UTILIZATION OF SCIENTIFIC METHODS TO IMPROVE URBAN SAFETY

           Vojtěch Stehel ^1, Radimír Novotný ^1, Aleš Kaňkovský^1, Veronika Sedláková^1

^1 The Institute of Technology and Businesses in Ceske Budejovice, Czech Republic


stehel@mail.vstecb.cz

Abstract. There are more than 6,000 municipalities in the Czech Republic, where concept of the
municipality is understood as a territorial self-government unit.

Thanks to operation possibilities of the local government, it is possible to solve the local
problems in these municipalities, including example: construction of various public spaces,
transportation, security, etc. This solution is often better than would be in the case with the
central management. However, this management has its limits associated with the budget parameters
of the given municipalities and it is very closely related to the political struggle in the given
locality. It is very common for the people who have no experience with the management of public
spaces and finances to be elected to the new positions of representatives and mayors. This state of
affairs may also lead to erroneous decisions that may have a long-term negative impact on the
citizens of the given municipalities, in view of their efforts to identify against previous
political competition.

The article makes extensive research into the application of scientific methods, approaches,
experiments and their validation that can be used in the context of urban safety. The focus of
these methods will be concentrated mainly on the municipalities (around 10,000 inhabitants).
Applied methods will also be discussed with view of limited resources (budget, personnel,
territorial, etc.). Part of the research article will be examples of good practice that have been
used primarily in foreign countries. There will also be a discussion on the implementation and
impact of these examples of good practice in cities in the Czech Republic. Implementation will be
discussed in two examples of specific cities (Milevsko, Chotěboř) with regard to their budget
constraints, state of roads and sidewalks and many other parameters. The aim of the article is to
get a basic overview of possible ways of solving local problems with a view to maximizing the
benefit for the citizen. The study was developed as part of the research project TL02000559 and
will be followed by further research with application potential.

1.  Introduction

Security can be understood from many points of view. The most basic view is the safety associated
with human survival, or with the ensuring of basic living needs. Another point of view can be
devoted to secure a property, or even in some sense we can talk about security in connection with
investments. The investments can be safe (almost certain) or, conversely, high-risk. In all
security categories, there are methods and systems that can increase the security or, at least, try
to quantify it in some way. If we look at the safety with the point of view of self-government
city, We will probably be concerned with the safety of human lives and the security associated with
protecting property.

The ensuring of basic living needs is realized by infrastructures, which are gradually expanding
and becoming more important with the development of humanity [1]. At the same time, these
infrastructures are becoming progressively more interconnected, thanks to the information
infrastructure [2]. Important infrastructures that ensure the basic living needs are called
critical infrastructures [3]. These infrastructures must face a number of threats (like congestion,
the environment, terrorism etc.) and therefore they are increasingly given attention. This leads to
analyses of the risks associated with the downtime of these infrastructures [4].

The infrastructures can be physically, cybernetically, geographically and logically dependent on
each other [5]. A number of publications are devoted to analyzing the risks associated with
interdependence [6, 7]. The infrastructures must also evolve in today's world so they are not a
static system. Yusta et al. [8] takes into account the dynamics of the development of this system
in risk analysis. Murray a Grubesic [9] deal with the protection based on the worst possible
scenario. This way may not be optimal, with regard to the operation of the entire infrastructure.

The safety of the inhabitant is influenced not only by the key infrastructure, but also by its
interaction with the environment. In a number of interactions, the individual does not get into
situations that could endanger his life. But in many cases, it is the other way around. First of
all it is about transportation, which is used by most of the population every day. In this case, it
can occur relatively easily to situations threatening human life. Other interactions are
situations, where the common inhabitant meets a criminal offense.

