Conference PaperPDF Available

Enriching Wayfinding Instructions with Local Landmarks


Abstract and Figures

Navigation services communicate optimal routes to users by provid- ing sequences of instructions for these routes. Each single instruction guides the wayfinder from one decision point to the next. The instructions are based on geometric data from the street network, which is typically the only dataset available. This paper addresses the question of enriching such wayfinding in- structions with local landmarks. We propose measures to formally specify the landmark saliency of a feature. Values for these measures are subject to hy- pothesis tests in order to define and extract landmarks from datasets. The ex- tracted landmarks are then integrated in the wayfinding instructions. A concrete example from the city of Vienna demonstrates the applicability and usefulness of the method.
Content may be subject to copyright.
Enriching Wayfinding Instructions with Local
Martin Raubal1 and Stephan Winter2
1 Institute for Geoinformatics, University of Münster, Robert-Koch-Str. 26-28,
48149 Münster, Germany
2 Institute for Geoinformation, Vienna University of Technology, Gusshausstr. 27-29,
1040 Vienna, Austria
Abstract. Navigation services communicate optimal routes to users by provid-
ing sequences of instructions for these routes. Each single instruction guides the
wayfinder from one decision point to the next. The instructions are based on
geometric data from the street network, which is typically the only dataset
available. This paper addresses the question of enriching such wayfinding in-
structions with local landmarks. We propose measures to formally specify the
landmark saliency of a feature. Values for these measures are subject to hy-
pothesis tests in order to define and extract landmarks from datasets. The ex-
tracted landmarks are then integrated in the wayfinding instructions. A concrete
example from the city of Vienna demonstrates the applicability and usefulness
of the method.
1 Introduction
Assume that you are spending a few days as a tourist in Vienna. You have just en-
joyed a cup of coffee in one of the traditional coffee houses, the Café Diglas, and you
start thinking about dinner. Your tourist guide recommends one of the current in-
restaurants, Novelli. Unfortunately, what you get is only the address, not the best or
any route from the Café Diglas to Novelli. This is a typical scenario for a navigation
service, a special case of a location-based service. Calling the service should provide
a route guide, delivered in real time and tailored for the user’s needs, in our case a pe-
Navigation services calculate an optimal route and provide a sequence of instruc-
tions for this route. Each single instruction guides the user from one decision point to
the next. Typically, the instructions use geometric data from the street network, which
is the only dataset available. This paper addresses the question of enriching such way-
finding instructions with landmarks. Research in spatial cognition has shown that
people use landmarks during spatial reasoning and communication of routes, therefore
this question is not only of theoretical but also of high practical importance. The main
challenges are the automatic definition and extraction of appropriate salient features,
i.e., landmarks, from the available datasets. Data providers offer so-called Points Of
Interest (POI) geo-coded in spatial datasets. These POI are hard-coded and pre-
defined. A navigation service can use them or not; no method is provided to measure
the attractiveness or relevance of a POI for being a landmark for a particular user in-
volved in a navigation task.
The hypothesis of this paper is that a formal model of landmark saliency based on
perceptual and cognitive concepts (i.e., vision and commonsense knowledge) allows
for an automatic generation of route instructions, which include landmarks and are
close to human communication. The goal is to improve existing navigation services
by concepts closer to the human user, adaptive for individual users, with flexibility for
different tasks.
Section 2 gives an overview of human wayfinding and highlights the importance of
landmarks for navigation. It further describes the components of wayfinding instruc-
tions and how they are used in a navigation service. In section 3 we present the prop-
erties of features used to measure their attractiveness as landmarks and describe data
sources from which property values can be derived. Section 4 explains hypothesis
testing as the method for defining and extracting landmarks from datasets and com-
bines the individual properties to form a global formal measure of landmark saliency
for a feature. A case study in section 5 is used to demonstrate the proposed method.
The final section gives conclusions and directions for future work.
2 Wayfinding, Navigation, and Landmarks
2.1 Human Wayfinding
Human wayfinding research investigates the processes that take place when people
orient themselves and navigate through space. Theories try to explain how people find
their ways in the physical world, what people need to find their ways, how they com-
municate directions, and how people’s verbal and visual abilities influence wayfind-
ing. Allen [1] and Golledge [9] describe wayfinding behavior as purposeful, directed,
and motivated movement from an origin to a specific distant destination, which can-
not be directly perceived by the traveler. Such behavior involves interactions between
the traveler and the environment. Human wayfinding takes place in large-scale spaces
([6], [14]). Such spaces cannot be perceived from a single viewpoint therefore people
have to navigate through large-scale spaces to experience them. Examples for large-
scale spaces are landscapes, cities, and buildings.
People use various spatial, cognitive, and behavioral abilities to find their ways.
These abilities are a necessary prerequisite to use environmental information or repre-
sentations of spatial knowledge about the environment. The spatial abilities are task-
dependent and seem to involve mainly four interactive resources: perceptual capabili-
ties, information-processing capabilities, previously acquired knowledge, and motor
capabilities [1]. As for the spatial abilities, the cognitive abilities also depend on the
task at hand. Finding one’s way in a city uses a different set of cognitive abilities than
navigating in a building.
Allen [1] distinguishes between three categories of wayfinding tasks: travel with
the goal of reaching a familiar destination, exploratory travel with the goal of return-
ing to a familiar point of origin, and travel with the goal of reaching a novel destina-
tion. A task within the last category, which is also the focus in this paper, is most of-
ten performed through the use of symbolic information. Here, we concentrate on
landmark-based piloting where success depends on the recognition of landmarks and
the correct execution of the associated wayfinding instructions.
2.2 Landmarks and Navigation
Among the different meanings of landmark is that of an object or structure that marks
a locality and is used as a point of reference [21]. The concept is bound to the promi-
nence or distinctiveness of a feature in a large-scale environment or landscape ([10],
[7]). Thus the landmark saliency of a feature does not depend on its individual attrib-
utes but on the distinction to attributes of close features. Being a landmark is a rela-
tive property.
Landmarks are used in mental representations of space [28] and in the communica-
tion of route directions ([31], [19], [17]). Route directions shall provide a ‘set of pro-
cedures and descriptions that allow someone using them to build an advance model of
the environment to be traversed’ ([23], p. 293). Landmarks support the building of a
mental representation of such an advance model. Studies show that landmarks are se-
lected for route directions preferably at decision points ([12], [23]). Another study has
shown that mapped routes enriched with landmarks at decision points lead to better
guidance, or less wayfinding errors, than routes without landmarks. Furthermore, dif-
ferent methods of landmark presentations were equally effective [4].
Lovelace et al. [17] distinguish between landmarks at decision points (where a re-
orientation is needed), at potential decision points (where a re-orientation would be
possible, but should not be done to follow the current route), route marks (confirming
to be on the right way), and distant landmarks. According to [18], distant landmarks
are only used in navigation by a novice for overall guidance. We call landmarks at
decision points and route marks the local landmarks with respect to a specific route.
