Crowd Anomaly Detection Systems Using RFID

and WSN Review

W. Raad A. Hussein M. Mohandes

Computer Engineering Dept. Electrical Engineering Dept. Electrical Engineering Dept.
KFUPM KFUPM KFUPM
Dhahran, Saudi Arabia Dhahran, Saudi Arabia Dhahran, Saudi Arabia
husseina@kfupm.edu.sa mohandes@kfupm.edu.sa
2021 4th International Symposium on Advanced Electrical and Communication Technologies (ISAECT) | 978-1-6654-3773-8/21/$31.00 ©2021 IEEE | DOI: 10.1109/ISAECT53699.2021.9668517

raad@kfupm.edu.sa

B. Liu A. Al-Shaikhi
Electrical Engineering Dept. Electrical Engineering Dept.
KFUPM KFUPM
Dhahran, Saudi Arabia Dhahran, Saudi Arabia
boliu@kfupm.edu.sa shaikhi@kfupm.edu.sa

Abstract— The security and safety of public places have been crowd anomaly detection using emerging technologies like
a concern to more entities in recent decades. Video surveillance Radio Frequency Identification (RFID).
is widely used to guarantee the security and safety of public A data-driven framework has been used for detecting
places. Yet, abnormal crowd movement detection and urban crowd anomaly, as shown in Fig. 1, which includes
estimation are essential in video surveillance to avoid incidents
three components:
like a stampede. The most challenging problem is detecting and
locating people in moving dense crowds with obstacles like • Urban data: are spatiotemporal data, along with
occlusion and blind spots where traditional video surveillance timestamps and location tags collected by cellular
techniques fail. This paper presents an extensive review of state- devices.
of-the-art advances in detecting abnormal behavior in dense • Urban dynamics: are variations in urban physical
crowds approaches. The techniques are based on range-free status, including urban mobility of crowds or traffic
localization for detecting the direction and speed of the crowd flow.
movement. Radio Frequency Identification (RFID) and
• Urban events: are related to social or individual
Wireless Sensor Networks (WSN) are surveyed. RFID is used by
analyzing the Received Signal Strength Indicator (RSSI) for activities taking place in urban areas. They
detecting orientation and speed of crowd movement and constitute the real causes of urban dynamics.
provides information like the crowd density, movement velocity, Compared with traditional techniques that rely on human
flow rate, and the number of persons passing to detect the onset observations, the data-driven frameworks are low-cost and
of a stampede. more efficient.

Keywords—crowd anomaly, RFID, WSN, RSSI, stampede

I. INTRODUCTION
Due to the unprecedented scale and speed of urbanization,
cities face the daunting task of keeping up with urban
dynamics. In addition to that, public places like airports, train
stations, and shopping centers are typically crowded. The
flow of people is growing consistently with the development
of urbanization. This growth causes an increase in the risk of
crowd stampedes in these places. Moreover, religious, sports
or social events are inherently overcrowded, and numerous
stampede accidents have occurred in such events. Examples
of stampede accidents during religious events include 700
people who died in a pilgrimage season in Saudi Arabia in
2015 and 115 people in 2013 during the Hindu festival of
Navatri [1]. Social incidents include 375 people who died in
2010 during Water Festival in Cambodia and 56 people killed
in a stampede at a funeral in Iran in 2020 [2]. Sports events Fig. 1. Data-Driven Urban Anomaly Detection [5]
include a stampede in a stadium in South Africa in 2001,
causing 43 people to die, and 93 people were killed in a Crowd detection and counting techniques can help prevent
stadium in Nepal in 1988 [3]. Therefore, the crowd must be catastrophic accidents by providing crucial information about
monitored efficiently for crowd anomaly detection to prevent the number of people and crowd density in a scene in real-
similar losses[4]. time.
Urban anomalies are typically unusual events, such as Due to recent advancements in information technology,
traffic congestion and unexpected crowd gathering, which techniques like smart video surveillance are needed for real-
may pose tremendous threats to public safety and stability if time monitoring of human activities such as crowd counting.
not timely handled. As a result, many researchers tackled These systems have gained more interest to overcome specific

