Intelligent Vehicular Safety System: A Novel Approach using IoT and CNN for Accident Detection and Rapid Rescue

© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 698
Intelligent Vehicular Safety System: A Novel Approach using IoT and
CNN for Accident Detection and Rapid Rescue
Riya Kapadia1 , Jash J. Joshi2
1Computer Engineering, K.J. Somaiya Institute of Technology, Sion, Mumbai, India
2Computer Engineering, K.J. Somaiya Institute of Technology, Sion, Mumbai, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Road accidents have become a serious issue for
the public. This paper offers a method for preventing
accidents that incorporates alcohol detection using a MQ3
alcohol sensor, followed by a message alert to a rescue
worker or family members. An SW-420 vibration sensor is
employed by the detecting component to recognize any
unexpected vibrations that might come from a collision.
Supervised deep learning CNN methods go along with this.
The front camera of the automobile is utilized to obtain a
picture of the accident site for the deep learning accident
prediction model. After a collision is discovered, notification
is delivered to the closest evacuation facility via GPS and
GSM devices. Once the Vehicle gets engaged in a collision,
the following vehicle will be alerted via VANET. The alcohol
sensor will then determine whether the driver has ingested
alcohol or not and if they need to operate the car in an
emergency while they are impaired. Then, as a consequence
of driving too rapidly, multiple accidents happened. So,
when the automobile exceeds the speed limit, an instant
warning will be transmitted by GSM Module. Finally,
Accident analysis device may be employed for smart cities
using supervised deep learning CNN algorithms.
Key Words: Internet of Things (IOT), speed limit,
MQ3 Alcohol Sensor, GSM, GPS, CNN and Accident
Detection.
1. INTRODUCTION
In contemporary society, anything and everything is now
reachable, particularly through transportation.
Transportation has varied over time, evolving since the
middle Ages carts towards the space vehicles of the 20th
millennium and beyond. Its swift development can be
ascribed to either fostering commerce or the demand for
speedy transit. Although many individuals believe that
transportation is a major advantage to humanity, others
are concerned that it can become a problem because of
excessive vehicle speeds and a lack of regard for traffic
safety rules.
In our everyday lives, we notice that the frequency of
accidents globally is escalating the fatal toll. According to
government estimates, accidents are believed to be the
cause of 140,000 recorded fatality cases every year.
According to statistics, rescue activities that are delayed
are the leading cause of mortality for accident victims. We
thus employed GSM and a MQ3 alcohol detector in our
suggested system to keep a watch on the automobile in
order to tackle this issue.
Accident fatality rates are decreased as well as the amount
of time an ambulance needs to travel to the hospital
thanks to IOT-based accident detection systems. Today,
since we have the Internet and there have been
technological developments in this sector, we are now in a
position to adopt IoT more efficiently and effectively.
Along with the usage of an IoT device in a car, we are also
in a good position to put technologies like machine
learning for sensor data and deep learning techniques like
CNN for image processing on the IoT board to deliver
significantly superior results. This powerful combo may be
perfectly utilized for road accidents and determine
accurately whether there has been an accident or not.
Once an accident is detected, GPS sensors can get the
actual location of the car. Once the accident data is sent to
a server, the server reads the GPS location of the vehicle
and spots the nearest hospital and emergency service.
Meanwhile, it sends a notification through an automated
call to the registered phone numbers by the user.
When a vehicle is engaged in an accident, an immediate
notification will be issued to the registered contact details.
Once an unplanned incident occurs the following car will
be alerted through VANET. If the driver is inebriated, the
notice will show promptly to the owner or the emergency
services. In parallel, the notice will be provided if the
automobile is involved in another collision. A prompt SMS
given to the registered mobile phone number when the car
exceeds the predetermined limit.
2. EXISTING METHODOLOGY
This study describes studies on the detection, localization,
reporting, modeling, and analysis of traffic accidents.
Accident detection systems based on smart phones have
been reported in various research. This study, Rajesh, G. et
al [1], offers a system that truly can identify traffic
problems and promptly communicate an emergency SMS
to the appropriate control center. It suggested Bhakat, A.
et al [2], a smart accident detection and rescue technology
that combines the IoTs (Internet of Things) and a system
with artificial intelligence to simulate the intellectual
processes of the human brain (AI). Choi, J.G and Kong [3],
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 11 | Nov 2023 www.irjet.net p-ISSN: 2395-0072

