ACCIDENT DETECTION USING MACHINE LEARNING METHOD

at

Sathyabama Institute Of Science and Technology

Submitted in partial fulfillment of the requirements for the award

of bachelor of engineering degree in computer science and
engineering

MANUKONDA TEJASREE - 38110307

MUCHUMARRI SAI JAHNAVI -38110330

DEPARTMENT OF COMPUTER SCIENCE AND

ENGINEERING
SCHOOL OF COMPUTING

SATHYABAMA INSTITUTE OF SCIENCE AND TECHNOLOGY

JEEPIAAR NAGAR ,RAJIV GANDHI SALAI,
CHENNAI-600119, TAMILNADU
MAY - 2022
 SATHYABAMA
INSTITUTE OF SCIENCE AND TECHNOLOGY
(DEEMED TO BE UNIVERSITY)
Accredited with “A” grade by NAAC
Jeppiaar Nagar, Rajiv Gandhi Salai, Chennai - 600119
www.sathyabama.ac.in

BONAFIDE CERTIFICATE

Certified that this project report titled “ACCIDENT DETECTION USING MACHINE
LEARNING METHODS” is the bonafide work of MANUKONDA TEJASREE , MUCHUMARRI
SAI JAHNAVI who carried out the project work under my supervision during the academic
year 2021-2022.

Internal Guide
Dr. A. Viji Amutha Mary, M.Tech., Ph.D

Head of the department

Dr. S. Vigneshwari M.E., Ph.D., and Dr. L. Lakshmanan M.E., Ph.D.,

Submitted for viva voce Examination held on :

Internal Examiner External Examiner

 DECLARATION

I, MANUKONDA TEJASREE , MUCHUMARRI SAI JAHNAVI hearby declare that the project
report entitled ACCIDENT DETECTION USING MACHINE LEARNING METHODS done by us
under the guidance of Dr. A. Viji Amutha Mary, M.Tech., Ph.Df is submitted in partial fulfillment of
the requirements for the award of Bachelor Of Engineering Degree in Computer Science and
Engineering.

DATE : 4/04/2022 M.Tejasree, M.Sai Jahnavi

PLACE : CHENNAI. Signature of candidate

2
 ABSTRACT

There are lots of studies about preventing or detecting the car accidents. Most of them includes sensing
objects which might cause accident or statistics about accidents. In this study, a system which detects
happening accidents will be studied. The system will collect necessary information from neighbor
vehicles and process that information using machine learning tools to detect possible accidents. Machine
learning algorithms have shown success on distinguishing abnormal behaviors than normal behaviors.
This study aims to analyze traffic behavior and consider vehicles which move different than current
traffic behavior as a possible accident. Results showed that clustering algorithms can successfully detect
accidents.The problem of deaths and injuries as a result of accidents is to be a global phenomenon.
Traffic safety has been a serious concern since the start of the automobile age, almost one hundred years
ago.It has been estimated that over 300,000 persons die and 10 to 15 million persons are injured every
year in road accidents throughout the world.Statistics have also shown that mortality in road accidents is
very high among young adults that constitute the major part of the work force. In order to overcome this
problem, there is need of various has road safety strategies and measure.Losses in road accidents are
unbearable, to the society as well as a developing country like us. So, it has become an essential
requirement to control and arrange traffic with an advanced system to decrease the number of road
accidents in our country. By taking simple precautions, based on prediction of a sophisticated system may
prevent traffic accidents. Moreover, to tackle this situation where every day so many people were killed
in a traffic accident. and day by day this rate is getting increased. The implementation of machine
learning is a functional and a great approach to take an accurate decision with the experience to manage
the current situation and the findings of the analysis part which we are suggested to traffic authorities.

3
 ACKNOWLEDGEMENT

I am pleased to acknowledge my sincere thanks to Board of Management of

SATHYABAMA for their kind encouragement in doing this project and for completing it
successfully. I am grateful to them.

I convey my thanks to Dr. T. Sasikala M.E., Ph.D, Dean, School of Computing, Dr. S.
Vigneshwari, M.E., Ph.D. and Dr. L. Lakshmanan, M.E., Ph.D., Heads of the
Department of Computer Science and Engineering for providing me necessary support
and details at the right time during the progressive reviews.

I would like to express my sincere and deep sense of gratitude to my Project Guide Dr. A.
Viji Amutha Mary, M.Tech., Ph.D for his valuable guidance, suggestions and constant
encouragement paved way for the successful completion of my project work.

I wish to express my thanks to all Teaching and Non-teaching staff members of the
Department of Computer Science and Engineering who were helpful in many ways for
the completion of the project.