Lynch [18] defines landmarks as external points of reference—points that are not
part of a route like the nodes in a travel network. He characterizes the quality of a
landmark by its singularity, where singularity is bound to a clear form, contrast to the
background, and a prominent location. The principal factor is the figure-background
contrast ([32], [22]). The contrast can be produced by any property, such as unique-
ness in form or function in the local or global neighborhood. Sorrows and Hirtle [29]
categorize landmarks into visual (visual contrast), structural (prominent location), and
cognitive (use, meaning) ones, depending on their dominant individual quality. A
landmark will be stronger the more qualities it possesses.
However, a formal measure for the landmark saliency of an object is still missing.
Research is done in mainly two directions: the investigation of what objects are se-
lected as landmarks in human route directions ([5], [23]) and the test of the success of
pre-selected landmarks ([4], [8]). Little research is concerned with the identification
of salient characteristics for the choice of landmarks for a route, as for instance in the
context of car navigation by Burnett ([2], [3]). This issue is investigated more in the
domain of robotics. Robots use automatic selection of landmarks for their self-
orientation and positioning. Landmarks in this context are merely feature details, such
as vertical lines, rather than features ([33], [16]). Such concepts do not seem appro-
priate for supporting human wayfinding.
Progress in telecommunication technology allows the enrichment of environments
with beacons that can act as active landmarks by attracting nearby mobile devices
[26]. Such landmarks are not perceived directly by humans but through their interac-
tion with software. Hence, active landmarks—although they can play a role in naviga-
tion—cannot be used as a reference for human users. Also, virtual landmarks, like vir-
tual information towers embedded in a model of the real world [24] cannot serve as
reference points for human wayfinders, because they have no physical counterpart.
2.3 Wayfinding Instructions
The basic assumption of this paper is that route directions enriched by local land-
marks are easier to understand than the ones, which are only direction- and distance
based. We propose the following formal model for the test of this assumption.
A route consists of a sequence of nodes and edges.
At nodes, the traveler needs information whether she shall continue moving in the
present direction, or turn. Hence, nodes are called decision points in this model.
One can distinguish between nodes where a re-orientation has to take place and
other nodes where re-orientation is not necessary.
Orientation and re-orientation shall be referenced to:
o Landmarks as anchors. This means we need at least one landmark at each deci-
sion point. If there are more, context-dependent selection criteria need to be ap-
plied to find the best one (direction of view, means of traveling, time of day).
o Egocentric cardinal orientations (front, back, left, right), assuming the orienta-
tion of the present direction.
Along edges no orientation action needs to take place. The traveler shall move
from the start node to the end node, i.e., from decision point to decision point.
Optionally, landmarks along edges (route marks) can be used to confirm to the
traveler that she is on the right track.
With these elements a general form of directions is the following (‘[]’ denoting re-
quired elements, ‘{}’ optional elements, ‘UPPER CASE’ language elements, and
‘lowercase’ variables):
[AT landmarki] +
{ONTO street name} +
{(PASSING | CROSSING) landmarkj}0…n +
[UNTIL landmarkk].
with ijk. There shall be no reference to distance information or cardinal directions.
Such survey information might nevertheless be useful for wayfinders and can easily
be integrated if needed. All that is left now is the automatic extraction of suitable
landmarks for use in these route directions.
3 Measures for the Attractiveness of Landmarks
The main contribution of this paper is a formal model of landmark saliency, which in-
cludes measures for the attractiveness of landmarks. In order to determine whether a
feature qualifies as an attractive landmark we specify properties, which determine the
strength of a landmark. In this section we identify such measures by taking into ac-
count the three types of landmarks as proposed by Sorrows and Hirtle [29]. Following
their framework we presume that the visual, semantic, and structural attraction of fea-
tures in geographic space determine their use as landmarks in human spatial reasoning
and communication. The final part of the section describes the necessary data sources
from which to derive values for the measures.
3.1 Visual Attraction
Landmarks qualify as visually attractive if they have certain visual characteristics
such as a sharp contrast with their surroundings or a prominent spatial location. Our
formal model of landmark saliency includes four measures regarding visual attraction:
Façade area, shape, color, and visibility. Table 1 shows the individual properties for
visual attraction of an object, gives an example for each kind, and describes how these
properties are measured. We propose to apply a statistical measure, i.e., a hypothesis
test (see 4.1), to find out whether the values of these properties are significantly dif-
ferent to objects in the local area, e.g., along the same street segment.
Façade Area. The façade area of an object is an important property for determining
its contrast to surrounding objects. People tend to easily notice objects whose façade
areas significantly exceed or fall below the façade areas of surrounding objects. In the
trivial case of a regularly shaped building (of rectangular form) the façade area is
calculated by multiplying its width and height.
Shape. Visual attraction of an object is also determined by its shape. Unorthodox
shapes amidst conventional rectangle-like shapes strike one’s eyes. We formally
specify the shape measure of an object by considering its shape factor and also the
deviation of its shape from that of a rectangle. The shape factor represents the
proportion of height and width. For example, skyscrapers have a high shape factor,
whereas long and low buildings have a low shape factor. The value of deviation is the
difference between the area of the minimum-bounding rectangle of the object’s façade
and its façade area. Notice that the value of deviation is not unique: Although two
objects are of different shape they can have the same value of deviation.
Color. An object can stand out from surrounding objects based on its color. For
example, imagine a red fire department building in the midst of a set of gray
buildings. We appoint a color value to each object by assigning decimal values from
the RGB color chart and then determine whether this color is different from the colors
of surrounding objects.
In principle, color is a property difficult to measure and compare. Perceived light is
a complex function of illumination, reflectance/absorption on surfaces, and receptive
abilities of the visual sense. As Mallot [20] states, it makes no sense to define a metric
in a color space, having no single basic color space. However, in the given context
one can reasonably argue for a white and not too specific illumination (daylight). If
the images are taken in diffuse daylight, color should be roughly comparable. In a
first approach, the color of a building can be measured globally (mean or median), by
a single [R,G,B] triple. A refined approach needs high-level knowledge for image in-
terpretation to select segments of the background color of a building for measurement.
Given a triple for all buildings in the neighborhood, a mean color can be estimated,
and distances (L2 norm) from the mean can be calculated for the hypothesis test.
Visibility. The final property to be measured is the prominence of spatial location.
We propose to measure this property by calculating two-dimensional visibility.
Visibility is considered for the space used in the actual mobility mode—for
pedestrians this is the public street space plus some private areas. It is assumed that
visibility is limited by recognizability, for which reason a pre-defined buffer zone
limits the considered space (and reduces computational complexity). The value of
visibility is then derived from the area of the space covered by the visibility cone of
the front side of a built object (Figure 1), which can then be ordered for the whole set
of objects within a street segment.
Fig. 1. Example for visibility area for ‘Singerstrasse 1’ within buffer zone (100m).
