IRJET - Accident Intimation System using Image Processing

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 03 | Mar 2020 www.irjet.net p-ISSN: 2395-0072
© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 1127
Accident Intimation System Using Image Processing
Promod Vishnu R1, Sathish Kumar A1, Vignesh M1, Gopinathan S2
1Computer Science and Engineering, Agni College of Technology
2Assistant Professor, Computer Science and Engineering Department, Agni College of Technology
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The road accident is the major cause of
unnatural death in the world. This is due to increase in the
population and number of vehicles the accident rate has been
increased. The major causes of death is due todelayinmedical
attention, this is because of delay in intimation. The proposed
system acts as the detecting the accident has occurred and
intimating to nearest hospital. The convolutional neural
network algorithm is used for detecting the accident has
occured and the simple mail transfer protocol is used for
communicating to nearest hospital. By this there will be no
delay in intimating the accident if occurred, the rate of death
will also be decrease.
Key Words: CNN, SMTP SERVER.
1. INTRODUCTION
The accident intimation system will analyzing the
accident detection through image based traffic
surveillance camera. This has always been a
challenging task because detecting the accident
accurately is not easy task for implementation. Hence
we need a system that can maximize the number of
frames per second and also achieving the acceptable
performance for the detection purpose.
Theimportantstageisvehiclecrashingmonitorsystem
that is useful for detecting the video by each frames
and also accuratelytrackingallthevehiclesacrosseach
frame. Tracking, vehicle detection, and change in
orientation can be determined the process of crash
detection. The tracking can be viewed as
correspondence problem in which the goal is to
determine that the vehicle detectinthenextframeisto
be given in the current frame. This will be calculating
the each and every frame and the detection will be
taking place. The task of tracking the is performed by
the system but it is quite challenging because it should
detect each and every frame in the video.
The convolutional Neural Network will be play the
important role in this detection process the working
process of the CNN algorithm are Convolution, Apply
the ReLu (Rectified Linear Unit), Pooling, Flattening,
Full connection, Softmax. These process will the taken
place in the CNN algorithm.
The three main items that are used in the process are
input image , feature detector and feature map .The
input image is the image for thedetectionpurpose.The
feature detector is a matrix format that will usually be
in the form of 3x3 and in some areas it could also be in
the form of 7x7 matrix. A feature detector is a nervous
system used for extracting the related behaviour .The
feature map will be detecting the location. Apply
ReLu(Rectified Linear Unit) is to increase the non-
linearity in the process.
The pooling function is to reduce the spatial size to
reduce the amount of computation in the network
.Each feature mapping will be operated by the pooling
layers in independent manner some types in pooling
are max pooling ,average pooling etc .Max pooling will
be selecting the maximum amount of elements from
region that has been covered . Average pooling is the
process of average number of elements will be taken
and will be computed.
Flattering, Full connection, Softmax all these three
processareclassificationprocessintheCNNalgorithm.
These are the various processes that are taken place in
the algorithm. The second main module is the SMTP
server which is a communication protocol used for
transmitting the electronic mail. This SMTP is used
here for intimating that the accident as occured by a
message.

II. DATASETS
The images for the vehicles are been collected. Each
and every image of the vehicleshouldbedifferentfrom
another , because only then they will be able to detect
all the vehicles. For this not only images of a particular
type of transport are been collected. For detecting the
accident also the data set are been collected only by
these set of image they can be identified that the
accident has been taken place or not. All these data set
will be trained in RCNN methodology for the accident
detection.
Fig1:
III. PROPOSED SYSTEM
The accident intimation system consist of two main:
modules accident detection and intimationofaccident.
In the detection part the dataset will the playing a
major role for recognizing the accident has occurred,
the set of images that have been trained in the RCNN
will be detectingthattheaccidenthasoccurred.Theset
of accident images will be given if any frame in the
video has the identical image set then it will be
identifying as accident has occurred. The second
module, accident intimation will be processed the
SMTP server. If the system has confirmed that the
accident has occurred then the SMTP will be sending a
mail to the nearest hospital
Fig2: BLOCK DIAGRAM OF THE PROPOSED SYSTEM
IV. MODULES
a. Input Image
b. Feature Extraction
c. Vehicle Detection
d. Accident Detection
e. SMTP Server
g. Output