2.  Transport and safety

This issue is addressed in the study by the authors Pucher and Dijkstra [10], who examine how to
improve the public health and safety of pedestrians and cyclists in American cities, according to
successful policies in Germany and the Netherlands. Very important for cities is to improve
conditions and reduce injuries or death of pedestrians and cyclists, thereby strengthening the
fight against dangerous overweight of the citizens and achieving movement and overall healthier
style in the everyday life of a person. In Germany and the Netherlands, the infrastructure used by
pedestrians and cyclists has been strengthened. This includes wide and well-lit sidewalk on both
road sides, clearly marked and raised pedestrian crossings with special lighting for good
visibility. Cycling Transport has a real coordinated network that covers urban and rural areas. In
addition, special bicycle lanes are built for bicycles leading directly to intersections that have
bicycle traffic signals and separate driving signs.

The Most important impact was the reduction in the speed of motor vehicles in residential areas,
which resulted in a reduction of the collision between the pedestrian or the cyclist with a
vehicle. That ensured the greater likelihood of survival. City structures are also oriented so that
the cyclists and pedestrians can easily and without problems get where they need to. Residential
and commercial buildings have sidewalks and cycle paths, it also includes safe crossing over the
transport ways like railway crossing or in case of highways a bridges only for pedestrian and
cyclists. The next point of the study is the transport education, which must be passed not only by
vehicle drivers, but it already started with children from 10 years of age.  They learn traffic
regulations, the right behavior for safe walking, cycling and anticipating dangerous situations,
and how to react properly on it. Last but not least, it is about traffic regulations and penalties
that benefit the cyclists and pedestrians, mainly in the Netherlands and Germany. In a traffic
accident that is caused by the cyclist or pedestrian motorists are always partly guilty. In case of
the accidents involving children or the elderly are totally guilty of the motorist who has to
predict the dangerous behavior of the pedestrians or cyclists. The conclusion is that American
cities should be more focused on strengthening the safety conditions for pedestrians and cycling.
Further build an attractive non-motorised transport network and thus more entice and attract its
citizens to greater movement and healthier lifestyles

The authors Vasilev, Pritchard and Jonsson [11] conducted research to test the lane for cyclists
and changes in user behavior. The Project was implemented in Trondheim, Norway, and was aimed at
encouraging the population to use publicly available and more mobility-active modes of transport.
On the basis of the research, the number of lanes in this area has been reduced from four to two a
two-way cycle path has been introduced. Research has shown that user behavior has changed
significantly and has contributed to a reduction in motorized traffic to improve cycling
conditions, which has led to greater interest in pedestrian and cycling.

Bicycle transport is used by more and more children and adults, and so the authors Zeuwts et al.
[12] focused on the perception of dangers from the perspective of young and adult cyclists. The
research depended on comparing visual search patterns and reaction times using a perception test
for cyclists. The test is based on recorded video clips, where each contains a dangerous traffic
situation that needs to be responded to. According to the results, it has been found that childhood
perception has a delayed reaction time than an adult, which may be the cause of the immature visual
system and the experience of a child. Therefore, a tailored educational program for children and
young cyclists could help to improve the perception and understanding of road hazards.

In the study of the authors Anagnostopoulos et al. [13] regard of the safety of cyclists at
crossroads, where the cyclists like to break the rules and often pass red, endangering not only
themselves, but also drivers of motor vehicles, is represented the mechanisms by which traffic
lights are able to perceive and react to cyclists. The Aim of the research is to give priority to
cyclists in front of the automobile transport, thereby reducing the accidents caused by cyclists
when crossing the red signal on traffic lights. According to the speed of the cyclist approaching
the intersection and using the app on the smartphone, which sends a signal to the traffic light, in
order to calculate the time of arrival of the cyclist, he could turn green. Cyclists would be the
first to drive without waiting, and there would be no collision with the motor vehicle, and this
would encourage people to use cycling more.