Other Visual Properties. Other properties of an object, such as its texture and
condition, may also influence the contrast to surrounding objects. The reasons for not
including them in our formal model are their subjectivity and lack of formality. The
texture of an object is often hard to identify, both from databases and in the real
world. The condition of an object refers to its age and cleanliness. Age is easy to
determine from a database, but often very hard to guess in the real world. For
example, a building may be very old but due to a recent renovation looks new.
Cleanliness is a subjective measure and therefore hard to specify within formal terms.
Although these properties can be measured objectively according to the specified
rules and criteria, their actual perception by wayfinders is influenced by temporal
constraints. For example, it can be hard to detect the color of a building at night (on
the other hand brightly illuminated buildings are very prominent at night) and the
visibility of a landmark decreases dramatically on a foggy day.
Table 1. Properties for visual attraction and how they are measured.
Properties for visual at-
Example Measurement
Façade area α = 25m * 15m =
= facadexx |
Shape factor
Shape deviation from
β1 = 15m / 25m = 0.6
β2 = 375sqm –
295sqm = 21%
β1 = height / width
β2 = (area of minimum
bounding rectangle – α) /
area of minimum bound-
ing rectangle
Color γ = [255, 0, 0] = red γ = [R, G, B]
Visibility δ = 2400 sqm visibleyx
3.2 Semantic Attraction
Our notion of semantic attraction is similar to that of cognitive attraction [29], which
focuses on the meaning of a feature. Semantic measures for the formal model of
landmark saliency comprise cultural and historical importance of an object, and ex-
plicit marks. The properties for semantic attraction, typical examples, and how these
properties are measured can be seen in Table 2.
Cultural and Historical Importance. Semantic attraction can result from the cultural
and historical importance of an object. For example, the ‘Looshaus’ in the first district
of Vienna is famous for its architectural style (Art Nouveau). This property can be
deduced from a database including cultural and historical objects. For the City of
Vienna such information is available from the so-called ‘Kulturgüterkataster’—a
database of cultural, archeological, and architectural treasures1. This database also
includes pictures of the objects. Here, we assign a Boolean value to each object:
‘True’ if it is of cultural or historical importance and ‘False’ otherwise. One could
refine this assignment by using a classification system similar to the ones of travel
guides for important sites. There, the cultural and historical importance of objects is
often measured on a predefined scale, e.g., from 1 to 5.
Explicit Marks. Marks such as signs on the front of a building explicitly specify its
semantics to the wayfinder. For example, a street sign gives information on what
street the building is located. If a building is marked as ‘coffee house X’ or ‘museum
Y’ then we can also know something about it, which cannot be inferred from its other
visual properties. From the point of view of the provider offering a service with
wayfinding instructions, the commercial semantics of buildings (which is most often
highlighted by explicit marks such as the sign of a supermarket), can be easily
extracted from the yellow pages of the area. An object with an explicit mark is
assigned the Boolean value ‘True’ whereas objects without explicit marks have the
value ‘False.’
Other Semantic Properties. Other measures for semantic attraction, which are not
included in our formal model, are prototypicality and implicit semantics. A
prototypical object is easy to recognize. For example, St. Stephen’s dome in Vienna is
regarded as a prototypical example for a church, whereas a modern church without a
steeple is not considered prototypical. Although categories and prototype effects have
been widely studied among psychologists and linguists ([27], [15]), extensive human
subjects testing concerning the prototypicality of objects along a given route would be
necessary to derive useful results, which could then be implemented in a database.
In the real world, people derive the meaning of an object from either implicit or
explicit semantics. For example, by looking through the windows of a building one
might detect people sitting at tables drinking coffee, talking to other people, or read-
ing newspapers. The conclusion would be that this is a coffee house, although the
building is not explicitly marked as such, therefore implicit semantics. Another exam-
ple is the conclusion that ‘this building must be the museum we are looking for, be-
cause there is a crowd of people lining up in front of the entrance.’ Implicit semantics
is difficult to specify because it is both user- and context dependent, and also tempo-
rally constrained, e.g., opening hours of a coffee house.
Table 2. Properties for semantic attraction and how they are measured.
Properties for semantic
Example Measurement
Cultural and historical
importance ε = T
ε = 1
(building very famous
for its architecture)
FT ,
Scale of importance:
1 (high) – 5 (low)
Explicit mark ζ = T
Sign on front of a
FT ,
3.3 Structural Attraction
A landmark is structurally attractive if it plays a major role or has a prominent loca-
tion in the structure of the spatial environment. Examples are intersections and down-
town plazas [29]. The structure considered here is the travel network of a traveler with
a single means of transport (fixed context). Corresponding to Lynch’s [18] elements
that structure a city, nodes, boundaries (edges), and regions (districts) are structural
elements that are perceivable and might become prominent due to their individual
structuring properties. In this paper we focus on local landmarks for wayfinding,
therefore only nodes and boundaries need to be considered. The individual properties
for structural attraction, examples for these properties, and how they are measured can
be seen in Table 3.
Nodes. Nodes in a travel network are its intersections. For car drivers nodes may be
street intersections, for pedestrians nodes may be places, and for business travelers
nodes may be airports. The central structural characteristic of a node is its grade of
connectivity, or, in terms of graph theory, its degree (Figure 2). The degree may be
additionally weighted by the quality of the incoming and outgoing edges. For
example, the street network hierarchy [30] allows for making a distinction between
two intersecting highways and two intersecting lanes in a street network. Weights
could be defined on a scale from 5 (highways) to 1 (footpaths), with state streets,
overland streets, and town streets in between.
Fig. 2. A T-junction (degree = 3) is quite common in street networks and therefore less remark-
able than a central place (degree = 4).
Boundaries. The perception of the structural properties of boundaries is linked to the
energy that has to be spent to cross them. We hypothesize that a boundary is the more
prominent the larger its resistance is. For example, the ‘Westbahn’ (westbound train
line) in Vienna separates two districts, and can be crossed only over two bridges or
through a tunnel of 2 km length. A similar perceivable boundary is the ‘Donaukanal’
(channel of the Danube) that separates the dense street networks of the first and
second districts by only a few links. Such barriers form significant shapes in city
maps: the travel networks show enclosed cells of large boundary edges with a small
distance between opposite edges, i.e., with a large form factor (Figures 3, 4). A
measure such as the product of cell size and form factor characterizes the structural
landmark saliency of the objects in these cells.
Other Structural Properties. The last of Lynch’s structural elements, i.e., regions,
corresponds to quarters, districts, and other areal subdivisions in the city. Such
elements may be useful landmarks in larger scale applications.
Fig. 3. Both cells have the same size but the right cell has a bigger form factor and creates the
larger barrier.
Fig. 4. The way from Spiegelgasse 9 to Dorotheergasse 7 (Jewish Museum) is long compared
to the distance. The shape of the block of buildings creates a barrier.
Table 3. Properties for structural attraction and how they are measured.