4
 TABLE OF CONTENTS
PAGE
CHAPTER NO TITLE
NO
ABSTRACT 3
LIST OF FIGURES 7
LIST OF TABLES 6
1 INTRODUCTION 8
1.1 Technologies used 9
1.1.1 Machine Learning 9
1.2 Importance of Machine learning 10
1.3 Machine learning process 10
1.4 Types.of Machine Learning 12
1.5 Supervised Machine Learning 12
1.6 Unsupervised Machine Learning 13

1.7 Reinforcement Machine Learning 13

2 LITERATURE REVIEW 15
2.1 Incident detection algorithm 15
2.2 Design of a system solution for relative
16
positioning
2.3 Vehicle - to - Vehicle safety messaging in
17
DSRC
2.4 Integrating dead reckoning,map matching and
18
GPS positioning
2.5 Challange for localisation-based Systems 19
3 PROJECT DESCRIPTION 20

5
 3.1 Existing system 20
3.2 Proposed system 20
3.3 Working 21
3.4 Module Description 21

3.4.1 Data cleaning and Data Preprocessing 22

3.4.2 Data cleaning 23

3.4.3 Data visualization 23
4 METHODOLOGY FLOW 24
24
4.1 System architecture
4.2 Input images 24
4.2.1 Accident dataset creation

4.2.2 Working
4.3 Disadvantages of Existing System
25

4.4 Advantages of Proposed System 26

4.5 Applications 26

5 TOOLS AND TECHNIQUES 27

5.1 Python 28
5.2 Pycharm 29
5.3 Software code 39

6 RESULTS 40

6
 6.1 Final results 43
7 CONCLUSION 44
7.1 Conclusion 44
7.2 Future work 44
7.3 Scope of further study 44
8 REFERENCES 46

7
 LIST OF FIGURES

FIGURE NO TITLE PAGE NO

1.1 Machine Learning process 10

1.2 Types of machine learning 12

4.1 System architecture 24

5.1 Software code and deployment 29

6.1 Accident occurred 40

6.2 Accident will happen 41

6.3 Accident occurred 42

6.4 No of accident occurred 43

8
 CHAPTER 1
INTRODUCTION

The problem of deaths and injuries as a result of accidents is to be a global

phenomenon. Traffic safety has been a serious concern since the start of the
automobile age, almost one hundred years ago.It has been estimated that over
300,000 persons die and 10 to 15 million persons are injured every year in road
accidents throughout the world.Statistics have also shown that mortality in road
accidents is very high among young adults that constitute the major part of the
work force. In order to overcome this problem, there is need of various has road
safety strategies and measure.Losses in road accidents are unbearable, to the
society as well as a developing country like us. So, it has become an essential
requirement to control and arrange traffic with an advanced system to decrease
the number of road accidents in our country. By taking simple precautions, based
on prediction of a sophisticated system may prevent traffic accidents. Moreover, to
tackle this situation where every day so many people were killed in a traffic
accident. and day by day this rate is getting increased.

1.1 TECHNOLOGIES USED:

To implement this project we have used Machine learning

1.1.1 Machine Learning

Machine Learning (ML) is coming into its own, with a growing recognition that
ML can play a key role in a wide range of critical applications, such as data
mining, natural language processing, image recognition, and expert systems.

9
" A computer program is said to learn from experience E with respect to some task
T and some performance measure P, if its performance on T, as measured by P,
improves with experience E.”

1.2 Importance of Machine Learning

Machine learning is important because it gives enterprises a view of trends

in customer behaviour and business operational patterns, as well as supports
the development of new products. Many of today's leading companies, such
as Facebook, Google and Uber, make machine learning a central part of their
operations.

1.3 Machine Learning Process

Machine learning workflow refers to the series of stages or steps involved in the
process of building a successful machine learning system.
The various stages involved in the machine learning workflow are-

Figure 1.1 Machine Learning Process

10
 Figure 1.2 Machine Learning Process
Data Collection- In this stage,

Data is collected from different sources.The type of data collected depends upon the type
of desired project.Data may be collected from various sources such as files, databases
etc.The quality and quantity of gathered data directly affects the accuracy of the desired
system.

Data Preparation- In this stage,

Data preparation is done to clean the raw data.Data collected from the real world is
transformed to a clean dataset.
Raw data may contain missing values, inconsistent values, duplicate instances etc.So, raw
data cannot be directly used for building a model.
Choosing Learning Algorithm- In this stage,
The best performing learning algorithm is researched.It depends upon the type of problem
that needs to solved and the type of data we have.If the problem is to classify and the data
is labeled, classification algorithms are used.
If the problem is to perform a regression task and the data is labeled, regression algorithms
are used.
If the problem is to create clusters and the data is unlabeled, clustering algorithms are
used.
Training Model- In this stage,
The model is trained to improve its ability.The dataset is divided into training dataset and
testing dataset.
The training and testing split is order of 80/20 or 70/30.It also depends upon the size of the
dataset.Training dataset is used for training purpose.Testing dataset is used for the testing
purpose.