suggested a smart accident detection and rescue solution
that employs an artificial intelligence system and the
Internet of Things (IoTs) to imitate the intellectual
processes of a human brain (AI). A combination of an
accelerometer and an ultrasonic sensor is proposed for
accident investigation Zou. X et al [4].
A modified Haddon matrix is offered to offer various
insights on the future wave of road safety studies during
the intelligent, networked, and automated vehicular
technology. In order to report a traffic collision more
quickly, the proposed operational concept of the system is
based on Convolutional networks and deep learning
techniques Harikrishnan. A et al [5]. Many lives could be
saved by implementing this approach. Whenever an
accident is detected, the method utilizes GPS to find and
alert a nearby hospital. In this research Wegmeth L et al
[6], they suggest a machine-learning (ML) framework
based on multiple in-vehicle sensors for autonomous car
accident detection. Their study centers on the
identification of actual driving collisions utilizing cutting-
edge feature extraction algorithms and ordinary
automobile sensors.
Inter-vehicle communication solutions like VANET
(Vehicular Ad-hoc Network) and IoV (Internet of Vehicles)
may be able to aid vehicles in reporting incidents via each
other when a reliable Internet connection is only
accessible to some nearby vehicles. It suggested Comi, A et
al [7], a descriptive statistic would be used to decide which
methods of data mining are appropriate for evaluating
road accidents, as well as to identify their most important
root causes and frequent trends. They presented M. U.
Ghazi and M. A. Khan Khattak et al [8], the system
efficiently distributing emergency alerts is a huge
difficulty as a result of the several issues generated by this
high traffic density. The dynamic character of the network
makes employing VANETs for data transfer especially
difficult.The proposed approach T. Yuan et al [9], based on
the GIS and Firefly Clustering algorithms, can aid in the
detection of city road "black spots' ' other than the
contributes to minimizing car accidents and preserving
sustainable urban growth. It indicates that B. Du, L. Yu, X.
Hu et al [10], timely accurate traffic accident prediction
has a great deal of potential to preserve public safety and
reduce financial losses. That depicts Z. Huang and S. Gao et
al [11], the recommended method gives considerable
benefits for large-scale urban passenger hotspots in terms
of clustering speed, accuracy, and visualization. They have
used Fang and D. Yan [12], the prediction of driver
attention is growing into a major research topic for driving
systems that are similar to human people.
This research is aiming to forecast the driver's focus in
scenarios that entail crashes (DADA). A method suggests,
Singh, G. Pal, M et al [13], a descriptive analysis will be
performed to find which data mining approaches are ideal
for evaluating road accidents, as well as to highlight their
most significant underlying causes and common patterns.
It proposes Zhang, X. Rane et al [14], have employed
Regular monitoring permits early machine failure
identification, which is offering advantages for industrial
automation superior process control. The proposed
Goerlandt.F,Li and Reniers [15], indicates that it is vital to
provide realistic accident prevention techniques in this
paper in accordance with the pertinent safety protection
criteria and the existing status of every industry. The
Proposed method M. Mythili et al [16], is to present an
elegant and safe biometric attendance technique for smart
classrooms utilizing fingerprint sensors. If the instructor is
not present in the classroom through the GSM module, an
SMS alarm is dispatched to the relevant class incharge. The
suggested methodology R. Sathya et al [17], is built upon a
distinct image processing method for detection stage and
identification system using support vector machine for
offline signature verification. Author Sathya et al., [18]
proposed a strong SSVM+Hybrid LUCNN idea that has
been designed to recognize vehicle number plates for
intelligent transportation systems. By observing machine
tool status, the suggested study Sathya et al [19] built an
intellectual (IoT) tool state supervising scheme to find
construction tradeoffs connected to sustainability and the
best machining settings.
3.PROPOSED ACCIDENT ANALYZING SYSTEM
The fundamental purpose of this Proposed System is to
design a real-time application that leverages GSM and
MQ3 Alcohol sensors to identify and decrease accidents.
This aims to incorporate three parts in this study. When a
vehicle is engaged in an accident, an SMS alert is
immediately sent via GSM to the registered phone number.
The novel proposed Accident Analyzing scheme
Architecture is exposed in Fig. 1. Some accidents happened
as a consequence of drunk driving. So, we can swiftly
detect whether a person has consumed alcohol or not with
the assistance of a MQ3 Alcohol Sensor.
As soon as alcohol is detected, the information is sent to
the rescue officer or the victim's family. to caution them
from driving near their family. Lastly, going too fast might
sometimes result in accidents.
As a result, we have to establish an 80 km/h speed
restriction. Through the use of engine control module
sensors and vehicle speed sensors, the registered
cellphone number will be alerted as soon as the speed
limits are exceeded. This endeavor will help to reduce
accidents, which in turn will reduce the number of deaths.

