2. Literature Survey
Mr. S. Kailasam, Mr. Karthiga, Dr. Kartheeban, R.M. Priyadarshani, K Anithadevi[
1
] states that due
to lack of attention, Drowsiness, and drunk driving are the major causes of road accidents, this paper
proposes preparing a system to prevent these circumstances. The proposed system herein aims at
preventing and controlling accidents by using a Night Vision Camera. This system monitors the driver’s
face when the car starts which mainly helps in observing continuously. It uses two functions: One to
detect the eye blinking, second is for reading the blinking. Automatic driving and braking systems are
also combined with a controlling system using python programming. Speed is automatically reduced
until the driver becomes alert and returns to consciousness. The proposed system alerts the driver
depending on his state, and makes sure that he is not drowsy. However, if the driver has a medical
condition or blinks at an abnormal rate despite not being drowsy, the system will give a false alarm. In
the worst case scenario, the driver happens to be in an accident, the system fails to detect the impact
and contact the concerned authorities. Lastly the system would constantly consume power and drain the
power supply since it monitors the driver continuously. Hence the outcome of not being able to
identifying the actual accident scenario made us reject the idea of adding face recognition to our system
as it would be costly, power- consuming and inefficient.
Rajvardhan Rish, Sofiya Yede, Keshav Kunal, Nutan V Bansode [
2
] proposed a system which states
that the leading cause of deaths in road accidents is due to delay in medical help. This can be prevented
by messaging the authorities and emergency contacts too on time. The system consists of GPS, GSM,
accelerometer and Arduino. It alerts nearest hospital, police headquarters, family and friends during the
time of mishap mainly by detecting changes in accelerometer. The system sends a google map link
using GPS module and Arduino. The vehicle sets the flag bit of the Arduino UNO as an accident is
identified until it detects abrupt deviation from the threshold values with the help of the measuring
system detector. Throughout the accident, the device sets the effective sensitive value for measuring
instrument detectors, unless a crash is observed. Once the accident or set bit is detected by the measuring
instrument detector, Arduino activates the GSM module, which has a manually saved signal of the
accident victim's emergency contact, and sends a pre-stored SMS to that contact.
Though this system works fine, it lacks the detection of rare minor accidents with no casualties. So,
it will eventually result in waste of resources and time in the case of minor accidents. Furthermore, it
uses Arduino UNO which is less powerful than the recent microcontrollers available in the market.
Hence, we decided to only take the system architecture components which would be beneficial to our
project in accuracy which are the following: GSM, GPS module, accelerometer.
Aarya D.S, Athulya C.K, Anas.P, Basil Kuriakose, Jerin Susan Joy, Leena Thomas [
3
] proposed a
system that states the vehicle accidents are one of the most leading causes of fatality. The period
between the occurrence of an accident and the dispatch of emergency medical services to the accident
site is a critical factor in accident survival rates.
Accident detection and messaging system will be stationed in vehicle itself which will be helpful
during the time of accident as hospital, police and emergency contact can be informed immediately.
The system is executed using GPS and GSM technology. A vibration sensor detects a collision using
piezoelectric effect; which is the ability of certain materials to generate an electric charge when they
are under mechanical stress. As soon as the collision is detected the GPS module locates the accident
(latitude and longitude) and sends a message to the hospital and the emergency contact using the GSM
module. The ambulance arrives to the location which is tracked by the GPS module and hence the victim
is treated as soon as possible reducing the help time. In case if there is a minor accident, the victim can
press a switch (button) to prevent the emergency contacts from being alerted. This system comprises of
Arduino, GPS, GSM and vibration sensor, which detects the accident and alerts the authorities
immediately, it also combats false alarms by using a switch provided for the driver. However, the system
does not provide the medical data and history of the victim and hence there could be a delay in the
victim’s treatment. We shall improvise our system in this scope.
Prashant Kapri, Shubham Patane, Arul Shalom [
4
] proposed a system which states that an accident
might occur at an isolated area where humans are absent to report any mishap. Inbuilt hardware modules
in luxury vehicles have recently been developed to detect and report accidents. Unfortunately, such
devices are both costly and immobile.
