution is to ask several persons to manu-

(a)

(b)

Figure 6: OGM Output Generated from OSM Indoor Data.

6a: Example OSM Indoor Map of a Floor in a Building. 6b:

OGM Generated from the given OSM Indoor Map Data.

ally draw paths in the map they think probable to be

frequently used. More appropriate, however, would

be automatised methods. One sensible solution is a

learning-based approach where most frequently used

paths are derived from the actual positioning output

and a kind of heat map is generated. The more detec-

tions are registered in a cell of an OGM, the higher

is the path probability, i. e. the heat, of this cell. A

further method could determine paths by applying

skeletonisation algorithms on the indoor map areas

whereas the remaining topological skeleton denotes

these paths. Finally, another option is to determine

direct paths between relevant objects, for example the

direct lines of sight from one door to other doors in a

room.

REFERENCES

Herrera, J. C. A., Hinkenjann, A., Pl

oger, P. G., and Maiero,

J. (2013). Robust indoor localization using optimal fu-

sion ﬁlter for sensors and map layout information. In

Occupancy Grid Map Generation from OSM Indoor Data for Indoor Positioning Applications

173

International Conference on Indoor Positioning and

Indoor Navigation, pages 1–8.

Herrera, J. C. A., Pl

oger, P. G., Hinkenjann, A., Maiero,

J., Flores, M., and Ramos, A. (2014). Pedestrian in-

door positioning using smartphone multi-sensing, ra-

dio beacons, user positions probability map and In-

doorOSM ﬂoor plan representation. In 2014 Interna-

tional Conference on Indoor Positioning and Indoor

Navigation (IPIN), pages 636–645.

Karney, C. F. F. (2013). Algorithms for geodesics. Journal

of Geodesy, 87(1):43–55.

Karney, C. F. F. (2019). GeographicLib.

https://geographiclib.sourceforge.io/.

Konolige, K. (1997). Improved Occupancy Grids for Map

Building. Autonomous Robots, 4(4):351–367.

Kurdej, M. (2015). Exploitation of Map Data for the Per-

ception of Intelligent Vehicles. PhD thesis, Univer-

sit

ede Technologie de Compi

egne, Compi

egne.

Kurdej, M., Moras, J., Cherfaoui, V., and Bonnifait, P.

(2012). Map-Aided Fusion Using Evidential Grids

for Mobile Perception in Urban Environment. In De-

noeux, T. and Masson, M.-H., editors, Belief Func-

tions: Theory and Applications, Advances in Intelli-

gent and Soft Computing, pages 343–350. Springer

Berlin Heidelberg.

Matthies, L. and Elfes, A. (1988). Integration of sonar and

stereo range data using a grid-based representation. In

1988 IEEE International Conference on Robotics and

Automation Proceedings, pages 727–733 vol.2.

Moravec, H. and Elfes, A. (1985). High resolution maps

from wide angle sonar. In 1985 IEEE International

Conference on Robotics and Automation Proceedings,

volume 2, pages 116–121.

Naik, L., Blumenthal, S., Huebel, N., Bruyninckx, H., and

Prassler, E. (2019). Semantic mapping extension for

OpenStreetMap applied to indoor robot navigation. In

2019 International Conference on Robotics and Au-

tomation (ICRA), pages 3839–3845.

OSM-Community (2019a). JOSM. https:

//josm.openstreetmap.de/.

OSM-Community (2019b). JOSM/Plugins/measurement

– OpenStreetMap Wiki. https://

wiki.openstreetmap.org/wiki/JOSM/Plugins/

measurement.

OSM-Community (2020). Indoor Map-

ping – OpenStreetMap Wiki.

https://wiki.openstreetmap.org/wiki/Indoor

Mapping.

Thrun, S. (2001). Learning occupancy grids with forward

models. In Proceedings 2001 IEEE/RSJ International

Conference on Intelligent Robots and Systems. Ex-

panding the Societal Role of Robotics in the the Next

Millennium (Cat. No.01CH37180), volume 3, pages

1676–1681 vol.3.

GISTAM 2020 - 6th International Conference on Geographical Information Systems Theory, Applications and Management

174