Properties for
structural attraction
Example Measurement
Nodes η = (4*2+4*2)
Second node in Figure 2
for pedestrians; all streets
are town streets (w=2)
η = (i+o)
Weighted incoming (i) and
outgoing (o) edges to and
from a node
Boundaries θ = 2500
Channel dividing a district
θ = cell size * form factor
Form factor: long side /
short side
3.4 Used Data
For the automatic selection of context-dependent landmarks for navigation a number
of data sources need to be available. According to the nature of landmarks, visual data
as well as semantic and structural data are required. Hence, city maps and street
graphs are complemented with images and content databases.
The following data sources are implied:
Digital city maps, such as the multipurpose map of Vienna. City maps provide
boundaries and classifications of the built areas. This information is useful for
model-driven image segmentation and rectification ([11], [25]).
Navigation graphs for the actual means of travel. Navigation graphs are needed for
route selection algorithms as well as for the route-specific classification into possi-
ble and real decision points.
Rectified, geo-referenced images of façades of each single building located at ele-
ments of the navigation graph. TeleInfo2 provides a complete coverage of geo-
referenced images for the street network in Germany and Austria. Figure 5 shows a
(distorted) 360°-view of an intersection demonstrating the richness and complexity
of this type of data.
Fig. 5. A 360°-view of the intersection Stephansplatz / Singerstrasse / Kärntner Strasse / Gra-
ben in Vienna.
Accessible databases such as yellow pages, or databases of cultural heritage, pro-
vide the required semantic content.
4 Assessment of Landmark Saliency
This section introduces the method used to assess the landmark saliency of a feature,
i.e., hypothesis testing. Applying this method to the properties presented in section 3
allows for defining a total measure of landmark saliency for each feature in a dataset.
4.1 Finding Landmarks
As the landmark saliency of a feature is bound to its prominence or distinctiveness, it
is straightforward to evaluate the distinction between the feature attributes and attrib-
utes of other features. A global landmark needs to be distinctive from all other fea-
tures, but our limitation to local landmarks allows a reduction to features that are
nearby. The computationally simplest approach to find the most distinctive feature at
a given location is a maximum (minimum) operator for each attribute, and also for the
total value of attraction. This procedure guarantees finding a local landmark in any
case, even if the difference from a local mean is small. However, the result cannot be
assessed in terms of significance.
For those measures that are continuous and a normal distribution can be assumed,
the assessment can be reached by hypothesis testing of the significance of deviations
from local mean characteristics [13]. Assuming a typical local appearance of objects
we may suppose a normal distribution for some of the characteristics. Further assum-
ing that there are outliers (namely the landmarks) the estimated mean of the distribu-
tion shall be determined by the median of all local observations. Also, the standard
deviation can be calculated. Both parameters—mean and standard deviation—depend
on the definition of a local neighborhood. This definition should be linked to the per-
ceptual capabilities of the human users in a specific mode of traveling, e.g., for pedes-
trians the neighborhood should be chosen smaller than for car drivers. The parameters
of a distribution are calculated once and then updated only when changes in the local
environment occur, which happens rarely. This means that the local neighborhood can
be bound to each feature, similar to a local filter operation. One could choose a rec-
tangle of a specific side length depending on the mode of travel.
Given the local parameters of the distribution for each characteristic, the feature at-
tributes can be tested for their difference from the mean. The hypothesis is that the
feature attributes deviate significantly from the local mean. If the hypothesis is re-
jected, the feature attribute is not significantly different from its surroundings. If the
hypothesis is accepted, the feature has some kind of landmark saliency, related to the
tested attribute. Type I errors (rejecting a correct hypothesis) lead to distinctive fea-
tures that are not detected. Type II errors (accepting a wrong hypothesis) lead to the
use of features as landmarks that are not distinctive. Type II errors are more expensive
because they lead to instructions, which are not useful. This means the power of the
test—the probability β of avoiding a Type II error—will be set high in the test proce-
4.2 Combination of Property Values for Measuring Landmark Saliency
The individual measures of properties shall now be combined to a global measure of
landmark saliency for each feature in a dataset. In a first computational step the vector
of property values is determined for each feature (Table 4). Then, for each feature and
each property the local mean and standard deviation are determined (see 4.1). Each
triple of value, local mean, and standard deviation, is subject to a hypothesis test that
determines whether a property value is significant (s=1) or not (s=0). The vector of
significance values can be grouped for visual, semantic, and structural significance.
With predefined weights for each group a total measure for the landmark saliency of a
feature can be calculated. The predefined weights allow for an adaptation to the con-
text (mode of travel) or individual user preferences.
Table 4. Deriving the total value of landmark saliency for a feature.
α sα
β1 sβ1
β2 sβ2
γ sγ
δ sδ
svis = (sα+
sγ+sδ) / 5
ε sε
attraction ζ sζ
ssem =
(sε+sζ) / 2 wsem s
η sη
attraction θ sθ
sstr =
(sη+sθ) / 2 wstr s
5 Wayfinding Instructions with Local Landmarks – An Example
This section demonstrates the applicability and usefulness of the presented approach
by showing an example from the introductory case study (section 1).
Fig. 6. The instruction at the decision point shall use the most salient feature at the decision
5.1 Description of the Situation
When taking the shortest path from the Café Diglas to the restaurant Novelli—both
located in Vienna’s first district—the wayfinder reaches at some point the intersection
of ‘Graben’, ‘Kärntner Strasse’, and ‘Stephansplatz’ (Figure 6 gives a panoramic
view). This decision point is used to demonstrate the selection of a landmark based on
the method developed in the paper. The instruction at the decision point has to direct
one to turn right, and the available data is evaluated for the automatic selection of a
local landmark.
5.2 Measures and Weights for the Extraction of a Local Landmark
The measures for the attractiveness are calculated for all features at the decision point.
Table 5 shows the individual property values and the total value of landmark saliency
for the ‘Haas’ building. The total value of landmark saliency is 1.8, which is the
maximum value for all features. The next total value is 1.2 for the ‘Bank Austria’
building. It is therefore recommended to use the ‘Haas’ building as a local landmark
in an instruction at this decision point.
Table 5. Deriving the total value of landmark saliency for the ‘Haas’ building.
α 17400 1
β1 0.62 1
β2 0 0
γ 21
δ 10600 1
svis = 0.8
wvis = 1
ε T 1
attraction ζ T 1
ssem = 2 / 2 wsem = 1 1
η - 0
attraction θ - 0
sstr =
0 / 2 wstr = 1 0
In the given example, the weights are set to wvis = 1, wsem = 1, and wstr = 1. Differ-
ent sets of weights could be selected for different user groups. For example, the set of
weights wvis = 3, wsem = 1, and wstr = 1 would reflect the visual capabilities of certain
users and might lead to a different resulting landmark (it would nevertheless not
change the outcome in this case).
5.3 The Wayfinding Directions
Having identified a local landmark at the decision point, an instruction can be created
following any grammar, for instance the grammar defined in section 2.3. The way a
landmark is communicated is not of central concern in this paper; however, at the
time of creating the instruction all significant properties of the landmark are known
and can be used.