They proposed a system in which the accidents are detected with the help of in-built sensors in the
smartphone and physical context information. The system comprises of the a server and the software.
The software acts as a sensing device as well as an interface for the third-party observers to contribute
information to the accident report. The software also uses Google maps on the smartphone for mapping
purposes. The map will allow other drivers to plan their route intelligently around an accident, hereby
reducing the congestion around that road via live update on map about accident. The client software
can access the data from phone database such as a contact list to inform emergency contacts. The
software will be installed in smartphone and will perform the task of detecting the accident and also
exchanging information with server. The detection will completely depend on the sensors built in the
phone which includes accelerometer sensor, GPS receiver, and microphone. Although in-vehicle
detection system provides essential information very quickly but unavailability of this system are
restricted by their non- portability and costs, whereas smartphone provides a promising platform with
same sensors at cheap price and portability benefits. The functionality of a smartphone will outperform
that of a traditional in-vehicle accident detection device. Whilst the idea of adding a smartphone was
great but the major drawback of this would be if the battery of the smartphone ran out or if Bluetooth
on the smartphone was disabled during the time of the accident, then this system fails. Hence, we
decided to not implement application software as it could potentially fail in major accidents which are
the main reason for the implementation of this project. Furthermore, if the device itself is damaged in a
major accident, then there would be no detection making this implementation useless.
Bruno Fernandes, Vitor Gomes, Joaquim Ferreira and Arnaldo Oliveira [
5
] proposed system with a
HDy Copilot, an Android application for accident detection integrated with multimodal alert
dissemination. An android mobile is also used so that the rider can use the app, receive road hazard
alerts from other vehicles in the area, and cancel countdown procedures if a false accident is detected.
The software used are IT2S ITS-G5, GPS, 2 radio frequency (RF) modules, a field programmable gate
array, USB. They demonstrated an accident detection system using an Android smartphone, ODB-II
data, and vehicular communications, all of which were integrated with the e-Call platform: When an
accident is identified, a DENM alert is sent to all nearby vehicles, along with SMS and voice calls to
the emergency number. However, the major drawback faced was the same that is damage of
smartphone/loss of signal which renders system useless.
According to Manuel Fogue, Piedad Garrida, and Francisco J. Martinez [
6,
] vehicular networks
would play a larger role in the Intelligent Transportation System sector, reducing road fatalities. The
majority of its uses, such as traffic control, fleet management, and navigation. These would be
dependent on vehicle-to-roadside infrastructure communication, or even vehicle-to-vehicle
communication. The introduction of sensoring technologies on-board vehicles, as well as peer-to-peer
mobile communication between vehicles, are expected to result in major safety improvements in the
near future. Based on the concepts of data mining and information inference, the proposed system
depicts a smart system that will automatically detect road accidents, alert them through vehicular
networks, and estimate their severity. On-Board unit (OBU) sensors, data acquisition unit, processing
unit, and wireless interfaces, V2V and V2I communications, Control Unit (CU), Information Discovery
in Databases are among the libraries and applications used to implement the system. Centered on V2V
and V2I communications, the authors developed and implemented a prototype for automatic accident
and assistance. After a thorough selection of relevant characteristics, the findings revealed that vehicle
speed is a critical factor in front accidents, but in side and rear-end collisions, the type of vehicle
involved and the speed of the struck vehicle are more significant than speed itself. The studied
classification algorithms do not show significant differences, but they do show that if we can identify
accidents based on the types of impacts, we can significantly improve the system's accuracy, particularly
in front crashes where the vehicle is typically the striking one. We disregarded use of KDD algorithm
as it didn’t provide significant difference and was inefficient as well as complex to implement.