Cycling has many benefits from a more active way of living through a drop in motor traffic to
reducing air pollution. Unfortunately, even this type of transport cannot do without the risk that
is most often associated with the risk of injury to the cyclist or his surroundings. Teschke et al.
[14] research route infrastructures and risks of cyclist injuries in Toronto and Vancouver. There
were selected 14 routes, where cyclists are often injured. Routes have been observed and recorded
data on their status, street type, street lighting, interroads, traffic volume or speed of
operation. By using of the SAS analysis version 9.2; SAS Institute, Cary, NC and logistic
regression model, it was found that it depends on the type of route and hence the risk of injury.
The Results prove that the routes that are on larger and quieter streets without parked cars and
with a bicycle path off the street indicate less risk of injury to cyclists than opposite street
type. The aim of the study is to highlight the importance of a safe track infrastructure for
cycling, reducing the risk of cyclist accidents and contributing to greater citizens' safety and
activity. Other authors examining this issue are Reynolds and et al. [15], who conducted a study
using articles and other published research on the impact of transport infrastructure on the safety
of cyclists. They characterized two categories of infrastructure: the first category relates to
crossroads (roundabout), and the other refers to roads, paths and alleys. Using these categories
has been found, that a major impact on the reduction of the risk of injury have, not only
well-built cycle paths, but also street lights, paved surfaces or proper traffic signs. It remains,
of Course, that the increased risk of accidents occurs in the mass transport infrastructure of
cyclists and motor vehicles, as well as in common terrain with pedestrians. The theme is related to
the study by the authors of Panter and et al. [16], the impact of new transport infrastructure on
pedestrians, cyclists and physical activity. The Study examines whether the issuance of new
transport infrastructure for pedestrians, cycling and public transport will contribute to an
increase in time spent walking, cycling and overall daily movement activity. The project depended
on the construction of a new bus network and the adjacent pedestrian and cycling route. Using
quasi-experimental analysis and recruitment of research participants, the authors evaluated the
effects of the newly built transport infrastructure. The study showed increased use of bus
transport, cycling and more walking activities after the construction of new transport
infrastructure.


3.  Limited Resources and Game theory

The economic reality of the self-governing units predicts that it has limited resources to
implement security measures. In the case of resource constraints, the question is how to allocate
these resources to maximize the effect. In other words, it is about to keep crime as low as it is
possible, or to minimize the number of negative externalities.

Effective allocation of limited security resources is addressed by Pita et al. [17]. In their
study, they examine the deployment of an algorithm for the allocation of security patrols to
protect the airport, which are limited. They use game theory - based approach to allocation -
specifically the Bayesian Stackelberg game [18]. The prerequisite is that the terrorists are
watching the patrol and trying to attack the currently vacant place.  Random allocation is the key
to increasing security as it increases the risk to attackers. However, even this random allocation
must be systematic so that the highest probability of attacking the place being guarded is as high
as possible. This fact is being solved by the Bayesian Stackelberg game.

In order to solve the Bayesinan game using the Conitzer and Sandholm method [18], we have to
perform Harsanyi transformation [19]. This will convert the Bayesian game (incomplete information)
into a normal form game. An example of transformation is shown in the following figure 1. In the
game we assume that there are two players. One player is the police, who has only one type of
behavior. The other player is the attacker. However, he may be of a different nature (terrorist,
thief, drunk,…). Each type of attacker has its own behavior pattern. In the example for simplicity,
we assume two types of attackers.


                                               c

                                                  d

                                              a

                                               2,1

                                                  5,0

                                              b

                                               1,0

                                                  3,2

Figure 1: example

The Transformation assumes that an attacker of type 1 will attack with a probability of a. Striker
2 will attack with a probability of 1-a. The Transformation is then shown in the following figure
2.


                      cc’

                                   cd'

                                                     dc'

                                                                   dd'

                     a

                      2a + (1-a), 1

                                   2, a

                                                     5a+(1-a),(1-a)

                                                                   5a+2(1-a),0

                     b

                      a, (1-a)

                                   a + 3(1-a), 2(1-a)

                                                     3a, 2a+(1-a)

                                                                   3,2

Figure 2: transformation

The problem with this resolve lies in the range of strategic choices that are represented by a
normal game. If I have n types of attackers with strategies, the resulting game is comprised of kn
strategies. In the real world, we are getting into extremely large structures.