In our case, the identified landmark is the ‘Haas’ building, which is known for its
controversial architecture (Figure 7). The instruction using the feature ‘Haas’ building
as a destination landmark is then:
AT previous landmark
TURN LEFT ONTO “Stephansplatz”
UNTIL “Haas building, a dark building of architectural
significance containing a (signed) Zara shop at the
Fig. 7. The most salient feature at the considered decision point: the ‘Haas’ building by archi-
tect Hans Hollein.
The example can be extended by showing the (optional) use of a route mark in the
instruction. With the landmark ‘St. Stephen’s cathedral’ along the considered route
segment, the instruction has the following form:
AT previous landmark
TURN LEFT ONTO “Stephansplatz”
PASSING “Stephansdom, a visually salient world cultural
heritage building”
UNTIL “Haas building, a dark building of architectural
significance containing a (signed) Zara shop at the
Note the chosen freedom to generalize different visually significant properties (façade
area, shape factor) into ‘visually salient.’
6 Conclusions and Future Work
In this paper we presented a method to automatically extract local landmarks from
datasets to be integrated in wayfinding instructions. Different individual properties for
the attractiveness of a landmark were first defined and then put together to form a
global measure of landmark saliency for each feature in a dataset. Hypothesis testing
was used to select the most significant landmark at each decision point for inclusion
in the wayfinding instruction. We applied the formal framework to an actual wayfind-
ing scenario to show the applicability and usefulness of our approach.
The work leads to many different directions for future research:
1. Activity- and profile-based selection of landmarks: Human subjects testing will
show how the weights of the attraction measures have to be adapted for different
modes of travel (e.g., pedestrian, bicycle, car) and user groups (e.g., tourist, busi-
ness traveler, handicapped).
2. One needs to find out how accurate the data have to be to get useful results.
3. The method needs to be implemented and applied to larger datasets in order to ana-
lyze performance and computational cost.
4. In the case study we have calculated measures of landmark saliency for individual
features only. How can aggregate landmarks (formed by connected objects such as
a block of buildings) be extracted automatically?
5. Different times (e.g., day- or nighttime) may require different landmarks therefore
the integration of temporal constraints into route instructions is necessary.
The major part of this work was done while the first author was at the Institute for
Geoinformation, Vienna University of Technology.
1. Allen, G., 1999: Spatial Abilities, Cognitive Maps, and Wayfinding - Bases for Individual
Differences in Spatial Cognition and Behavior. In: Golledge, R. (Ed.), Wayfinding Behavior
- Cognitive Mapping and Other Spatial Processes, Johns Hopkins University Press, Balti-
more, pp. 46-80.
2. Burnett, G., 1998: Turn Right at the King's Head. Drivers' Requirements for Route Guidance
Information. Ph.D., Loughborough University.
3. Burnett, G.; Smith, D.; May, A., 2001: Supporting the navigation task: Characteristics of
'good' landmarks. In: Hanson, M.A. (Ed.), Contemporary Ergonomics 2001, Taylor & Fran-
cis, London, pp. 441-446.
4. Deakin, A., 1996: Landmarks as Navigational Aids on Street Maps. Cartography and Geo-
graphic Information Systems, 23 (1): 21-36.
5. Denis, M.; Pazzaglia, F.; Cornoldi, C.; Bertolo, L., 1999: Spatial Discourse and Navigation:
An Analysis of Route Directions in the City of Venice. Applied Cognitive Psychology, 13:
6. Downs, R.; Stea, D., 1977: Maps in Minds: Reflections on Cognitive Mapping. Harper and
Row, New York.
7. Fellbaum, C. (Ed.), 1998: WordNet: An Electronic Lexical Database. The MIT Press, Cam-
bridge, Massachusetts, 423 pp.
8. Fontaine, S.; Denis, M., 1999: The production of route instructions in underground and urban
environments. In: Freksa, C.; Mark, D. (Eds.), Spatial Information Theory. Lecture Notes in
Computer Science, 1661. Springer, Berlin, pp. 83-94.
9. Golledge, R., 1999: Human Wayfinding and Cognitive Maps. In: Golledge, R. (Ed.), Way-
finding Behavior - Cognitive Mapping and Other Spatial Processes, Johns Hopkins Univer-
sity Press, Baltimore, pp. 5-45.
10. Golledge, R., 1993: Geographical perspectives on spatial cognition. In: Gärling, T.;
Golledge, R. (Eds.), Behaviour and environment: Psychological and geographical ap-
proaches, Elsevier, Amsterdam, pp. 16-46.
11. Haala, N.; Anders, K.-H., 1996: Fusion of 2D-GIS and Image Data for 3D Building Recon-
struction. International Archives of Photogrammetry and Remote Sensing, XXXI (B3): 285-
12. Habel, C., 1988: Prozedurale Aspekte der Wegplanung und Wegbeschreibung. In: Schnelle,
H.; Rickheit, G. (Eds.), Sprache in Mensch und Computer, Westdeutscher Verlag, Opladen,
pp. 107-133.
13. Koch, K., 1999: Parameter Estimation and Hypothesis Testing in Linear Models. Springer,
Berlin, 333 pp.
14. Kuipers, B., 1978: Modeling Spatial Knowledge. Cognitive Science (2): 129-153.
15. Lakoff, G., 1987: Women, Fire, and Dangerous Things: What Categories Reveal About the
Mind. The University of Chicago Press, Chicago, 614 pp.
16. Livatino, S.; Madsen, C., 2000: Acquisition and recognition of visual landmarks for
autonomous robot navigation, International Symposium on Intelligent Robotic Systems,
Reading, UK, pp. 269-279.
17. Lovelace, K.; Hegarty, M.; Montello, D., 1999: Elements of Good Route Directions in Fa-
miliar and Unfamiliar Environments. In: Freksa, C.; Mark, D. (Eds.), Spatial Information
Theory. Lecture Notes in Computer Science, 1661. Springer, Berlin, pp. 65-82.
18. Lynch, K., 1960: The Image of the City. MIT Press, Cambridge, 194 pp.
19. Maaß, W.; Schmauks, D., 1998: MOSES: Ein Beispiel für die Modellierung räumlicher
Leistungen durch ein Wegebeschreibungssystem. Zeitschrift für Semiotik, 20 (1-2): 105-
20. Mallot, H.; Allen, J., 2000: Computational Vision. MIT Press, Cambridge.
21. Merriam-Webster, 2001: Merriam-Webster's Collegiate Dictionary. Merriam-Webster, Inc.
22. Metzger, W., 1936: Gesetze des Sehens. Senckenberg-Buch, VI. W. Kramer & Co.,
Frankfurt am Main, 172 pp.
23. Michon, P.-E.; Denis, M., 2001: When and Why are Visual Landmarks Used in Giving Di-
rections? In: Montello, D. (Ed.), Spatial Information Theory. Lecture Notes in Computer
Science, 2205. Springer, Berlin, pp. 292-305.