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

Authorized licensed use limited to: Shah and Anchor Kutchhi Engineering College. Downloaded on August 14,2023 at 10:31:46 UTC from IEEE Xplore. Restrictions apply.
 challenges such as moving/non-moving object detection,
multi-person tracking, classification, complex human motion
analysis, and activity understanding. In this work, we survey
various device-free RFID techniques for detecting the onset of
a stampede in moving dense crowds, and state-of-the-art
advances in video surveillance methods to count people in
dense crowds.
II. RFID AND WSN FOR STAMPEDE DETECTION
A. RFID for Stampeded Detection
The RFID technology incorporates the use of Fig. 2. Passive RFID Based System for Crowd detection
electromagnetic or electrostatic coupling in the Radio
Frequency (RF) of the electromagnetic spectrum. It is used B. WSN for Stampeded detection
for the identification of people, objects, or animals. Every Another popular technology for tracking is the Wireless
RFID system consists of two basic components: a tag and a Sensor Network (WSN). It is defined as a set of autonomous
reader. The data is stored on tags; these tags may be either nodes, which communicate among each other using wireless
active, passive, or battery-assisted passive (BAP). The BAP media. WSNs constitute infrastructure-less and self-
and active tags utilize batteries to communicate over a more configured networks to monitor environmental conditions,
extended range up to 2 km. including the concentration and motion of objects.
Active and BAP tags have a disadvantage of the finite The sensors pass their data cooperatively through the
battery life and are larger and more expensive than passive network to the central location where it is observed and
tags due to their integrated power cells. Nevertheless, they analyzed. A wireless sensor node is equipped with a
allow wider ranges, are less sensitive to electromagnetic computing module, sensing module, radio transceiver, and
noise, and offer higher transmission rates [6]. In RFID-based power component. The power for each sensor node is
tracking systems, every user wears an active or passive RFID supplied from a battery. Nodes in a WSN have limited
tag. Once the tag is in the detection range of an RFID reader, processing speed, storage capacity, and communication
the reader transmits an interrogation signal to the tag and bandwidth. The sensor nodes self-organize a suitable
receives a reflected electromagnetic signal from the tag. infrastructure of the network, often with multi-hop
RFID-based systems detect people in large, crowded events communications. WSNs are utilized in various tracking
like the sports or pilgrimage seasons [7-13]. applications ranging from industrial, environmental,
Unlike battery-powered tags, passive tags use the RF agricultural, and social events [12, 18, 19]. WSNs are often
signal received from the reader to produce power and densely deployed in the field. Sensor nodes often vary in size
transmit/receive data. The RFID system, particularly the ranging from a shoebox to a tiny grain of dust. It is possible
passive RFID, is a low-cost technology for energy harvesting, to deploy sensor nodes in ad-hoc or planned schemes. An ad-
saving lots of power compared to wireless sensor solutions hoc deployment is suitable for large size terrain for
and Global Positioning System (GPS) technologies. Passive unattended applications. In this scheme, maintenance is
RFID systems at the UHF band have the highest coverage complex for a large number of nodes [20]. Fig. 3 shows a
range for this type of tag. The communication between the WSN for crowd event detection.
tag and the reader takes place using an electromagnetic
backscatter up to a maximum range of 25 meters depending
upon the tag and the reader antennas' specifications. An RFID
reader can read/write data to the tags using wireless
communication in a single hop. An RFID reader reads
information stored on the tag and passes it to a back-end
computer for further analysis. RFID middleware has the
function of collecting and processing the data. An RFID
reader broadcasts to all tags within its range, selecting a
particular tag for communication, and exchanging
information with the selected tag.
One of the important applications of RFID is real-time
tracking of people such as the elderly or patients. Passive
RFID is usually used for people's identification based on the
ID number of the tag they hold [12, 14, 15]. RFID-based
localization technologies are popular for indoor tracking
schemes since GPS-based localization systems perform Fig. 3. Wireless Sensor Network
poorly in location tracking indoor. In addition, they are vital
enablers for many novel applications, mainly due to their C. Hybrid RFID/WSN Systems
drastically reduced cost[9, 16, 17]. See Fig. 2 for a passive WSN and RFID are complementary as they were originally
RFID-based system for crowd detection. designed with different functionalities (RFID for
identification while WSN for sensing). For this reason, the
integration of both RFID and WSN provides a significantly

Authorized licensed use limited to: Shah and Anchor Kutchhi Engineering College. Downloaded on August 14,2023 at 10:31:46 UTC from IEEE Xplore. Restrictions apply.
 improved performance. This integration incorporates smart data signatures for large groups are handled. This latter
fixed and mobile devices resulting in novel applications algorithm was validated on real-time experiments for
covering a variety of industrial, commercial, and human continuous movement of up to 75 people with an accuracy of
sensing domains. It enables the RFID to work in multi-hop to 90%. See Fig. 5 for an RSSI-based system for detecting the
extend applications to operate in a broader area. Furthermore, direction and speed of crowd movement and the onset of the
it enhances the capability of these two combined key stampede.
technologies to collect, aggregate, and analyze
environmental, real-time data about objects, their
localization, and ecological conditions.
For example, see Fig. 4 for utilizing a hybrid RFID &
WSN, including ZigBee for crowd detection. Next, we will
describe briefly some of the work done for each of these
technologies.