Vehicular networks will play an increasing role in the Intelligent Transportation Systems (ITS) area
and hence reduce road fatalities. Most ITS applications, such as road safety, fleet management, and
navigation. These will depend on communication between the vehicle and the roadside infrastructure
(V2I), or even directly between vehicles (V2V). The integration of sensoring capabilities on-board of
vehicles, along with peer-to-peer mobile communication among vehicles, forecast significant
improvements in terms of safety in the near future. Piedad G., Manuel F., Francisco J. M. [
7
] proposed
a framework defines a smart system that automatically detects, notifies, and estimates the magnitude of
road accidents across vehicle networks on the basis of the principle of data mining and information
inference. On-Board Unit (OBU) sensors, the data acquisition unit, the processing unit and wireless
interfaces, V2V and V2I communications, Control Unit (CU), Information Discovery in Databases are
the libraries and applications used to implement the system (KDD). A prototype for automated accident
and assistance based on V2V and V2I communications has been developed and implemented by the
authors. Results showed that vehicle speed is a key factor in front crashes after a careful selection of
relevant characteristics, but the type of vehicle involved and the speed of the hitting vehicle are more
significant in side and rear-end collisions than speed itself. The studied classification algorithms do not
show remarkable differences, but they demonstrate that, if we are able to classify the accidents
depending on the types of impacts, we can noticeably increase the accuracy of the system, especially
for front crashes where the vehicle is usually the striking one. We disregarded use of KDD algorithm
as it didn’t provide significant difference and was inefficient as well as complex to implement.
The late arrival of an ambulance is the leading cause of death in road accidents. If aid comes quickly,
more lives can be saved. In some cases, injuries are minor, but the ambulance's late arrival can result in
casualties. Usman Khalil, Tariq Javid, and Adnan Nasir [
8
] have proposed a system for sending accident
messages to emergency services using the Vehicular AD-Hoc Network (VANET). VANET assists these
services in finding the best route to their destination by using the ABEONA algorithm, a traffic signal
module. This function will significantly reduce the time it takes for an ambulance to arrive at the
accident scene. VANET, eCALL mobile application, and ultrasonic sensor are among the system's
features. Accident detection using VANET is the most effective approach of all because it not only
detects an accident, but also provides the ambulance with the best path to get to the accident site as
quickly as possible. However, the system requires the machinery to be installed in both the vehicle and
the ambulance, which is not always feasible. As a result, we determined that VANET will not be
implemented in our scheme. In addition, if a cell phone is damaged in an accident, the application will
be useless, rendering this solution useless.
OEMs do not instal such detection systems in most low-end cars in developed countries. Based on
data obtained from mobile sensors, this algorithm confirms an accident. This is a less expensive choice.
Smartphone, GPS receiver, accelerometer, magnetometer, and gyroscope are the components used to
implement the device proposed by Harit S., Ravi K.R., Archana K. [
9
]. It uses a collision index and a
crash detection algorithm. A few common crash detection systems are also smartphone-based and use
MEMS sensor data collected in real-time while driving. However, data obtained from smartphone
builtin sensors may contain significant noise, and therefore most current solutions may reduce detection
accuracy and lead to false positives. To address the limitations of using a smartphone as a data capture
source, the proposed solution employs collision index measurement, known false positive identification,
and removal, as well as additional confirmation with position data from the smartphone. However, since
this scheme does not have the victim's medical history, care could be delayed. This solution will not be
able to offer what it promises in incidents where the smartphone is totally destroyed. As a result, a
smartphone sensor-based application might not be able to match the precision of expensive in-vehicle
systems fitted with high-end accident detection technology, as they have an obvious advantage of
providing direct access to the vehicle.
Hossam M. Sherif, M.Amer Shedid, and Samah A. Senbel [
10
] suggested a solution in which three
key components, namely the Node algorithm, Router algorithm, and Coordinator algorithm, are
required to facilitate the accident report. During a road accident, the Real Time Traffic Accident
Detection System (RTTADS) can intelligently notify the accident site through a wireless interface. It
will also notify the appropriate authorities. It will not only tell you how many people have died, but also
what kind of emergency services are needed. RFID sensors, an RF module, a wireless module, a crash
sensor, a rollover sensor, a fire alarm sensor, a weight sensor, and a microcontroller are among the
system's hardware components. The proposed system would detect an accident in real time and send
information to the supervisory programme about the accident site, vehicle speed (before the impact
sensors are triggered), number of passengers in the vehicle, crash sensors that have been activated (front,
back, right side, and left side), rollover sensor status, and fire alarm sensor status. The rule-based system
scans the estimated number of ambulance cars required for injured patients, as well as whether or not a
fire truck is required. We liked the concept of using three separate algorithms to organise the process
of sending a message, but it can be easily implemented using a GPS and GSM module, and false alarms
can be avoided using updated technology like the fingerprint sensor used in our proposed system.