11
1.4Types of Machine Learning

There are 3 types of Machine Learning

1. Supervised Machine Learning
2. Unsupervised Machine Learning
3. Reinforcement Machine Learning

Figure 1.2 Types of Machine Learning

1.5 Supervised Machine Learning :

Supervised machine learning is based on supervision. It means in the supervised
learning technique, we train the machines using the "labelled" dataset, and based on
the training, the machine predicts the output. Here, the labelled data specifies that
some of the inputs are already mapped to the output. More preciously, we can say;
first, we train the machine with the input and corresponding output, and then we ask
the machine to predict the output using the test dataset.
Supervised machine learning can be classified into two types of problems, which are
given below:
a) Classification : Classification algorithms are used to solve the classification
problems in which the output variable is categorical, such as "Yes" or No,
Male or Female, Red or Blue, etc

12
 b) Regression : Regression algorithms are used to solve regression problems in
which there is a linear relationship between input and output variables. These
are used to predict continuous output variables, such as market trends,
weather prediction, etc.

1.6 Unsupervised Machine Learning :

In unsupervised learning, the models are trained with the data that is neither
classified nor labelled, and the model acts on that data without any supervision.
The main aim of the unsupervised learning algorithm is to group or categories the
unsorted dataset according to the similarities, patterns, and differences.
Unsupervised Learning can be further classified into two types, which are given
below:
a) Clustering : The clustering technique is used when we want to find the
inherent groups from the data. It is a way to group the objects into a cluster
such that the objects with the most similarities remain in one group and have
fewer or no similarities with the objects of other groups.
b) Association : Association rule learning is an unsupervised learning technique,
which finds interesting relations among variables within a large dataset. The
main aim of this learning algorithm is to find the dependency of one data item
on another data item and map those variables accordingly so that it can
generate maximum profit.

1.7 Reinforcement Machine Learning :

Reinforcement learning works on a feedback-based process, in which an AI agent
(A software component) automatically explore its surrounding by hitting & trail,

13
taking action, learning from experiences, and improving its performance.
Reinforcement learning is categorized mainly into two types of methods/algorithms:

a) Positive Reinforcement Learning: Positive reinforcement learning specifies

increasing the tendency that the required behaviour would occur again by
adding something. It enhances the strength of the behaviour of the agent and
positively impacts it.
b) Negative Reinforcement Learning: Negative reinforcement learning works
exactly opposite to the positive RL. It increases the tendency that the specific
behaviour would occur again by avoiding the negative condition

14
 CHAPTER -2 LITERATURE

REVIEW

2 LITERATURE SURVEY:

2.1 Angshuman, G., 2004. An Incident Detection Algorithm Based On

a Discrete State Propagation Model of Traffic Flow.
Proposed work
Incidents are defined as random and nonrecurring events such as vehicular
crashes,
disabled vehicles, spilled loads, temporary maintenance and construction
activities, and other unusual events that disrupt the normal flow of traffic.
Incident-related congestion is a common occurrence on busy roadways.
The number of incidents per million-vehicle-miles has been reported to be
between 20 and 200 and lane-blocking incidents lasting more than 45
minutes per 100 million-vehicle-miles have been reported to be 1.09. It has
been estimated that 52 to 58 percent of the traffic congestion in urban
areas is due to incidents, amounting to 2 billion vehicle-hours of delay and
a cost of $40 billion in terms of hours of delay and excess fuel consumption
in 2001, as reported in the 2003 Annual emergency services, lost time and
reduction in productivity, increased cost of goods and services, reduced air
quality, increased vehicle maintenance costs and reduced quality of life.

15
 2.2 : Kukshya, V., Krishnan, H., Kellum, C., n.d. Design of a system
solution for relative positioning of vehicles using vehicle-to-vehicle radio
communications during GPS outages.

Proposed work
Active safety applications for vehicles have been at the forefront of the
automotive safety community for a number of years. Cooperative collision
warning based on vehicle-to-vehicle radio communications and GPS
systems is one such promising active safety application that has evoked
considerable interest among automobile manufacturers and research
communities worldwide. In this paper, we address one of the key functional
components of the cooperative collision warning application, which is,
accurate estimation of relative positions of all the neighboring vehicles based
on real-time exchange of their individual GPS position coordinates, and then
propose a novel system solution for achieving the same (relative positioning
functionality) during persistent GPS outages. Based on survey results, we
also qualitatively assess various radio based ranging and relative positioning
techniques, experimentally evaluate the received signal strength based
ranging technique, and comment on their suitability for our proposed
solution.