Fig. 5. Utilizing RSSI for Detecting the Direction of Motion

Many researchers have used wireless sensor networks

based on ZigBee protocol using RSSI combined with RFID
bracelets for crowd estimation. In this case, the system works
by installing readers at portals or gates where pedestrians
pass. Wang et al. [23] have conducted a survey comparing
various crowd estimation techniques, including WSN, RFID,
Wi-Fi, and others, for activity recognition purposes taking
Fig. 4. A Hybrid RFID & WSN
advantage of human body attenuation/or channel fading of
Chen et al. [21] proposed a passive RFID technology for wireless radio. Compared with traditional activity
counting the number of persons in a single channel by recognition, radio-based recognition exploits radio
analyzing the Received Signal Strength Indicator (RSSI). communication to obtain high recognition accuracy, reduce
RSSI is an indicator of the power level received from a energy costs, and preserve privacy.
transmitting antenna. It has been observed that unique Yamin et al. [24] used a combination of various
identifiable signatures of RSSI variations are correlated with technologies for detecting stampedes in crowds, including
the number of people crossing a gate or a portal where an RFID, WSN, IP cameras, and smart devices. They proposed
RFID reader is installed. Therefore, RSSI can monitor a an image-processing algorithm for detecting the onset of the
moving crowd and provide information like the crowd stampede in crowds. The algorithm is based on processing
density, movement velocity, flow rate, and the number of images sent through IP camera to the central processing unit,
persons passing to detect the onset of a stampede. Moreover, counting the number of heads. Once the number of heads
suppose each person wears an RFID tag. In that case, the exceeds a certain threshold, it indicates the onset of stampede,
parameters of pedestrians' movement, i.e., the time and and notifications are sent to the person in charge to act. As
location of a particular person, are obtained when they pass part of the healthcare management subsystem, the authors
in front of the corresponding antenna. Gupta et al. [22] used integrated several mobile applications and developed
device-free techniques to monitor crowd movement by applications dealing with relief issues, blood donations,
distributing RFID tags along the pathways. They focused on complaints, and alerts, and utilizing mobile phones as a
detecting the speed of a moving crowd based on the sensor device.
signatures of the RSSI variations. However, the increase of RFID utilization for tracking people and vehicles can
the crowd density beyond a specific limit tends to block the effectively manage the crowd, including the stampedes and
tag reading completely. Thus, getting large periods of very disasters. The RFID transponders are easily linked to WSN,
low RSSI levels or no reading instead of a series of leading GPS, or cellular (3G, 4G, and 5G) networks. The engineering
and trailing edges. Two algorithms have been developed to process of selecting one of these technologies would highly
solve this problem. The first is a spatial occupancy-based depend on the geographical terrain where the RFID
approach to remove the effect of velocity on active intervals. transponders are deployed. In urban areas, the width of
However, this works only for a limited crowd density. Thus, streets, density of buildings, and congestion would have to be
a time difference waveform correlation algorithm using the considered. Utilizing WSN for aggregating signals from
divide and conquer approach was developed, where the test RFID transponders provides highly accurate latitudinal and