Fig. 1. The entire suggested Accident Analyzing System
Architecture
3.1 Hardware Elements Included in the Accident
Detection Phase and Sensing Phase.
1. MQ3 Alcohol Sensor: The presence of alcohol gas is
detected [21], using this low-cost semiconductor sensor.
This sensor's sensitive SnO2 has a conductivity that is
lower than that of pure air. Its conductivity rises when
alcohol content is observed. Both analogue and digital
output are supplied by this module. It can simply be
connected to microcontrollers like the Raspberry PI,
Arduino boards, and others. If the individual is boozed
while driving then the quick message will be sent to the
rescue officers or family.
2. Radio Frequency Identification: RFID employs
electromagnetic combination in the radio range zone of
the electromagnetic spectrum to globally recognize a thing
or human individual. It will be delivered the signal via the
Radio Waves. Each and every automobile will have a
unique RFID Tag to receive the signals from the accident
vehicle to alert the incoming vehicle.
3. VANET: A vehicle ad hoc network approach is made up
of numerous moving or stationary automobiles that are
connected by a wireless network (VANET). VANETs were
mostly employed until recently to increase driver comfort
and safety in moving autos. Through the VANET, the cars
on the road were connected to a wireless medium. The
function of VANET will be demonstrated in Fig.2.
Fig. 2. The role of VANET
This will assist us to deliver the alarm signal through
waves. The VANET will be coupled with the RSU (Road
Side Unit). Once the motor begins VANET will be activated
through RSU.
4. GSM Module (SIM900A): The GSM module is switched
on if the deep learning model identifies an accident, and
employing location tracking, the data is provided as a text
message to the neighborhood emergency centre. The
message and collectively, the CGI is transmitted. This
message is a base transceiver station's internationally
unique identity. MNC, LAC, CI and MCC (Mobile Country
Code) are its four component sections (Cell Identification).
With the use of GPS, the CGI may be used to trace the
location.
5. Engine Control Module and Vehicle Speed Sensors: The
actuators in a combustion engine are governed by an ECU,
also referenced as an ECM, to guarantee optimal engine
performance. Then, the standard speedometer is replaced
by a speed sensor. It rotates while being attached into an
electrical connection that may provide a signal to a
computer. In this way, the sensor sends data for
calculating your car's speed. It also informs whether you
need to alter the transmission speed or shift levels.
3.2 Software Elements.
In this case, Tensor Flow works as the backend while
Keras serves as the neural network API. Tensor Flow is a
complete open resource Artificial Intelligence framework.
TensorFlow gives a large, complete network of
procedures, databases, and group of people qualities that