16
2.3 Xu, Q., Mak, T., Ko, J., Sengupta, R., 2004. Vehicle-to-vehicle
safety messaging in DSRC. AC

The design of layer-2protocols for a vehicle to send safety messages

to other vehicles. The target is to send vehicle safety messages with
high reliability and low delay. The communication is one-to-many,
local, and geo-significant. The vehicular communication network is
ad-hoc, highly mobile, and with large numbers of contending nodes.
We design several random access protocols for medium access
control. The protocols fit in the DSRC multi-channel
architecture.Analytical bounds on performance of the addressed
protocols are derived. Simulations are conducted to assess the
reception reliability and channel usage of the protocols. The
sensitivity of the protocol performance is evaluated under various
offered traffic and vehicular traffic flows. The results show our
approach is feasible for vehicle safety messages in DSRC.

2.4 Krakiwsky, E.J., Harris, C.B., Wong, R.V.., 1988. A Kalman filter for
integrating dead reckoning, map matching and GPS positioning, in: ,
IEEE Position Location and Navigation Symposium, 1988. Record.
Navigation into the 21st Century. IEEE PLANS ’88. Presented at the ,
IEEE Position Location and Navigation Symposium, 1988. Record.
Navigation into the 21st Century. IEEE PLANS ’88, IEEE, pp. 39-46.
To make driving easier and safer, modern vehicles are equipped with driver
support systems. Some of these systems, for example navigation or

17
curvature warning systems, need the global position of the vehicle. To
determine this position, the Global Positioning System (GPS) or a Dead
Reckoning (DR) system can be used. However, these systems have often
certain drawbacks. For example, DR systems suffer from error growth with
time and GPS signal masking can occur. By integrating the DR position
and the GPS position, the complementary characteristics of these two
systems can be used advantageously. In this thesis, low cost in-vehicle
sensors (gyroscope and speedometer) are used to perform DR and the
GPS receiver used has a low update frequency. The two systems are
integrated with an extended Kalman filter in order to estimate a position.
The evaluation of the implemented positioning algorithm shows that the
system is able to give an estimated position in the horizontal plane with a
relatively high update frequency and with the accuracy of the GPS receiver
used. Furthermore, it is shown that the system can handle GPS signal
masking for a period of time. In order to increase the performance of a
positioning system, map matching can be added. The idea with map
matching is to compare the estimated trajectory of a vehicle with roads
stored in a map data base, and the best match is chosen as the position of
the vehicle. In this thesis, a simple off-line map matching algorithm is
implemented and added to the positioning system. The evaluation shows
that the algorithm is able to distinguish roads with different direction of
travel from each other and handle off-road driving.

18
2.5 Boukerche, A., Oliveira, H., Nakamura, E., Loureiro, A., 2008.
Vehicular Ad Hoc Networks: A New Challenge for Localization-Based
Systems. Computer Communications 31, 2838–2849.

The new kind of ad hoc network is hitting the streets: Vehicular Ad Hoc
Networks. (VANets). In these networks, vehicles communicate with each
other and possibly with a roadside infrastructure to provide a long list of
applications varying from transit safety to driver assistance and Internet
access. In these networks, knowledge of the real-time position of nodes is
an assumption made by most protocols, algorithms, and applications. This
is a very reasonable assumption, since GPS receivers can be installed
easily in vehicles, a number of which already comes with this technology.
But as VANets advance into critical areas and become more dependent on
localization systems, GPS is starting to show some undesired problems
such as not always being available or not being robust enough for some
applications. For this reason, a number of other localization techniques
such as Dead Reckoning, Cellular Localization, and Image/Video
Localization has been used in VANets to overcome GPS limitations. A
common procedure in all these cases is to use Data Fusion techniques to
compute the accurate position of vehicles, creating a new paradigm for
localization in which several known localization techniques are combined
into a single solution that is more robust and precise than the individual
approaches. In this paper, we further discuss this subject by studying and
analyzing the localization requirements of the main VANet applications.

19
 CHAPTER - 3 PROJECT DESCRIPTION

3.1 EXISTING SYSTEM:

In existing method Data Mining techniques are used to identify the locations
where high frequency accidents are occurred and analyze them to identify the
factors that have an effect on road accidents at that locations. The first task is to
divide the accident location into k groups using the k-means clustering algorithm
based on road accident frequency counts. Then, association rule mining algorithm
applied in order to find out the relationship between distinct attributes which are in
accident data set and according to that know the characteristics of locations.