An alternative approach to the solution is decomposed optimal Bayssiaon Stackelberg solver (DOOBS).
The key to the DOBSS decomposition is the observation that evaluating the leader strategy against a
Harsanyi-transformed game matrix is equivalent to evaluating against each of the game matrices for
the individual follower types [20]

The Bayssiaon Stackelberg game pilot Application for the airport safety management in LA has had
extremely successful results after the pilot operation and continues its continued use [17].

4.  Neural networks to detect risk behaviour and prediction of risks

Neural networks can be very useful and can be used in systems to help keep safety in citites. This
issue deals with Murugan et al. [21] who examined inappropriate behaviour or anomalies in the
pedestrian zone using a regional scalable convolution neural network called RS-CNN. For the
calculation, the statistical model px (x) was determined for the distribution of the X measurement
under normal conditions. Abnormalities are defined as measurements that are less than the threshold
in this model. This is equal to the statistical hypothesis test:

·         H0 : x is obtained from px (x)

·         H1 : x is made out of a non-information separation px (x) a 1

Where px (x) < v, where v indicates the normalization constant of the non-informative separation
with the rule of minimum probability of test failure and rejection of the null hypothesis.

The first phase of the process is a segmentation, which moves where the video footage is structured
to identify area of anomalies. RS-CNN functions are calculated calculated for correlation
filtering. In the second phase occurs to marking classes, when the market labels are assigned to
each observed area in the video frame. Then the anomaly is detected. The detection level enters a
set of images in the video sequence with the observed areas. The output will then be anomaly images
with the appropriate label. To this end, the RS-CNN features were extracted, followed by the
mapping process of the regions of interest and marking. When the observed areas were found within
the frame, the relevant mark was allocated with the detection speed.

Anomalies are detected using this model, which are unusual or do not belong to the limitary zone
for pedestrians, such as cars, cyclists or inappropriate or brutal behaviour of pedestrians. This
technique can be utilized in systems to monitor abnormal activities in the video.

One of the most research parts is the movement of pedestrians using neural networks, which makes
sense for a number of applications, and which play an important role in the automotive industry or
robotics. Dai, Duan et al. [25] deal with infrared rays illuminated by pedestrians at night, which
can be seen through the visible spectrum camera over longer distances, helping to identify
pedestrians. The researchh uses a self-acting softmax with a 9-layer model of convolutionn neural
Network (CNN).

The result of the research contributes to reducing the collision of a motor vehicle with the
pedestrian at night, when the walker without the use of reflexive elements is not well visible.

The Same topic was examined by the authors Li et al [30], using a neural network to detect
pedestrians, which can be used in car driving assistance systems. The method uses neural functions
from fully convolution networks (FCN), detects pedestrians using the boundaries of green or red
rectangles. Green point to pedestrians who are nearby and red rectangles show, so put on fake
people, which means a poster with a character. The research uses three FCN models, which are
trained on the PASCAL VOC Segmentation command.

FCN is softer and more structural than traditional convolphic networks. The advantage of using a
neural network is to automatically learn the types of elements and thus achieve excellent
performance in detecting pedestrians. Research exceeds most traditional methods that use
hand-crafted functions.

Another author who dealt with the detection of pedestrians is Tome et al. [23], which proposes a
deep learning system that is able to customize a conventional network for a given task It Performs
a rigorous analysis and optimization of each step to achieve the accuracy of the task, approaching
the most modern approaches, while also requiring a low computational time. The research was based
on convolutional neural networks. These networks rely only on large-scale training data files and a
link-based training practice, along with an optimization algorithm. The first layers of the network
determine low-level concepts (edges, details), the final layers are able to combine low-level
functions to identify complex visual concepts. The ultimate research system is a combination of
LDCF region design algorithm and deep convolutional neural network.