Ms. Sharmila S. Gaikwad [
11
] proposes an agent-based framework for the extremely complex and
variable sense of healthcare emergency decision support. It emphasizes the value of using mobile agents
to help the real-time deployment of an emergency service, rather than just hypotheses. A mobile agent
is an autonomous program capable of transporting itself entirely under its own control between network
nodes, carrying with it the data and execution status required to resume execution at the destination host
from where it left off on the original host. Hence this agent makes a decision on when and where to
move and how to go about the execution without consulting the user repeatedly. Along with mobility,
they are also capable of performing dynamic and intelligent inference tasks during their execution which
makes them a great approach not only in situations of urgent medical help but also in other areas like
military to educational.
Md. Syedul Amin, Jubayer Jalil, M. B. I. Reaz [
12
] have stated that speed is one of the most
important and basic risk factors in driving. Not only does it affect the severity of a crash, but it also
raises the likelihood of a crash. If emergency service could get accident reports and reach it in time,
more lives could have been saved. The GPS tracks the speed of a vehicle and, through a microcontroller
chip, compares it with the previous speed every second. Whenever the speed is below the stated speed,
an accident is believed to have occurred. Through using the GSM network, the device will then submit
the accident position obtained from the GPS along with the time and speed. They propose to use the
ability of a GPS receiver to track a vehicle's speed and detect an accident based on the speed monitored
and send the position and time of the accident from the GPS data processed by a microcontroller to the
Warning Service Center using the GSM network. Kinetic energy is converted into destructive forces
that cause damage to the passengers as well as to the vehicle when an accident occurs. The system uses
GPS, GSM and a microcontroller unit for hardware and the software components are accident detection
algorithm, speed measurement, detection and reporting by MCU, and finally the data interpretation unit.
It will also show the previous speed of the vehicle before committing the accident. This data will help
the Alert Service Center to assess the severity of the accident basing on the speed and also it can initiate
a voice call. The detailed algorithm sure reaps a good result but it is time consuming and hence, only
necessary steps can be initiated keeping speed in mind.
Nagarjuna R Vatti, Prasanna Lakshmi Vatti, Rambabu Vatti, Chandrashekhar Garde [
13
] have
provided some shocking statistics of accident rates in India and also talked about the reasons of increase
in road accidents that is caused by improper construction and low maintenance of the roads as well as
overcrowding and increasing count of vehicles. Other than this, the youth succumb to their injuries on
roads because of rash driving, drunken driving. The pre-existing types of accident prevention systems
installed in cars are air bags, ABS. The projected system relies on the thought that the accident is
detected by vibration and rotating mechanism sensors once a close review of current systems and
literature surveys, and a message is straight away sent to the emergency contact numbers victimization
the GSM module beside the situation found by the GPS module. If the vehicle gets any head-on collision
the vibrations area unit created. If the vibrations exceed a threshold price they're detected and treated as
a heavy accident case. The gyro detector can notice if the vehicle has toppled or atilt by an outsized
angle. All told cases, the system can anticipate ten seconds. If button is ironed by the motive force
among ten seconds the system considers that accident isn't serious and it resets back to traditional
operation. During this system, the guts rate detector can notice the speed of the motive force only if the
accident has occurred. By causation this data, the hospitals can get to grasp concerning the condition of
the motive force and consequently they'll react to assist the motive force. They need to conduct
experiments by implementing the system in a very toy automobile and determine that the system is
functioning properly. This is by far the best and most efficient as well as cost friendly approach.