3.2 PROPOSED SYSTEM:

Now in this method classification techniques will be using for

identifying the accident prone area's. The accident data records which can
help to understand the characteristics of many features like drivers
behavior, roadway conditions, light condition, weather conditions and so
on. This can help the users to compute the safety measures which is useful
to avoid accidents. The data set can be analyzing based on Yolo(You only
look once) algorithm will gives the accurate dataset. The models are
performed to identify statistically significant factors which can be able to
predict the probabilities of crashes and injury that can be used to perform a
risk factor.

20
3.3 WORKING

In this project, we will be developing a system which will be constantly

monitor vehicles to check whether accident occurred or not. first we need to collect
the data from the website.In this project we have collect the data from kaggle.after
we need topreprocess the data, because when we collect the data from. the
website there is some missing data or some aspects will occur, so to avoid this we
need to preprocess the data ,this is called as data preprocessing Then apply the
Machine learning algorithms to predict the output . after prediction we will save
the data .Thus, helps to effectively monitor accidents among the general public .

3.4 MODULE DESCRIPTION

• Data collection and Data Preprocessing

• Data cleaning
• Visualization

3.4.1 Data collection and Data Preprocessing :

Collecting data for training the ML model is the basic step in the machine learning
pipeline. The predictions made by ML systems can only be as good as the data on which
they have been trained. Following are some of the problems that can arise in data
collection:
Inaccurate data. The collected data could be unrelated to the problem statement.
Missing data. Sub-data could be missing. That could take the form of empty values in
columns or missing images for some class of prediction.

21
Data imbalance. Some classes or categories in the data may have a disproportionately high
or low number of corresponding samples. As a result, they risk being under-represented in
the model.
Data bias. Depending on how the data, subjects and labels themselves are chosen, the
model could propagate inherent biases on gender, politics, age or region, for example.
Data bias is difficult to detect and remove.
Data Preprocessing : Real-world raw data and images are often incomplete, inconsistent
and lacking in certain behaviors or trends. They are also likely to contain many errors. So,
once collected, they are pre-processed into a format the machine learning algorithm can
use for the model.
Pre-processing includes a number of techniques and actions:
Data cleaning. These techniques, manual and automated, remove data incorrectly added or
classified.
Data imputations. Most ML frameworks include methods and APIs for balancing or filling
in missing data. Techniques generally include imputing missing values with standard
deviation, mean, median and k-nearest neighbors (k-NN) of the data in the given field.

22
 3.4.2 Data cleaning :

Data cleaning is one of the important parts of machine learning. It plays a significant
part in building a model. It surely isn’t the fanciest part of machine learning and at the
same time, there aren’t any hidden tricks or secrets to uncover. However, the success or
failure of a project relies on proper data cleaning. Professional data scientists usually
invest a very large portion of their time in this step because of the belief that “Better data
beats fancier algorithms”.

3.4.3 Data visualization :

Data visualization is the graphical representation of information and data in a pictorial or
graphical format(Example: charts, graphs, and maps). Data visualization tools provide an
accessible way to see and understand trends, patterns in data, and outliers. Data
visualization tools and technologies are essential to analyzing massive amounts of
information and making data-driven decisions. The concept of using pictures is to
understand data that has been used for centuries. General types of data visualization are
Charts, Tables, Graphs, Maps

23
 CHAPTER -4
METHODOLOGY FLOW

This chapter describes the overall design. It also describes each module that
is to be implemented.

4.1 SYSTEM ARCHITECTURE

Figure: 4.1: System Architecture

4.2 Input images:

Our Accident detection data set

This dataset consists of 1,376 images belonging to two classes:

● Accident : 690 images

● Non - Accident : 686 images

Our goal is to train a custom deep learning model to detect whether a accident has
occurred or not .

24
4.2.1 Accident dataset creation :

To create this dataset, Then creating a custom computer vision Python script to add
accidents to them, thereby creating an artificial (but still real-world applicable)
datasetThis method is actually a lot easier than it sounds once you apply landmarks to the
problem.

4.2.2 Working :

First we collect the data from CCTV cameras.Then we apply data preprocessing
technique.after preprocessing the data we train the system by using data
set(accident) & (non accident).after completion of training we will apply YOLO
algorithm to detect the object.By using YOLO algorithm it will divide into frames
then the frames will detect whether the accident is happened or not.