Authorized licensed use limited to: Shah and Anchor Kutchhi Engineering College. Downloaded on August 14,2023 at 10:31:46 UTC from IEEE Xplore. Restrictions apply.
 longitudinal coordinates. Nevertheless, the installation of nature of human movement. Understanding human behavior
WSN is costly as well as hazardous in terrain with limitations. is still one of the challenges for computer vision. Other
methods are proposed in the literature for this purpose.
III. DISCUSSION Artificial intelligence-based techniques open the avenue for
Table I summarizes an extensive comparison between WSN breakthroughs in dense moving crowds and harsh weather
and RFID technologies in detecting the onset of stampede in conditions. Machine learning techniques come at the
crowds [6-19]. forefront for solving pending problems in traditional
computer vision paradigms to detect anomalies in dense
TABLE 1. A COMPARATIVE SURVEY BETWEEN RFID AND WSN TECHNIQUES crowds.
Factor WSN RFID
Another open problem is the gap between existing
Monitor and sense researches and the final goal of diagnosing the underlying
Localization & problems. While urban anomaly detection and prediction help
Main Function environmental
identification
parameters us be aware of anomalous events, analyzing the root causes
Reflects RF signal helps to decide what kind of remedy should be taken, mainly
Aggregate, Process, transmitted from a if the various events are on different emergency levels and
Tasks
Transfer, and Store reader for localization of
the attached object
should be handled in different ways. However, finding the
Limited range of underlying problems is difficult because the causes of urban
communication With passive tag (2-3 anomalies can be reflected by different urban data and have a
Range using the multi hub meters), Active tag significant spatial and temporal misalignment.
for accessing sink (100-200 meters)
nodes IV. CONCLUSIONS
Safety, wellbeing, Track and trace,
Application healthcare, smart healthcare, crowd The paper has shed light on the latest literature review,
grid estimation, etc. particularly utilizing state-of-the-art advances in RFID to
Wi-Fi 802.11 detect the onset of a stampede in dense crowds by using RSSI
Protocol WLAN, ZigBee RFID protocols
802.15.4
to detect the speed and direction of a moving crowd.
Communication Furthermore, this work paves the way for further extensive
Multi hup Single hub
and connection research to find cost-effective, privacy conserving, robust,
Mobility Static Mobile and scalable hybrid techniques for public safety. RFID range-
free techniques can be utilized to obtain the highest accuracy
Deployment Random or fixed Attached to objects
in detecting the onset of the stampede in dense moving
Power and Battery needed for Batteryless for passive crowds under abnormal environmental conditions to
energy sensors tags
Range, Power, Communication, guarantee public safety.
Limitation Architecture, Real- Cost of Active tags, Regarding future work related to open problems, urban big
time Applications Security & Privacy data-based methods are the future and a promising trend of
urban anomaly analysis. Artificial Intelligence (AI) and
A significant number of studies have been done in recent Machine Learning (ML) are expected to impact future
years. However, several open problems have not been well directions in this field.
addressed, such as precise prediction and underlying problem
diagnosis. ACKNOWLEDGMENT
The source of the problems is the difficulty brought up by The Deanship of Research at King Fahd University of
the complexity of big urban data. The main challenge here is Petroleum & Minerals supported this work under grant
that urban data comes from different sources, posing a RG1424.
challenge to process efficiently large volumes of data
suffering from spatial and temporal misalignment. The real- REFERENCES
time fusion and spatiotemporal alignment of big urban data [1] (2015, October 25, 2021). Timeline: Deadliest stampedes.
from different sources like videos of pedestrian flows and Available: https://www.bbc.com/news/world-11819540
[2] (2020, October 25, 2021). 56 Killed in Stampede at Funeral for
vehicle flows is still an unsolved problem. Reliability Iranian Qasem Soleimani. Available:
detection is another open problem. https://time.com/5760392/qasem-soleimani-funeral-stampede/
False-positive and false-negative reports can both lead to [3] (1988, October 25, 2021). 93 Die in Nepal Stadium Stampede :
wrong decisions. The impacts of urban anomalies are usually Soccer Fans Rush to Locked Exits in Sudden Hailstorm.
Available: https://www.latimes.com/archives/la-xpm-1988-03-
composed of components reflected by various types of data. 13-mn-1821-story.html
To ensure the reliability of an urban anomaly detection [4] K. R. Ahuja and N. N. Charniya, "A Survey of Recent Advances
system, we should consider multiple sources of data. For in Crowd Density Estimation using Image Processing," in 2019
example, the audiences attending a concert can choose International Conference on Communication and Electronics
Systems (ICCES), 2019, pp. 1207-1213.
multiple types of transportation such as subway, taxi, or [5] M. Zhang, T. Li, Y. Yu, Y. Li, P. Hui, and Y. Zheng, "Urban
shared bicycles. Merely depending on one source might lead Anomaly Analytics: Description, Detection and Prediction,"
to underestimating the crowd density in a concert or even IEEE Transactions on Big Data, pp. 1-1, 2020.
failure to report the event. Hence, multiple data sources [6] M. El-Absi, A. A. Abbas, A. Abuelhaija, K. Solbach, and T.
Kaiser, "Distance and Tag Aware Localization in Indoor
should be effectively merged. Terahertz Systems," in 2018 First International Workshop on
The analysis of crowd behavior is a very vital topic of Mobile Terahertz Systems (IWMTS), 2018, pp. 1-5.
research in computer vision. Modeling and classifying human [7] M. Mohandes, "A case study of an RFID-based system for
activities are not trivial due to the randomness and complex pilgrim identification abd tracking," Sustainable Radio
Frequency Identification Solutions, INTEC, pp. 87-104, 2010.