T Kalyani, S Monika, B Naresh, Mahendra Vucha[
14
] have proposed three main components in this
system namely vibration sensors, GPS and GSM module. When a vehicle meets with an accident, the
vibration sensor will read the impact and Arduino will then compare it with the threshold value set in
the program. If the value exceeds the threshold value then GPS will generate the current location and
GSM will send the alert message to respective authorities with the help of Arduino. Vibration sensor is
used to detect the accident. One of the major gaps with this system is that it doesn’t provide any solution
to a minor accident situation where there isn’t any serious damage or any casualties. So in such cases if
the system doesn’t provide a solution to this problem this will ultimately result in waste of resources
and time. Also this system doesn’t provide any medical history of the victim which eventually results
in the delay which will ultimately hinder the cause.
Bharath P, Saravanan M, Aravindhan K [
15
] have performed a detailed literature survey of 19 papers
reflecting ideas on accident detection using machine learning. Machine learning is an artificial
intelligence (AI) technology that gives systems the ability to learn and develop from practice
automatically without being programmed specifically. Machine learning focuses on the development
of a computer program that can access and use knowledge to learn for itself. The paper that stands out
among them is the one where the dissertation system anticipates and predicts the destination location.
Using a Damped area simulation, it tests four progressive scholarship approaches, namely, variable
compound regression, spherical-spherical regression, irregular circular K-NN dropping off, and a
gaggle of those techniques for his or her effectiveness in determining the matter of destination
prediction.
Wen-Kai Tai, Hao-Cheng Wang, Cheng-Yu Chiang, Chin-Yueh Chien, Kevin Lai, and
TsengChang Huang [
16
] have stated that, traffic accidents are happening every day, still there is not an
efficient way to investigate, record and measure such parameters with precision. The authors have come
up with a system to extend the exactness and viability of knowledge retrieved in traffic accident
investigation, shorten the investigation time, ensure the safety of investigators, decrease social value,
and improve the general service quality of traffic accident investigations. The planned system contains
four phases: prepared, Record, Measure, Archive, and Edit phases. The intent of this system is merging
the practicality of drones associated mobile devices to form a system that may collect and record proof
each fleetly and expeditiously. If the investigator must sketch the accident diagram, he will use the
functions of Accident Diagram App (ADA). This technique provides correct and correct readings, with
the assistance of overhead pictures enamored by the assistance of drones; the gap measured is correct
with average error of +-5cm.This paper provides an excellent system which could be able to detect and
investigate traffic accidents happening in the city, and help save time. It uses drones to investigate the
accident spot which collects overhead photos and provides a function ADA to do so efficiently.
However, if the accident is very fatal, this system would not inform any hospital or such designated
authorities.
Barış Guksa and Burcu Erkmen [
17
] have proposed a system which would not only inform about
the accident, but also detect if the driver is feeling sleepy and is not in a condition to drive, hence
decreasing the possibility of an accident. To observe the driver's standing, the projected device uses a
smartphone, which can do this with its front camera. The user is asked to require a photograph once the
program is opened. Within the photograph, the situation of the face is set. Once the situation of the face
is set, solely the face is cut and it's saved as a replacement photograph by the system. Then, the quantity
of gap and shutting eyes in a very minute is saved within the system. If additional or less the quantity
of eye gap and shutting in a very minute, that is completely different from the conventional, is detected,
the motive force controls himself by causing him a voiced warning message speech “are you sleepy?”
is aimed. The smartphone ought to have a front camera; associate ALS (Ambient light-weight Sensor),
a gyro detector associated with a measuring system and a recipient IC (Integrated Circuit), which is
required for GPS signals to be processed within the phone. The system proposed by the author could be
efficient to prevent accidents from happening. As this system uses smartphone’s sensors for detection,
it is mandatory that the driver has a smartphone, since smartphones are very common these days, they
could be useful in such a system, but if the driver forgets to bring his smartphone, nonetheless does not
carry a smartphone, this could render the whole system inefficient.