The resulting system exceeds alternative approaches based on manual and learned functions, in
reasonable computational complexity. The version is able to detect pedestrians in real time on
modern hardware.

The topic of city security from the point of view of criminology was dealt with in the article by
Andersson et al [26], who using the 4-Cardinal Siamese convolutional Neural Network (4-CSCNN), to
investigated criminal behavior that is influenced by the environment where to criminal offences.
The analysis works with street level images to classify low and high criminality.

Using Neural networks we can predict the magnitude of the earthquake, as Adeli and Panakkat [28]
show in their article. They explore the class of neural networks used in the classification
problems of probability neural networks to predict earthquake quantities. A Model of probabilistic
neural networks provides good predicted precision for earthquakes between 4.5 and 6.0. Conversely,
when the magnitude of the earthquake is greater than 6.0, it does not provide good results for the
earthquake prediction. In an earthquake greater than 6.0 must be used model of the recurrent neural
network, which was previously developed by authors significantly more successful in the prediction
of earthquakes of magnitude between 6.0 and 7.5.


5.  Industry 4.0 and safety

Smart City is currently a very popular Area of research. In some cases, even voices say that it is
an overstated term. Smart City Solutions should ultimately bring benefits to the citizen in a given
locality. The Author Alsarhan et al [31] deals with the traffic control safety management. In their
analysis, they deal with the problem of limited available radio frequency spectrum. In the
analyses, a paid spectrum was used to minimise traffic accidents and communications across
transport participants. Numerical studies have shown that this approach could reduce accidents.
Tokoda et all. [32] in its publication deals with the intelligent transport and intelligent vehicle
system, which can improve both throughput and transport safety. The article deals mainly with
methodological approach to the issue.

The Contribution of Adb-Elha Taha [33] introduces new short-term metrics for road safety
assessment. The Metric uses machine learning in the core design of metric calculations using the
new Hidden Markov models (HMM) application. The impact of the proposed architecture is demonstrated
through a security planning application based on the current state.

Wan et al. [34] deals with emergencies that need to be effectively detected for  their subsequent
rapid resolution and removal. To this end, it proposes a structure composed of a central agent and
three layers – layers of unmanned aerial vehicles (UAVS), layers of multiple robots and layers of
sensor networks.

The authors Velladurai et al. [35] in their publication are dedicated to an intelligent system for
detecting toxic gases. Microcontrollers would be equipped with a GSM module, which would send a
monitoring report that could react quickly if a certain threshold is exceeded.

Srivastava et al. [36] In their overview map systems for security based on neural networks. The
authors Lacinak and Ristvej [37] discuss what it means to introduce the smart cities concept. Then
they are focused on the area of security, which is often underestimated in this respect. Finally,
they accentuate the importance of simulations.

6.  Conclusion

The article aims to perform a research of the current state and methods in the field of urban
safety. To this end, documents have been analysed that define the key infrastructures that ensure
basic human needs. In addition to the delimitation, approaches to their protection and
interdependence were analysed. In addition, systems and measures in the field of transport have
been analyzed, which can have a positive impact on increasing safety. Transport was chosen with
regard to the number of external deaths according to the Czech Statistical Office.

The scientific methods have analysed neural networks that can predict or classify certain
structures on the basis of identical parameters. This means that they can predict potentially risky
situations and thus allow authorities to take appropriate measures with a proper priority. Another
approach that has been analyzed is the game theory, which optimizes resource allocation while
protecting certain infrastructure. In this case, analyses were used that are practically used in
the protection of airports and which are inherently applicable to a number of criminal situations.

At least, the Smart city system related to urban safety, were analyzed. These systems often
intertwined with neural network methods or game theory. However, these were practical applications
related to the subject of the article.

The analyses will lead to the design of a safety city concept with regard to available of the local
government resources. Proposals will be directed to the cities with a population below 10,000.
citizens.