allow researchers to overcome the restrictions of machine
learning and implementers will promptly and efficiently
develop and organize machine learning starts.
4.IMPLEMENTATION FLOW OF PROPOSED
SYSTEM
The recommended idea comprises two steps, sensing
stage and avoidance stage. An Arduino Uno is connected to
the sensors for the preventive phase, according to the
block diagram mentioned in Fig. 3.
Fig. 3. Flowchart of accident detection and sensing.
Step 1: The Arduino Uno is interfaced with the alcohol
Sensor (MQ3), the SW 420 range of vibration sensor, and
the GSM module for delivering messages (L293D motor
driver).
Step 2: Using a USB connection, the raspberry pi module
receives the conclusion of the SW 420 range of this sensor.
If the output is designated as logic high, that will turn on
because it is linked to the raspberry pi module. The scene
is taken by this suggested module, and utilizing the deep
network layer constructed in the component layer, the
image is compared with the specified dataset.
Step 3: After the boozed individual rides a car and meets
with an accident the quick communication will be sent to
the rescue team.
Step 4: When the speed limit is outside the range the
immediate Alert notice will be issued to the relatives or
rescue squad.
5. RESULTS
1. Results of Vehicular Accident Detecting: The sensing
phase involves accident identification, detection and
tracking with utilizing a deep network layer and
intimation of vehicle accident incidents to the nearby
rescue center. The function of the system will be clearly
depicted in Fig.4. The SW420 vibration sensor is utilized to
establish collision detection. When an abnormal jerk or
vibration is detected that exceeds the specified threshold,
the sensor output spikes, triggering the buzzer. We have
set the threshold value at 80000 for prototype
demonstration purposes. The rescue team will then get the
alert message quickly.
Fig. 4. Accident Detection using GSM.
2. Results of Vehicular Alcohol Sensing: Internal
initialization has set the MQ3 sensor's threshold at 500
ppm. The GSM will be engaged by RSU once the alcohol
concentration surpasses the threshold. The nominated
cellphone number will thereafter get the alert message
over GSM. The exact phenomenon has been proven using
an L293D motor driver. An alcohol based hand sanitizer
has been deployed for testing reasons; when it is
discovered, a notification is sent.
3. Results of Accident Recognition using CNN: It is
recommended that a deep learning Convolutional neural
network be used for each frame of a motion picture that
has been trained to discriminate between accident- and
non-accident-related video frames. It has been established
that Convolutional Neural Networks [20], offer a rapid and
reliable way for identifying pictures. For considerably
smaller datasets, CNN-based image classifiers have
achieved accuracy levels of exceeding 95% and
unnecessary preprocessing than other image classifying
algorithms.
4. Results of Speed Alert Detection: The RSU will activate
both the Engine Management Module and Vehicle Speed
Sensor when the car exceeds the speed restriction
80km/h. The limit of the vehicle fixed is indicated in Fig.5.

Fig. 5. Speed Guard.
The GSM Module will then activate the two sensors. The
rescue squad will instantly get an alarm message if the
GSM Module senses vibration.
6. CONCLUSION
In comparison to the present methods, the proposed
solution is much more trustworthy and may be more
efficient. We are able to track, monitor, and establish the
whereabouts of both persons and vehicles. The primary
benefit of this research is that it allows us to prevent the
loss of life by stopping intoxicated drivers from driving
automobiles and alerting them to a speed alert. When
compared to the present methods, the innovative accident
sensing and vehicle accident detection method delivers
better accuracy. In the future, if a car is stolen, we will be
able to follow its whereabouts in real-time using a GPRS
module.
7. REFERENCES
[1] Rajesh, A.R. Benny, A. Harikrishnan, J.J. Abraham, N.P.
John, “A Deep Learning based Accident Detection System”.
In- Proceedings of the 2020 International Conference on
Communication and Signal Processing pp. 1322–1325, 28–
30th July 2020.
[2] A. Bhakat, N. Chahar, V. Vijayasherly, “Vehicle Accident
Detection and Alert System using IoT and Artificial
Intelligence” In: Proceedings of the 2021 Asian Conference
on Innovation in Technology, pp. 1–7, 27–29th August
2021.
[3] J.G. Choi, C.W. Kong, G. Kim, S. Lim, “Car crash detection
using ensemble deep learning and multimodal data from
dashboard cameras”. In: Expert Scheme for Young
Scientists and Technologists Appl. 183, 115400 in 2021.
[4] X. Zou, H.L. Vu, H. Huang, Fifty Years of “Accident
Analysis & Prevention: A Bibliometric and Scientometric
Overview” Anal. Prev. 144, 105568 in 2020.
[5] A. Harikrishnan, J.J. Abraham, N.P. John, “A Machine
learning based Accident Detection System”. In:
Proceedings of the 2021 International Conference on
Communication and Signal Processing, pp. 1325–1327,
28–30th July 2021.
[6] H.H. Pour, F. Li, L. Wegmeth, C. Trense, R. Doniec, M.
Grzegorzek, R. Wismüller “A Machine Learning
Framework for Automated Accident Detection Based on
Multimodal Sensors” Sens. 2022, pp. 1–21 in 2022.
[7] A. Comi, A. Polimeni, C. Balsamo “Road Accident
Analysis with Data Mining Approach: Evidence from
Rome” Transp. Res. Procedia 2022, 62, 798–805 in 2022.
[8] M. A. Khan Khattak, A. W. Malik, M. U. Ghazi, M. S.
Ramzan, and B. Shabir “Emergency message dissemination
in vehicular networks: A review”, In: IEEE Access, vol. 8,
pp. 38606–38621 in 2020.
[9] T. Yuan, T. Shi, and X. Zeng “Identifying urban road
black spots with a novel method based on the firefly
clustering algorithm and a geographic information
system,” Sustainability, vol. 12, pp. 2091 in 2020.
[10] B. Du, L. Sun, L. Han, W. Lv, and X. Hu "Deep
spatiotemporal graph convolutional network for traffic
accident prediction” Neurocomputing, vol. 423, pp. 135–
147, Jan. 2021.
[11] Huang. Z, Gao.S, Cai. C, Zheng. H and Pan. Z, "A rapid
density method for taxi passengers hot spot recognition
and visualization based on DBSCAN+" Scientific Reports,
vol. 11, pp. 1-13 in 2021.
[12] Fang. J, Yan. D, Qiao. J, and Xue. J “DADA: Driver
attention prediction in driving accident scenarios” IEEE
Trans. Intell. Transp.Syst., early access, with doi:
10.1109/TITS.2020.3044678 on Jan1 2021.
[13] G. Singh, M. Pal, Y. Yadav, T. Singla "Deep neural
network-based predictive modeling of road accidents"
Neural Comput. Appl. 2020, pp 12417–12426, 2020.
[14] X. Zhang, K.P.Rane, I. Kakaravada, M. Shabaz Research
on "vibration monitoring and fault diagnosis of rotating
machinery based on IOT" Nonlinear Eng. 2021, pp. 245–
254 in 2021.