24. Nicklas, D.; Großmann, M.; Schwarz, T.; Volz, S.; Mitschang, B., 2001: A Model-Based
Open Architecture for Mobile, Spatially-Aware Applications. In: Jensen, C. et al. (Eds.),
Advances in Spatial and Temporal Databases. Lecture Notes in Computer Science, 2121.
Springer, Berlin, pp. 117-135.
25. Pasko, M.; Gruber, M., 1996: Fusion of 2D GIS Data and Aerial Images for 3D Building
Reconstruction. International Archives of Photogrammetry and Remote Sensing, XXXI
(B3): 257-260.
26. Pradhan, S.; Brignone, C.; Cui, J.-H.; McReynolds, H.; Smith, M., 2001: Websigns: Hyper-
linking Physical Locations to the Web. IEEE Computer Journal, 34 (8): 42-48.
27. Rosch, E., 1978: Principles of Categorization. In: Rosch, E.; Lloyd, B. (Eds.), Cognition
and Categorization, Lawrence Erlbaum Associates, Hillsdale, New Jersey, pp. 27-48.
28. Siegel, A.; White, S., 1975: The development of spatial representations of large-scale envi-
ronments. In: Reese, H. (Ed.), Advances in child development and behavior, 10. Academic
Press, New York, pp. 9-55.
29. Sorrows, M.; Hirtle, S., 1999: The Nature of Landmarks for Real and Electronic Spaces. In:
Freksa, C.; Mark, D. (Eds.), Spatial Information Theory. Lecture Notes in Computer Sci-
ence, 1661. Springer, Berlin, pp. 37-50.
30. Timpf, S.; Frank, A., 1997: Using Hierarchical Spatial Data Structures for Hierarchical Spa-
tial Reasoning. In: Hirtle, S.; Frank, A. (Eds.), Spatial Information Theory. Lecture Notes in
Computer Science. Springer, Berlin, pp. 69-83.
31. Werner, S.; Krieg-Brückner, B.; Mallot, H.; Schweizer, K.; Freksa, C., 1997: Spatial Cogni-
tion: The Role of Landmark, Route, and Survey Knowledge in Human and Robot Naviga-
tion. In: Jarke, M.; Pasedach, K.; Pohl, K. (Eds.), Informatik '97. Springer, pp. 41-50.
32. Wertheimer, M., 1925: Über Gestalttheorie. Philosophische Zeitschrift für Forschung und
Aussprache, 1: 39-60.
33. Yeh, E., 1995: Toward Selecting and Recognizing Natural Landmarks. Yale Technical Re-
port 9503, Center for Systems Science, Yale University.
... In order to facilitate the challenging navigation process, wayfinders use landmarks as visual cues to organize their spatial knowledge and orient themselves in the environment (Couclelis et al., 1987;Montello, 2005). Landmarks are objects in the environment that are easily identifiable, have high contrast with their surroundings, have prominent characteristics, and can be used as anchor and reference points for orientation, wayfinding, and communication (Lynch, 1960;Raubal & Winter, 2002;Richter & Winter, 2014;Sorrows & Hirtle, 1999). For instance, in the above example, a landmark could be a unique building with a characteristic facade. ...
... For instance, in the above example, a landmark could be a unique building with a characteristic facade. Given their role as visual cues, landmarks hold high practical importance for human wayfinding and spatial knowledge acquisition (Couclelis et al., 1987;Raubal & Winter, 2002;Richter & Winter, 2014;Siegel & White, 1975;Sorrows & Hirtle, 1999). ...
... Current mobile navigation aids do not provide what is needed to acquire spatial knowledge during navigation (Ishikawa, 2018); they seem to shift users' attention away from the environment and task-relevant landmarks Gardony et al., 2015), thereby compromising wayfinders' ability to incorporate landmarks into spatial learning (Dahmani & Bohbot, 2020;Ishikawa, 2018;Siegel & White, 1975). As a result, there have been many design recommendations for integrating landmarks into future mobile maps for effective communication to pedestrian navigators (Raubal & Winter, 2002;Richter & Winter, 2014;Thrash et al., 2019;Yesiltepe et al., 2021). However, how to graphically visualize landmarks on mobile maps for effective communication to pedestrian navigators remains an open question (Richter & Winter, 2014). ...
Full-text available
Even though they are day-to-day activities, humans find navigation and wayfinding to be cognitively challenging. To facilitate their everyday mobility, humans increasingly rely on ubiquitous mobile maps as navigation aids. However, the over-reliance on and habitual use of omnipresent navigation aids deteriorate humans' short-term ability to learn new information about their surroundings and induces a long-term decline in spatial skills. This deterioration in spatial learning is attributed to the fact that these aids capture users' attention and cause them to enter a passive navigation mode. Another factor that limits spatial learning during map-aided navigation is the lack of salient landmark information on mobile maps. Prior research has already demonstrated that wayfinders rely on landmarks—geographic features that stand out from their surroundings—to facilitate navigation and build a spatial representation of the environments they traverse. Landmarks serve as anchor points and help wayfinders to visually match the spatial information depicted on the mobile map with the information collected during the active exploration of the environment. Considering the acknowledged significance of landmarks for human wayfinding due to their visibility and saliency, this thesis investigates an open research question: how to graphically communicate landmarks on mobile map aids to cue wayfinders' allocation of attentional resources to these task-relevant environmental features. From a cartographic design perspective, landmarks can be depicted on mobile map aids on a graphical continuum ranging from abstract 2D text labels to realistic 3D buildings with high visual fidelity. Based on the importance of landmarks for human wayfinding and the rich cartographic body of research concerning their depiction on mobile maps, this thesis investigated how various landmark visualization styles affect the navigation process of two user groups (expert and general wayfinders) in different navigation use contexts (emergency and general navigation tasks). Specifically, I conducted two real-world map-aided navigation studies to assess the influence of various landmark visualization styles on wayfinders' navigation performance, spatial learning, allocation of visual attention, and cognitive load. In Study I, I investigated how depicting landmarks as abstract 2D building footprints or realistic 3D buildings on the mobile map affected expert wayfinders' navigation performance, visual attention, spatial learning, and cognitive load during an emergency navigation task. I asked expert navigators recruited from the Swiss Armed Forces to follow a predefined route using a mobile map depicting landmarks as either abstract 2D building footprints or realistic 3D buildings and to identify the depicted task-relevant landmarks in the environment. I recorded the experts' gaze behavior with a mobile eye-tracer and their cognitive load with EEG during the navigation task, and I captured their incidental spatial learning at the end of the task. The wayfinding experts' exhibited high navigation performance and low cognitive load during the map-aided navigation task regardless of the landmark visualization style. Their gaze behavior revealed that wayfinding experts navigating with realistic 3D landmarks focused more on the visualizations of landmarks on the mobile map than those who navigated with abstract 2D landmarks, while the latter focused more on the depicted route. Furthermore, when the experts focused for longer on the environment and the landmarks, their spatial learning improved regardless of the landmark visualization style. I also found that the spatial learning of experts with self-reported low spatial abilities improved when they navigated with landmarks depicted as realistic 3D buildings. In Study II, I investigated the influence of abstract and realistic 3D landmark visualization styles on wayfinders sampled from the general population. As in Study I, I investigated wayfinders' navigation performance, visual attention, spatial learning, and cognitive load. In contrast to Study I, the participants in Study II were exposed to both landmark visualization styles in a navigation context that mimics everyday navigation. Furthermore, the participants were informed that their spatial knowledge of the environment would be tested after navigation. As in Study I, the wayfinders in Study II exhibited high navigation performance and low cognitive load regardless of the landmark visualization style. Their visual attention revealed that wayfinders with low spatial abilities and wayfinders familiar with the study area fixated on the environment longer when they navigated with realistic 3D landmarks on the mobile map. Spatial learning improved when wayfinders with low spatial abilities were assisted by realistic 3D landmarks. Also, when wayfinders were assisted by realistic 3D landmarks and paid less attention to the map aid, their spatial learning improved. Taken together, the present real-world navigation studies provide ecologically valid results on the influence of various landmark visualization styles on wayfinders. In particular, the studies demonstrate how visualization style modulates wayfinders' visual attention and facilitates spatial learning across various user groups and navigation use contexts. Furthermore, the results of both studies highlight the importance of individual differences in spatial abilities as predictors of spatial learning during map-assisted navigation. Based on these findings, the present work provides design recommendations for future mobile maps that go beyond the traditional concept of "one fits all." Indeed, the studies support the cause for landmark depiction that directs individual wayfinders' visual attention to task-relevant landmarks to further enhance spatial learning. This would be especially helpful for users with low spatial skills. In doing so, future mobile maps could dynamically adapt the visualization style of landmarks according to wayfinders' spatial abilities for cued visual attention, thus meeting individuals' spatial learning needs.