7.  Acknowledgment

The Article was created with the support of the Technolgic agency of the Czech Republic in solving
the grant TL02000559.

References

[1]          Kathi, B. A. (2006). Critical Path: A Brief History of Critical Infrastructure
Protection in the United States. Washington D.C.: Spectrum Publishing Group. First printing, 2006,
220 p. ISBN 978-0-913969-06-9.

[2]          Pederson, P. et al. (2006). Critical Infrastructure Interdependency Modeling: A Survey
of U.S. and International Research. USA, Idaho: Idaho National Laboratory, 2006. 116 p.

[3]          Critical Foundations: Protecting America`s Infrastructures (1997). The Report
of the President`s Commission on Critical Infrastructure Protection, 1st. edition. Washington D.C.:
United States Government Printing Office, 1997, 101 p.

[4]          HROMADA, M., LUKAS, L. (2012). Multicriterial evaluation of critical infrastructure
element protection in Czech Republic. In International Conference on Advanced Software Engineering
and Its Applications, ASEA 2012, and the 2012 International Conference on Disaster Recovery and
Business Continuity, DRBC 2012, Held in Conjunction with GST 2012. South Korea: Jeju Island, 28th
November – 2nd December, pp. 306-309. ISBN 978-161804004-6.

[5]          RINALDI, S.M., PEERENBOOM, J.P., KELLY, T.K. (2001). Identifying, Understanding and
Analyzing Critical Infrastructure Interdependencies. IEEE Control Systems Magazine, 2001, Vol. 21,
No. 6, pp. 11-25. ISSN 1066-033X. DOI: 10.1109/37.969131

[6]          HOKSTAD, P., UTNE, I. B., VATN, J. (2013). Risk and Interdependencies in Critical
Infrastructures: A Guideline for Analysis. Springer, 2013. 252 p. ISBN 978-1-4471-4661-2.

[7]          RINALDI, S.M. (2004). Modeling and simulating critical infrastructures and their
interdependencies. In Proceedings of the 37th Annual Hawaii International Conference on System
Science. 2004, 8 p.

[8]          YUSTA, J., CORREA, G., LACAL-ARÁNTEGUI, R. (2011). Methodologies and applications for
critical infrastructure protection: State-of-the-art. Energy Policy, 2011. Vol. 39, No. 10, pp.
6100-6119. ISSN 0301-4215.

[9]          MURRAY, A. T., GRUBESIC, T. H. (2012). Critical Infrastructure Protection: The
Vulnerability Conundrum. Telematics and Informatics, 2012, Vol. 29, No. 1, pp. 56-65.
DOI:10.1016/j.tele.2011.05.001

[10]        PUCHER J.,  DIJKSTRA L., 2003. „Promoting Safe Walking and Cycling to Improve Public
Health: Lessons From The Netherlands and Germany“, American Journal of Public Health 93, č. 9, s.
1509-1516. https://ajph.aphapublications.org/doi/full/10.2105/AJPH.93.9.1509

[11]        VASILEV, M.; PRITCHARD, R.; JONSSON, T., 2018., „Trialing a Road Lane to Bicycle Path
Redesign—Changes in Travel Behavior with a Focus on Users’ Route and Mode Choice.“ Sustainability ,
10, 4768. Dostupné z: https://www.mdpi.com/2071-1050/10/12/4768/htm#B27-sustainability-10-04768

[12]        ZEUWTS L., VANSTEENKISTE P., DECONINCK F., CARDON G., LENOIR M., 2017, „Hazard
perception in young cyclists and adult cyclists“, Accident Analysis & Prevention, 64-71., Dostupné
z: https://www-sciencedirect-com.ezproxy.techlib.cz/science/article/pii/S0001457516301476