[15] F. Goerlandt, J.Li, G. Reniers "Virtual Special Issue:
Mapping Safety Science" Reviewing Safety Research. Saf.
Sci. 2021, pp. 105278 in 2021.
[16] Andrea Vilan and Pamela Walker, “A Review of CPS
Design and Vulnerability Analysis", vol.2, no.3, pp. 110-
119, July 2022. doi: 10.53759/181X/JCNS202202014.
[17] R. Sathya, S. Ananthi, R. Rupika, N. Santhiya, and K.
Lavanya, "Average Intensity Sign Feature-based Offline
Signature Verification for Forgery Detection using
Machine Learning" International Conference on
Augmented Intelligence and Sustainable Systems 2022, pp.
325-330, doi: 10.1109/ICAISS55157.2022.10010812.
[18] Sathya Ramasamy, Ananthi Selvarajan, Vaidehi
Kaliyaperumal, and Prasanth Aruchamy “A hybrid
location-dependent ultra CNN based vehicle number plate
recognition approach for intelligent transportation
systems” Concurrency and Computation: Practice and
Experience, Jan 2023, doi: 10.1002/cpe.7615.
[19] V. N. Kumar, G. Singh, S. Rudresha, and S. Sampath
Kumar, "Tool condition monitoring by quality during the
micro milling process by using Internet of things and
Artificial Intelligence" 6th International Conference on
Electronics, Communication and Aerospace Technology in
2022, pp. 01-04, doi:
10.1109/ICECA55336.2022.10009545.
[20] Pandiaraja. P, Muthumanickam. K, “Convolutional
Neural Network-Based Approach to Detect COVID-19 from
Chest X-Ray Images” In: Lecture Notes in Networks and
Systems, vol 370. Pp. 231-245, 2022.
[21] P. Vijayalakshmi, K. Muthumanickam, G. Karthik, S.
Sakthivel, “Diagnosis of Infertility from Adenomyosis and
Endometriosis through Entroxon Based Intelligent Water
Drop Back Propagation Neural Networks” Journal of
Intelligent & Fuzzy Systems, vol. 43, no. 3, pp. 2243-2251,
2022.

Intelligent Vehicular Safety System: A Novel Approach using IoT and CNN for Accident Detection and Rapid Rescue

Recommended

Recommended

More Related Content

Similar to Intelligent Vehicular Safety System: A Novel Approach using IoT and CNN for Accident Detection and Rapid Rescue

Similar to Intelligent Vehicular Safety System: A Novel Approach using IoT and CNN for Accident Detection and Rapid Rescue (20)

More from IRJET Journal

More from IRJET Journal (20)

Recently uploaded

Recently uploaded (20)

Intelligent Vehicular Safety System: A Novel Approach using IoT and CNN for Accident Detection and Rapid Rescue