... In the literature, many empirical studies suggest that the following factors affect wayfinder's perceived complexity for the wayfinding task, including personal experience and expertise, physical environments' characteristics, and route features (Fang, Li, & Shaw, 2015;Giannopoulos, Kiefer, Raubal, Richter, & Thrash, 2014;Schmidt, Beigl, & Gellersen, 1999). Indeed, the human perceived complexity of road intersections for making decisions depends not only on the structural configuration of road networks but also on the visual sensing of the real scenes, as well as the availability of semantic features (e.g., well-known points of interest (POIs)) in the environment (Guan et al., 2022;Raubal & Winter, 2002;Sorrows & Hirtle, 1999). Additionally, different passing directions at a road intersection provide pedestrians with different perceptual information ('environmental cues'). ...
... Inspired by Raubal and Winter (2002), a well-cited study on combining visual, structural, and semantic features, and many other related studies (e.g., Nuhn (2020)), this study employs a weighted sum model that assigns different weights to the three dimensions. Each feature is normalized to the interval [0,1] using Min-Max feature scaling. ...
The complexity of urban physical environments at road intersections is a primary factor characterizing the difficulty of wayfinding, which is a fundamental spatial activity of human beings in cities. A complex intersection may increase the difficulty of understanding the environment, which may result in incorrect turning decisions and even bring road safety issues. Existing methods measure the complexity of road intersections by solely considering their visual or structural features. More importantly, they only output a single complexity value for each intersection, failing to differentiate the decision-making complexity based on the specific entry/exit branches of a passing branches. This study proposes a computational model to quantify the fine-grained decision-making complexity of road intersections for the navigation data models and navigation systems based on specific passing branches, using the visual, structural, and semantic features from human perspectives. For each pair of two branches (i.e., one entry and one exit) passing through the road intersection, the model will output a specific decision-making complexity score. Furthermore, this study develops a route planning algorithm for generating the minimum complexity route to serve relevant navigation applications. This study contributes to human-centered route planning and communication, as well as enabling potential innovative applications in traffic safety studies and sustainable urban and environmental development.
... According to the literature, there are two different landmarks based on their visibility. A local landmark where people experience landmarks visible only from a limited area and only from speci c directions (Raubal & Winter, 2002;Schwering et al., 2013). The second type is a global landmark, visible from any/every location in a setting (Lynch, 1960). ...
Full-text available
Researchers believe that wayfinding and landmark identification can be enhanced using route instruction and a bird's eye view. It remains an open question whether a bird's eye view or a route instruction would reduce cognitive load in spatial landmark identification. In addition, the effect of environmental colour on human landmark identification during navigation is unclear. The study was conducted with a Virtual environmental (VE) paradigm, and Sixty-six college students (46 males and 20 females) between the ages of 18–35 years volunteered as participants. Participants were randomly assigned to four groups (Instruction- Bird's eye, Instruction- No Bird's eye, No Instruction- Bird's eye, and No Instruction - No Bird's eye). The results of the independent between-group ANOVA yielded a statistically significant effect, F (3, 56) = 3.75, p = 0.01, η2 = 0.16 on coloured environmental conditions. Compared to the B/W condition, coloured environments support landmark identification only in the initial stages of wayfinding. Moreover, the visual trajectory analysis indicates that the number of deviations in the shortest route is less in B/W conditions than in coloured conditions. The study results demonstrated the importance of route instruction on landmark identification under coloured and B/W environments. The results also indicate that the wayfinding time can be reduced by providing clear route instructions in a declarative format.
... Landmarks are geographic features that structure human mental representation of space by acting as anchors or reference points for orientation, wayfinding, and route communication (Richter & Winter, 2014). Therefore, landmarks hold high practical importance for human wayfinding (Raubal & Winter, 2002). Despite the evidence that landmark visualization can mediate spatial learning from navigation aids in the general population (see reviews by Richter & Winter, 2014;Yesiltepe et al., 2021), how to effectively visualize landmarks on mobile map displays is still an open research question (Richter & Winter, 2014). ...
Full-text available
Humans increasingly rely on GPS-enabled mobile maps to navigate novel environments. However, this reliance can negatively affect spatial learning, which can be detrimental even for expert navigators such as search and rescue personnel. Landmark visualization has been shown to improve spatial learning in general populations by facilitating object identification between the map and the environment. How landmark visualization supports expert users’ spatial learning during map-assisted navigation is still an open research question. We thus conducted a real-world study with wayfinding experts in an unknown residential neighborhood. We aimed to assess how two different landmark visualization styles (abstract 2D vs. realistic 3D buildings) would affect experts’ spatial learning in a map-assisted navigation task during an emergency scenario. Using a between-subjects design, we asked Swiss military personnel to follow a given route using a mobile map, and to identify five task-relevant landmarks along the route. We recorded experts’ gaze behavior while navigating and examined their spatial learning after the navigation task. We found that experts’ spatial learning improved when they focused their visual attention on the environment, but the direction of attention between the map and the environment was not affected by the landmark visualization style. Further, there was no difference in spatial learning between the 2D and 3D groups. Contrary to previous research with general populations, this study suggests that the landmark visualization style does not enhance expert navigators’ navigation or spatial learning abilities, thus highlighting the need for population-specific mobile map design solutions.