4.3 DISADVANTAGES OF EXISTING SYSTEM:

● Time consuming Process

● More prone to damages

25
4.4 ADVANTAGES OF PROPOSED SYSTEM:

● Less prone to damages

● Faster
● Less hassle

4.5 APPLICATIONS:

● Used in traffic .

● Used in Highways .
● We can also use this at driving schools and vehicle exhibitions

26
 CHAPTER-5
TOOLS AND TECHNIQUES

SOFTWARE REQUIREMENTS

● Python

● Pycharm

5.1. Python

In this project, python is used as the programming language. In technical

terms, Python is an object-oriented, high-level programming language with
integrated dynamic semantics primarily for web and app development. It is
extremely attractive in the field of Rapid Application Development because it
offers the dynamic typing and binding options. Python is relatively simple, so it's
easy to learn since it requires a unique syntax that focuses on readability.
Developers can read and translate Python code much easier than other languages.
In turn, this reduces the cost of program maintenance and development because it
allows teams to work collaboratively without significant language and experience
barriers.
Python supports the use of modules and packages, which means that
programs can be designed in a modular style and code can be reused across a
variety of projects. Once you've developed a module or package you need, it can
be scaled for use in other projects, and it's easy to import or export these modules.
One of the most promising benefits of Python is that both the standard library and
the interpreter are available free of charge, in both binary and source form. There is

27
 no exclusivity either, as Python and all the necessary tools are available on all
major platforms. Therefore, it is an enticing option for developers who don't want
to worry about paying high development costs.

If this description of Python over your head, don't worry. You will
understand it soon enough. What you need to take away from this section is that
Python is a programming language used to develop software on the web and in app
form, including mobile. It's relatively easy to learn, and the necessary tools are
available to all free of cost.

5.2 PYCHARM

PyCharm is the Python IDE by JetBrains, designed for professional Python

developers. Industry-leading code completion, code navigation, safe refactoring, and
smart debugging are just a few important features that contribute to make
professional software development a more productive and enjoyable experience.
PyCharm Professional Edition comes with wide support for Python web
frameworks, modern JavaScript development, as well as with advanced database
tools and scientific tools integrations.

Pros:

● "I have been using PyCharm since over 4 years now and It is one of the best
IDEs out there for python developers. Great product with auto-complete
features."

28
 ● "Also PyCharm have some tools to find errors and helps you to correct them. I also
like the style and colors very nice for me”.

● "The best all in one IDE out there, the python supporting features are great and it
has a many templates for different projects for ease of architecture."

● "PyCharm is probably the best IDE for Python projects as it has so many Python
orientated features. Personally, I only use PyCharm for the occasional Python
project and I am very satisfied with that."

5.3 Software Code Or Deployment:

import cv2
import os
def getFrame(sec, path, out_path):
vidcap = cv2.VideoCapture(path)
vidcap.set(cv2.CAP_PROP_POS_MSEC,sec*1000)
hasFrames,image = vidcap.read()
if hasFrames:
st_path = os.path.join(out_path, "image"+str(count)+".jpg")
cv2.imwrite(st_path, image) # save frame as JPG file
return hasFrames

temp_dir = 'test_input'
output_dir = 'output_frames'

29
for video_file in os.listdir(temp_dir):
path = os.path.join(temp_dir, video_file)
print(path)
output_dir_folder = os.path.join(output_dir, video_file)
os.mkdir(output_dir_folder)
sec = 0
frameRate = 2.0 #//it will capture image in each 0.5 second
count=1
success = getFrame(sec, path, output_dir_folder)
while success:
count = count + 1
sec = sec + frameRate
sec = round(sec, 2)
success = getFrame(sec, path, output_dir_folder)

import cv2
import tensorflow as tf
import os
from skimage.transform import resize
import numpy as np
from PIL import Image
model = tf.keras.models.load_model('model_train')
categories = ["Accident", "Non-Accident"]

30
main_dir = "test_frames"
for dirs in os.listdir(main_dir):
test_dir = os.path.join(main_dir, dirs)
print(test_dir)
a=0
for img_t in os.listdir(test_dir):

image = cv2.imread(os.path.join(test_dir, img_t), cv2.IMREAD_GRAYSCALE)

image = resize(image, (128, 128))
image = image / 255
img_arr = image.reshape([1] + list(image.shape) + [1])
predict_x= model.predict(img_arr)
predict_class=np.argmax(predict_x,axis=1)
if categories[predict_class[0]]=='Accident':
print(img_t, "\t", categories[predict_class[0]])
im=Image.open('test_frames/{}/{}'.format(dirs,img_t))
im.show()
import cv2
import os
def getFrame(sec, path, out_path):
vidcap = cv2.VideoCapture(path)
vidcap.set(cv2.CAP_PROP_POS_MSEC,sec*1000)
hasFrames,image = vidcap.read()
if hasFrames:
st_path = os.path.join(out_path, "image"+str(count)+".jpg")
cv2.imwrite(st_path, image) # save frame as JPG file