Adnan Bin Faiz, Ahmed Imteaj, Mahfuzulhoq Chowdhury [
18
] have stated that, these days,
smartphones have become an everyday requirement and this paper therefore proposes a device that uses
a smartphone to detect whether an accident using the sensors of the smartphone has occurred. The
proposed system uses the GPS receiver of the phone to detect a sudden rapid change in deceleration
that happens at the time of an accident. It also reads the change in pressure from the pressure sensor and
the angle of tilt from an accelerometer sensor in a Smartphone. After detecting these three conditions,
the android app would send an alert to the emergency contact. To prevent any false alarm, the system
is provided with a switch which, if pressed, would prevent sending an alert to the emergency contact.
This could be useful in case of wrong readings in the sensors, or if the accident is a minor one and not
fatal. This device would require a constant internet connection in order to work smoothly. The system
used in this paper can give systematic results and ensure the safety of the driver using smartphones
which have become a daily necessity in today’s life. Since the smartphone’s sensors could provide false
data sometimes, this system also comes with a switch, which could be used by the driver in case of
wrong readings. Considering this system requires a constant internet connection, which may not be
possible in remote areas, this may be a drawback in its working. In such a situation, the system could
fail.
Miss. Priyanka M. Sankpal, Prof. P. P. More [
19
] have implied that this system aims at preventing
accidents by detecting the driver’s state i.e., if the driver is feeling drowsy or is not in the condition to
drive. IR reflective obstacle sensors are used to detect the position of the eyes. If eyes are opened then
the sensor gives the low signal to the controller, and if the eyes are closed till 4 sec then the sensor gives
high signal to the controller. The PIC controller receives input from the sensor, and the controller gets
the high or low signal and processes it to make the buzzer ON or OFF, depending on the buzzer the
LCD displays “Driver Slept”, if the buzzer is ON. Hence the driver is notified and a major mishap could
be prevented from happening. Since drowsiness is a major cause of accidents, this system provides a
very efficient way to reduce such incidents. Even if the system provides a way to prevent accidents
from happening due to drowsiness, there could be other causes which may not be detected and since the
system does not provide detecting the damage and notifying the emergency contacts, this could be a
major drawback.
Dnyanesh Dalvi, Vinit Agrawal, Sagar Bansod, Apurv Jadhav, Prof. Minal Shahakar [
20
] have
suggested a system which uses an android application to automatically detect and report an accident
using a smartphone. An Android application was built and connected to a USB port in order to include
an accident detection mechanism and eCall implementation. A GPS receiver, two Radio Frequency
modules, a Field Programmable Gate Array for lower MAC, baseband PHY layer use and Universal
Serial Bus connections are the highlighted features of the device that are helpful to the project. Due to
its hardware resources and software capabilities, a smartphone has been chosen as an Application Unit
(AU), even because it has a linear accelerometer of 3 axes. For eCall implementation, the GSM/GPRS
capabilities are very useful.
The android application is divided into two modules: Design- Related to the elements of the
Graphical User Interface (GUI), such as the added icons, colors, layouts, pictures and visual effects,
and Core- Mostly related to the functionality of the application. It enables the GUI components, as well
as Autonomous Accident Detection (aAD) and eCall processing, to perform the tasks needed when
requested. The center of the eCall system is the Accident Detection Algorithm, which focuses on
providing the application to automatically detect vehicle accidents. This system also offers a warning
notification which the driver can press if there are any false alarms or if the accident is not fatal. Overall
this system could work adeptly with the help of a smartphone which would contain the android
application. But since it uses a smartphone, there comes the issue of network connectivity, which could
be a problem in remote areas.
P. Kaliuga Lakshmi, C.Thangamani [21] have presented a system which detects an accident using
MEMS sensor which would detect the car roll over and give it as an input to the controller for
processing. When a rollover is detected, the system would detect the location using a GPS module and
send a message to the medical authorities using the GSM module. Using Vehicle to Vehicle (V2V) or
Vehicle to Road (V2R) communication, vehicles may communicate with each other. It could have
certain applications like security distance warning, vehicle collision warning, map location, driverless
vehicle etc... Most applications include the measurement of traffic speed and travel time. This
calculation helps roadway users to identify which routes to use. This calculation can be saved to analyze
traffic speed and travel time for different time intervals. This paper suggests an approach which could
be efficient and easy to implement, but it does not provide a switch or any kind of system to prevent a
false alarm.