... More recently, attention has also been drawn to reduce the adverse effects of GPS-enabled navigation aids on spatial learning, e.g., by geographic information scientists (GIScientists) (Wenig et al., 2017), cognitive scientists (Ruginski et al., 2019), and map user interface (UI/UX) designers (Ricker and Roth, 2018;Thrash et al., 2019;Li, 2020;Fabrikant, 2022). Among the ideas proposed, the appropriate inclusion and display of landmarks on GPS-enabled mobile maps has gained particular traction among cartographers and navigation researchers in GIScience (Raubal and Winter, 2002;Duckham et al., 2010;Credé et al., 2020;Keil et al., 2020;Liu et al., 2022). ...
Full-text available
The continuous assessment of pedestrians’ cognitive load during a naturalistic mobile map-assisted navigation task is challenging because of limited experimental control over stimulus presentation, human-map-interactions, and other participant responses. To overcome this challenge, the present study takes advantage of navigators’ spontaneous eye blinks during navigation to serve as event markers in continuously recorded electroencephalography (EEG) data to assess cognitive load in a mobile map-assisted navigation task. We examined if and how displaying different numbers of landmarks (3 vs. 5 vs. 7) on mobile maps along a given route would influence navigators’ cognitive load during navigation in virtual urban environments. Cognitive load was assessed by the peak amplitudes of the blink-related fronto-central N2 and parieto-occipital P3. Our results show increased parieto-occipital P3 amplitude indicating higher cognitive load in the 7-landmark condition, compared to showing 3 or 5 landmarks. Our prior research already demonstrated that participants acquire more spatial knowledge in the 5- and 7-landmark conditions compared to the 3-landmark condition. Together with the current study, we find that showing 5 landmarks, compared to 3 or 7 landmarks, improved spatial learning without overtaxing cognitive load during navigation in different urban environments. Our findings also indicate a possible cognitive load spillover effect during map-assisted wayfinding whereby cognitive load during map viewing might have affected cognitive load during goal-directed locomotion in the environment or vice versa. Our research demonstrates that users’ cognitive load and spatial learning should be considered together when designing the display of future navigation aids and that navigators’ eye blinks can serve as useful event makers to parse continuous human brain dynamics reflecting cognitive load in naturalistic settings.
This study is aimed to report the perspective of people with V.I about their use of spatial representation(cognitive maps) for determining the support level and applicability of cognitive maps while traveling on real routes. This research utilized the methodology of qualitative research in the form of semi structured interviews in order to obtain insights of persons having visual impairment about their navigation in different environments. The interviews were conducted with 20 persons having visual impairment working in different fields of life. The data analysis was done through thematic analysis approach. The participants were intended to answer the questions about the representation and understanding of space, the support level of these representations while traveling, and the role of other senses for successful navigation. The results showed that persons having visual impairment do have spatial representation and the creation of spatial representations does not require visual experience. The study suggested the strong need to assess the way in which persons having visual impairments develop cognitive maps of their environment because it is extremely significant both theoretically and practically and throws light on the role of sensory modalities in the development of mental mapping skills which can in turn recommend how spatial developments might be nourished.
Landmark buildings are salient features for spatial cognition on maps. Distinctive outlines are the major visual characteristics that separate landmark buildings from their surrounding environments. The automatic symbolization of landmark outlines facilitates recognition and map production. As users often recognize landmarks by the outlines of their façades from a street view, this study proposes an automatic method for automatically generating representations of the outlines of landmark buildings in four steps: (1) extract outlines from street-view photographs using GrabCut method, (2) vectorize the extracted building outlines, (3) simplify outline shapes, and (4) symbolize the simplified building outlines in three dimensions (3D). We used the proposed method to generate test data with symbolized outlines for eight buildings in a real-world environment for a wayfinding experiment in which the subjects used the building representations to identify landmark buildings and evaluated their perception of the generated maps. The subjects successfully recognized these buildings based on the symbolized outlines on a map, expressed satisfaction with the manually generated 3D symbols, and reported the same or similar ease of building recognition using 2D or 3D symbolized outlines.
Full-text available
In this report, we present the project URWalking conducted at the University of Regensburg. We describe its major outcomes: Firstly, an indoor navigation system for pedestrians as a web application and as an Android app with position tracking of users in indoor and outdoor environments. Our implementation showcases that a variant of the $$A^*$$ A ∗ -algorithm by Ullmann (tengetriebene optimierung präferenzadaptiver fußwegrouten durch gebäudekomplexe , 2020) can handle the routing problem in large, levelled indoor environments efficiently. Secondly, the apps have been used in several studies for a deeper understanding of human wayfinding. We collected eye tracking and synchronized video data, think aloud protocols, and log data of users interacting with the apps. We applied state-of-the-art deep learning models for gaze tracking and automatic classification of landmarks. Our results indicate that even the most recent version of the YOLO image classifier by Redmon and Farhadi (olov3: An incremental improvement. arXiv, 2018) needs finetuning to recognize everyday objects in indoor environments. Furthermore, we provide empirical evidence that appropriate machine learning models are helpful to bridge behavioural data from users during wayfinding and conceptual models for the salience of objects and landmarks. However, simplistic models are insufficient to reasonably explain wayfinding behaviour in real time—an open issue in GeoAI. We conclude that the GeoAI community should collect more naturalistic log data of wayfinding activities in order to build efficient machine learning models capable of predicting user reactions to routing instructions and of explaining how humans integrate stimuli from the environment as essential information into routing instructions while solving wayfinding tasks. Such models form the basis for real-time wayfinding assistance.
This chapter describes the geographical perspectives on spatial cognition. Geography has, as one of its major emphases, the discovery and explanation of spatial patterns of specific features, functions, phenomena and interactions in environments of many scales. These patterns are generally identified at scales well beyond the perceptual domain. The focus of the chapter is to articulate some of the fundamental or primitive elements that are embedded in the physical and built environment that should have counterparts in cognized space. Such an emphasis helps explain the geographer's perspective in environmental cognition generally and in spatial cognition in particular. The chapter presents some of the fundamental or primitive elements of physical reality that have been identified by geographic research, and to suggest how they should evidence themselves in the cognitive domain. It also discusses some specific properties of spatial knowledge from a geographical perspective, to advance hypotheses about the way these properties manifest themselves in the content of cognized space, and to provide evidence wherever possible of the explicit and implicit involvement of geographers in identifying the nature and content of spatial cognition. Finally, the chapter examine show geographers have used spatial cognition in dealing with real world problems.
The abstract for this document is available on CSA Illumina.To view the Abstract, click the Abstract button above the document title.