31
 return hasFrames

temp_dir = 'test_input'
output_dir = 'output_frames'

for video_file in os.listdir(temp_dir):

path = os.path.join(temp_dir, video_file)
print(path)
output_dir_folder = os.path.join(output_dir, video_file)
os.mkdir(output_dir_folder)
sec = 0
frameRate = 2.0 #//it will capture image in each 0.5 second
count=1
success = getFrame(sec, path, output_dir_folder)
while success:
count = count + 1
sec = sec + frameRate
sec = round(sec, 2)
success = getFrame(sec, path, output_dir_folder)
import tensorflow as tf

from tensorflow.keras.models import Sequential

from tensorflow.keras.layers import Conv2D, MaxPooling2D, Activation, Dropout,
Dense, Flatten
import numpy as np
import matplotlib.pyplot as plt

32
import cv2
from skimage.transform import resize

train = tf.keras.models.load_model('model_train')
categories = ["Accident", "Non-Accident"]

img_test = np.load("labels/test_features.npy")
label_test = np.load("labels/test_labels.npy")
img_test = img_test / 255
img_test = img_test.reshape(list(img_test.shape) + [1])
print(img_test.shape)
train.evaluate(img_test, label_test)[1]
print("index \t label \t predicted")
#classes = train.predict_classes(img_test)
classes = train.predict(img_test)
np.argmax(classes)
for i in range(len(label_test)):
print(i+1,"\t", categories[label_test[i]])
#, "\t", categories[classes[i]]
import tensorflow as tf

from tensorflow.keras.models import Sequential

from tensorflow.keras.layers import Conv2D, MaxPooling2D, Activation, Dropout,
Dense, Flatten
import numpy as np
import matplotlib.pyplot as plt

33
import cv2
from skimage.transform import resize

X = np.load("labels/train_features.npy")
print("St 1")
Y = np.load("labels/train_labels.npy")
print("St 2")
X = X / 255
print("St 3")
X = X.reshape(list(X.shape) + [1])
print("St 4")
print(X.shape)

#creating my model
model = Sequential();

#adding first layer i.e convolutional layer, passing input_shape only in first layer
model.add(Conv2D(32, (5 , 5), activation="relu", input_shape=(128, 128, 1)))

#adding a pooling layer

model.add(MaxPooling2D(pool_size = (2 , 2)))

#adding another convolutional layer

model.add(Conv2D(32, (5 , 5), activation="relu"))

#adding another pooling layer

34
model.add(MaxPooling2D(pool_size = (2 , 2)))

#adding flatteing layer i.e. fully connected layer

model.add(Flatten())

#Embedding nuerons using dense layer

model.add(Dense(1000, activation="relu"))

#Adding a dropout with 50% droupout rate

model.add(Dropout(0.5))

#Embedding nuerons using dense layer

model.add(Dense(500, activation="relu"))

#Adding a dropout with 50% droupout rate

model.add(Dropout(0.5))

#Embedding nuerons using dense layer

model.add(Dense(250, activation="relu"))

#Embedding nuerons using dense layer

model.add(Dense(2, activation="softmax"))

#compiling the model

model.compile(loss = "sparse_categorical_crossentropy", optimizer = "adam",
metrics = ["accuracy"])

35
#training the model
train = model.fit(X, Y, batch_size = 256, epochs = 10, validation_split=0.2,
shuffle=True)
model.save('model_train')

#plotting the model accuracy

plt.plot(train.history["accuracy"])
plt.plot(train.history["val_accuracy"])
plt.title("Model Accuracy Visualization")
plt.ylabel("Accuracy")
plt.xlabel("Epoch")
plt.legend(['train', 'val'], loc="upper right")
plt.show()
import numpy as np
import random
import matplotlib.pyplot as plt
import os
import cv2
from skimage.transform import resize

Train_DIR = r"/home/sparsh/Desktop/Accident_Detection/Accident-Dataset/train"
Test_DIR = r"/home/sparsh/Desktop/Accident_Detection/Accident-Dataset/test"
categories = ["Accident", "Non-Accident"]
test_data = []
train_data = []