[13]        ANAGNOSTOPOULOS T., FERREIRA D., SAMODELKIN A., AHMED M., KOSTAKOS V., 2016,
,,Cyclist-aware traffic lights through distributed smartphone sensing“, Pervasive and Mobile
Computing, 22-36., Dostupné z: https://www-sciencedirect
com.ezproxy.techlib.cz/science/article/pii/S1574119216000249

[14]        TESCHKE K., a kol., 2012., „Route Infrastructure and the Risk of Injuries to
Bicyclists: A Case-Crossover Study“, American Journal of Public Health®, Electronic ISSN:
1541-0048, Dostupné z: https://ajph.aphapublications.org/doi/10.2105/AJPH.2012.300762#

[15]        REYNOLDS C., a kol., 2009., „The impact of transportation infrastructure on bicycling
injuries and crashes: a review of the literature“, Environmental Health, ISSN: 1476-069X, Dostupné
z: https://ehjournal.biomedcentral.com/articles/10.1186/1476-069X-8-47

[16]        PANTER J. a kol., 2016, „Impact of New Transport Infrastructure on Walking, Cycling,
and Physical Activity“, American Journal of Preventive Medicine. Published by Elsevier Inc., 45-53,
Dostupné z: https://www.ajpmonline.org/article/S0749-3797(15)00622-4/fulltext

[17]        PITA J., M. JAIN, F. ORDÓÑEZ, Ch. PORTWAY, M. TAMBE, C. WESTERN, P. PARUCHURI, S.
KRAUS, Using Game Theory for Los Angeles Airport Security, AI Magazine, 30 (2009), pp. 43-57

[18]        Conitzer, V. a T. Sandholm, Computing the optimal strategy to commit to, Proceedings of
the ACM Conference on Electronic Commerce, 2006 (2006), pp. 82-90

[19]        Harsanyi, J. C., and Selten, R. 1972. A Generalized Nash Solution for Two-Person
Bargaining Games with Incomplete Information. Management Science 18(5): 80–106.

[20]        Paruchuri, P.; Pearce, J. P.; Marecki, J.; Tambe, M.; Ordóñez, F.; and Kraus, S. 2008.
Playing Games for Security: An Efficient Exact Algorithm for Solving Bayesian Stackelberg Games. In
Proceedings of the 2008 International Conference on Autonomous Agents and Multiagent Systems.
Richland, SC: International Foundation for Autonomous Agents and Multiagent Systems.

[21]        MURUGAN B. S., ELHOSENY M., SHANKAR K., UTHAYAKUMAR J., 2019., Computers & Electrical
Engineering., Region-based scalable smart system for anomaly detection in pedestrian walkways,
146-160, [cit. 2019-02-27], Dostupné z:
https://www-sciencedirect-com.ezproxy.techlib.cz/science/article/pii/S0045790618331847

[22]        ZHANG Y., ZHOU CH., CHANG F., KOT A. C., 2019. Neurocomputing, Multi-resolution
attention convolutional neural network for crowd counting, 144-152, [cit. 2019-02-27], Dostupné z:
https://www-sciencedirect-com.ezproxy.techlib.cz/science/article/pii/S0925231218312542

[23]        TOME D., MONTI F., BAROFFIO L., BONDI L., TAGLIASACCHI M., TUBARO S., 2016, Signal
Processing: Image Communication, Deep Convolutional Neural Networks for pedestrian detection,
482-489, [cit. 2019-02-26], Dostupné z:
https://www-sciencedirect-com.ezproxy.techlib.cz/science/article/pii/S0923596516300637

[24]        YANG B., CAO J., WANG N., ZHANG Y., ZOU L., 2018, Signal Processing: Image
Communication, Counting challenging crowds robustly using a multi-column multi-task convolutional
neural network,118-129, [cit. 2019-02-27], Dostupné z:
https://www-sciencedirect-com.ezproxy.techlib.cz/science/article/pii/S0923596518302194