Authorized licensed use limited to: Shah and Anchor Kutchhi Engineering College. Downloaded on August 14,2023 at 10:31:46 UTC from IEEE Xplore. Restrictions apply.
 [16] Q. Ma, X. Li, G. Li, B. Ning, M. Bai, and X. Wang, "MRLIHT:
[8] A. H. Al-Hashedi, M. R. M. Arshad, A. S. Baharudin, and H. H. Mobile RFID-Based Localization for Indoor Human Tracking,"
Mohamed, "RFID applications in Hajj management system," in Sensors, vol. 20, 2020.
2013 IEEE International Conference on RFID-Technologies and [17] Y. Zhang, M. G. Amin, and S. Kaushik, "Localization and
Applications (RFID-TA), 2013, pp. 1-6. Tracking of Passive RFID Tags Based on Direction Estimation,"
[9] Y. Duroc and S. Tedjini, "RFID: A key technology for International Journal of Antennas and Propagation, vol. 2007, p.
Humanity," Comptes Rendus Physique, vol. 19, pp. 64-71, 017426, 2007/12/24 2007.
2018/01/01/ 2018. [18] M. Mohandes, M. Haleem, M. Deriche, and K. Balakrishnan,
[10] E. A. Khan, "An RFID-Based System for Pilgrim Management "Wireless Sensor Networks for Pilgrims Tracking," IEEE
in King Abdul Aziz International Airport," in 2011 International Embedded Systems Letters, vol. 4, pp. 106-109, 2012.
Conference on Information Management, Innovation [19] M. Mohandes, M. A. Haleem, M. Kousa, and K. Balakrishnan,
Management and Industrial Engineering, 2011, pp. 124-129. "Pilgrim Tracking and Identification Using Wireless Sensor
[11] A. Tufail, "Pilgrim Tracking and Location Based Services Using Networks and GPS in a Mobile Phone," Arabian Journal for
RFID and Wireless Sensor Networks," International Journal of Science and Engineering, vol. 38, pp. 2135-2141, 2013/08/01
Computer Science and Network Security (1738-7906), vol. 18, 2013.
pp. 112-119, 06/30 2018. [20] D. G. Pascual, P. Daponte, and U. Kumar, Handbook of Industry
[12] M. Mohandes, "Pilgrim tracking and identification using the 4.0 and SMART Systems: CRC Press, 2019.
mobile phone," in 2011 IEEE 15th International Symposium on [21] L. Chen, W.-g. Song, and F.-z. Huo, "Study of Personnel
Consumer Electronics (ISCE), 2011, pp. 196-199. Monitoring and Counting in Single Channel based on RFID
[13] Y. Fan and Q. Liang, "An Improved Method for Detection of the Technology," Procedia Engineering, vol. 71, pp. 529-536, 2014.
Pedestrian Flow Based on RFID," in 2017 IEEE International [22] G. Gupta, V. Bhope, J. Singh, and A. R. Harish, "Device-Free
Conference on Computational Science and Engineering (CSE) Crowd Count Estimation Using Passive UHF RFID
and IEEE International Conference on Embedded and Technology," IEEE Journal of Radio Frequency Identification,
Ubiquitous Computing (EUC), 2017, pp. 69-72. vol. 3, pp. 3-13, 2019.
[14] A. Soman and S. Jacob, "An Efficient and Decisive Crowd [23] S. Wang and G. Zhou, "A review on radio based activity
Management System Based On RFID Technology," recognition," Digital Communications and Networks, vol. 1, pp.
International Journal of Research Ideas and Innovations in 20-29, 2015/02/01/ 2015.
Technology, vol. 4, p. 4, 2018. [24] M. Yamin and Y. Ades, "Crowd Management with RFID and
[15] M. Mohandes, "RFID-based system for pilgrims identification & Wireless Technologies," in 2009 First International Conference
tracking", Journal of Appli. Comput. Electromagn. Soc. (ACES), on Networks & Communications, 2009, pp. 439-442.
vol. 25 (3), pp. 273-282, 2010.

Authorized licensed use limited to: Shah and Anchor Kutchhi Engineering College. Downloaded on August 14,2023 at 10:31:46 UTC from IEEE Xplore. Restrictions apply.