a. Input Image
The input video will be reading by each and every
frame. Thus by this the input dataset and if any frame
in the video is been matched then the accident will be
detected. The input video will run using matlab.
Fig:3
b. Feature Extraction:
If the input data is too large in size then the feature
extraction will be helpful .For the convolutional neural
network the SURF (Speeded Up Robust Feature)
algorithm is used for extraction.
c. Vehicle detection
The vehicle detection is been detected by the trained
data set .The annotation part, reading the video will be
processed by matlab .Thus if it identified if any vehicle
passes through it, then it will be automatically check
the trained data set that is the process is been match. If
they are similar to each other than detected object is
said to be a vehicle.
Fig4:
d. Accident Detection
The accident detection will be performed by the
detecting each and every frame in the video .If will be
detecting all the vehicle in each and every frame .If the
accident has occurred the it will be reading the trained
data set, if they are similar then the accident will be
detected
Fig:5
e. SMPT Server
The SMTP server will be acting as a indicator to the
hospital if accident has occured .This SMTP server will
be activated only if anyaccidenthasbeendetected.The
Email id of the hospital will be initialized in the code.
Thus the process of intimationwillbeintimatedonlyto
that particular mail id.
g .Output
As input is taken in form of frames from the video. The
image frames andthetraineddatasetwillbecompared
and the output will be detected.
Fig:6

Fig:7
V. CONCLUSION
Thus by detecting the accident and intimating it to the
hospital will reduce the number of unnatural death.
This can be proposed easily in surveillance camera, so
this system can be placed all over the city. Thus it will
be helping the people who met with the accident for
implementation purpose.
VI. RESULT
In this paper, we have proposed the accident detecting
using CNN algorithm developed in matlab, and
intimating that the accident has occurred by using
SMTP Server.
VII. REFERENCES
[1] Vaishnavi Ravindran , Lavanya Viswanathan,Dr.
Shanta Rangaswamy “A Novel Approach to Automatic
Road-Accident Detection using Machine Vision
Techniques” (IJACSA) International Journal of
Advanced Computer Science and Applications, Vol. 7,
No. 11, 2016
[2] Department of Transportation National Highway
Traffic Safety Administration, Traffic safetyfacts2011.
[Online]. Available: http://www-nrd.
nhtsa.dot.gov
[3] Z. Sun, G. Bebis, and R. Miller, “On-road vehicle
detection: A review,” IEEE Trans. Pattern Anal. Mach.
Intell., vol. 28, no. 5, pp. 694–711,
May 2006.
[4] A. Barth and U. Franke, “Tracking oncoming and
turning vehicles at intersections,” in Proc. 13th Int.
IEEE ITSC, Sep. 2010, pp. 861–868.
[5] S. Sivaraman, B. T. Morris, and M. M. Trivedi,
“Learning multi-lane trajectories using vehicle-based
vision,” in Proc. IEEE Int. Conf. Comput. Vision
Workshop, 2011, pp. 2070–2076.
[6] D. Kasper, G. Weidl, T. Dang, G. Breuel, A. Tamke,
and W. Rosenstiel,“Object-orientedBayesiannetworks
for detection of lane change maneuvers,” in Proc. IEEE
IV, Jun. 2011, pp. 673–678.
[7] X.Mao, D. Inoue, S. Kato, and M. Kagami,
“Amplitude-modulated laserradarforrangeandspeed
measurement in car applications,” IEEE Trans. Intell.
Transp. Syst., vol. 13, no. 1, pp. 408–413, Mar. 2012.
[8] S. Tokoro, K. Kuroda, A. Kawakubo, K. Fujita, and H.
Fujinami, “Electronically scanned millimeter-wave
radar for pre-crash safety and adaptive cruise control
system,” in Proc. IEEE Intell. Veh. Symp., Jun. 2003,
pp. 304–309.
[9] J. Levinson, J. Askeland, J. Becker, J. Dolson, D. Held,
S. Kammel, J. Kolter, D. Langer, O. Pink, V. Pratt, M.
Sokolsky, G. Stanek, D. Stavens, A. Teichman, M.
Werling, and S. Thrun, “Towards fully autonomous
driving: Systems and algorithms,” in Proc. IEEE IV, Jun.
2011, pp. 163–168.
[10] S. Sato, M. Hashimoto, M. Takita, K. Takagi, and T.
Ogawa, “Multilayer lidar-based pedestrian tracking in
urban environments,” in Proc. IEEE IV, Jun. 2010, pp.
849–854.

IRJET - Accident Intimation System using Image Processing

More Related Content

What's hot (18)

Similar to IRJET - Accident Intimation System using Image Processing (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET - Accident Intimation System using Image Processing