[25]        DAI X., DUAN Y., HU J., LIU S., HU C., ON Y., CHEN D., LUO CH., MENG J., 2019, Infrared
Physics & Technology, Near infrared nighttime road pedestrians recognition based on convolutional
neural network, 25-32, [cit. 2019-02-26], Dostupné z:
https://www-sciencedirect-com.ezproxy.techlib.cz/science/article/pii/S135044951830690X

[26]        ANDERSSON V.O., BIRCK M.A.F., ARAUJO R.M. (2017) Investigating Crime Rate Prediction
Using Street-Level Images and Siamese Convolutional Neural Networks. In: Barone D., Teles E.,
Brackmann C. (eds) Computational Neuroscience. [cit. 2019-02-26], ISBN: 978-3-319-71011-2. Dostupné
z: https://link-springer-com.ezproxy.techlib.cz/chapter/10.1007%2F978-3-319-71011-2_7

[27]        PALOCSAY, S,W.,WANG, P., BROOKSHIRE, R. G., 2000, Socio-Economic Planning Sciences,
Predicting criminal recidivism using neural networks, 271-284, [cit. 2019-02-26], Dostupné z:
https://www-sciencedirect-com.ezproxy.techlib.cz/science/article/pii/S0038012100000033

[28]        ADELI, H., PANAKKAT, A., R., 2009, Neutral networks, A probabilistic neural network for
earthquake magnitude prediction, 1018-1024, [cit. 2019-02-26], Dostupné z:
https://www-sciencedirect-com.ezproxy.techlib.cz/science/article/pii/S0893608009000926

[30]        LI CH., WANG X., LIU W., 2017, Neurocomputing, Neural features for pedestrian
detection, 420-432, [cit. 2019-02-26], Dostupné z:
https://www-sciencedirect-com.ezproxy.techlib.cz/science/article/pii/S0925231217302710

[31]        Ayoub Alsarhan, Ahmed Y. Al-dubai, Geyong Min, Albert Y. Zomaya, Mohammad Bsoul, A New
Spectrum Management Scheme for Road Safety in Smart Cities, IEEE Transactions on Intelligent
Transportation Systems, 19 (2018), pp. 3496-3506

[32]        Daniel Tokody, Attila Albini, Laszlo Ady, Zoltan Rajnai, Ferenc Pongracz, Safety and
Security Through the Design of Autonomous Intelligent Vehicle Systems and Intelligent
Infrastructure in the Smart City, Interdisciplinary Description of Complex Systems, 16 (2018), pp.
384-396

[33]        Abd-Elhamid M. Taha, An IoT Architecture for Assessing Road Safety in Smart Cities,
Wireless Communications & Mobile Computing, (2018), pp. 11

[34]        Shuo Wan, Jiaxun Lu, Pingyi Fan, Khaled B. Letaief, To Smart City: Public Safety
Network Design for Emergency, IEEE Access, 6 (2018), pp. 1451-1460

[35]        V. S. Velladurai, M. Saravanan, R. Vigneshbabu, P. Karthikeyan, A. Dhlipkumar, Human
Safety System in Drainage, Unused Well And Garbage Alerting System for Smart City, IEEE:
International Conference on I-SMAC (IOT in Social, Mobile, Analytics and Cloud) (I-SMAC), (2017),
pp,. 6-9

[36]        Shweta Srivastava, Aditya Bisht, Neetu Narayan, Safety and Security in Smart Cities
Using Artificial Intelligence-A Review, IEEE: Proceedings of the 7th International Conference on
Cloud Computing, Data Science and Engineering,  (2017), pp. 130-133

[37]        M. Lacinák, J. Ristvej, Smart City, Safety and Security, Procedia Engineering, 192
(2017), pp. 522-527

[38]        Novotný, Petr. Určování regionálních subjektů a prvků kritické infrastruktury.
[Disertační práce]. Ostrava: Vysoká škola báňská – Technická univerzita Ostrava, Fakulta
bezpečnostního inženýrství, 2017. 147 s.