36
def getTrainData():
i=0
for index, category in enumerate(categories):
path = os.path.join(Train_DIR , category)
for img in os.listdir(path):
i += 1
print(i)
try:
image = cv2.imread(os.path.join(path, img),
cv2.IMREAD_GRAYSCALE)
image = resize(image, (128, 128))
train_data.append([image , index])
except Exception as e:
pass

def getTestData():
i=0
for index, category in enumerate(categories):
path = os.path.join(Test_DIR , category)
for img in os.listdir(path):
i += 1
print(i)
try:

37
 image = cv2.imread(os.path.join(path, img),
cv2.IMREAD_GRAYSCALE)
image = resize(image, (128, 128))
test_data.append([image , index])
except Exception as e:
pass
print("Getting Train Data")
getTrainData()

print("Getting Test Data")

getTestData()

random.shuffle(train_data)
random.shuffle(test_data)

train_features = []
train_labels = []
test_features = []
test_labels = []

for feature, label in train_data:

train_features.append(feature)
train_labels.append(label)

train_features = np.array(train_features)
train_labels = np.array(train_labels)

38
print(type(train_features))

for feature, label in test_data:

test_features.append(feature)
test_labels.append(label)

test_features = np.array(test_features)
test_labels = np.array(test_labels)
print(type(test_features))

print("Saving Numpy Arrays")

np.save("train_features", train_features)
np.save("train_labels", train_labels)
np.save("test_features", test_features)
np.save("test_labels", test_labels)

39
 CHAPTER-6
FINAL RESULTS
We take the input video from a Source (CC TV camera �) and divide the video
into severalframes.On each frame a Vehicle is detected whether Accident has
occurred or not. If accident occurred it will print the image number with “accident”
message.
6.1 FINAL RESULTS

Fig 6.1 : Accident occurred

When we run the we got this screenshot because of accident

has occurred , System will take the Video from CC TV cameras
and cut the video in such frames then each frame will be
detected using YOLO Algorithm . In this frame two vehicles got

40
smashed to each other .

Fig 6.2 Accident will happen

When we run the we got this screenshot because of accident

has occurred , System will take the Video from CC TV cameras
and cut the video in such frames then each frame will be
detected using YOLO Algorithm . In this frame vehicle is going
in wrong direction soo Accident is going to happen so it will
detected that also .

41
 Comment [1]: Accident occurred

Fig : 6.3 Accident occurred.

When we run the we got this screenshot because of accident

has occurred , System will take the Video from CC TV cameras
and cut the video in such frames then each frame will be
detected using YOLO Algorithm . In this frame two vehicles got
smashed to each other .

42
 Fig 6.4 : No of Accident occurred

In this picture we got 4 images displayed with name accident , because in that
video 4 accidents has happened so we got displayed that.

43
 CHAPTER – 7

7.1 CONCLUSION

Accident detection operation is not an easy task to handle;it can be an extremely

complicated process when it comes toreal time applications, which is the main
reason why it is notimplemented yet on a large scale. The proposed system will
help to improve the present scenarios.

7.2 FUTURE WORK

The proposed system deals with the detection of the accidents. But this can
be extended by providing medication to the victims at the accident spot. By
increasing the technology we can also avoid accidents by providing alerts
systems that can stop the vehicle to overcome the accidents.

7.3 SCOPE OF FURTHER STUDY

The proposed system deals with the detection of the accidents. But this
can be extended by providing medication to the victims at the accident spot.
By increasing the technology we can also avoid accidents by providing alerts
systems that can stop the vehicle to overcome the accidents.

44
 CHAPTER - 8

REFERENCE
1. Kardas K, Cicekli NK. SVAS: surveillance video analysis system.
Expert Syst Appl. 2017;89:343–61.

2. Wang Y, Shuai Y, Zhu Y, Zhang J. An P Jointly learning perceptually

heterogeneous features for blind 3D video quality assessment.
Neurocomputing. 2019;332:298–304 (ISSN 0925-2312).

3.Tzelepis C, Galanopoulos D, Mezaris V, Patras I. Learning to detect

video events from zero or very few video examples. Image Vis Comput.
2016;53:35–44 (ISSN 0262-8856).

4. P.A. Viola and M.J. Jones, ―Rapid object detection using a boosted
cascade of simple features,‖ in Proc. CVPR, 2001, no. 1, pp. 511–518.

5.“Global status report on road safety 2015”, World Health

Organization,2016. [Online]. Available:
http://www.who.int/violence_injury_prevention/road_safety_status/2015
/en/. Accessed: 22- Mar- 2016

6. V. Goud, “Vehicle accident automatic detection and remote alarm

device,” International Journal of Reconfigurable and Embedded
Systems (IJRES), vol. 1, no. 2, Jul. 2012.

7. B. Prachi, D. Kasturi and C. Priyanka, ―Intelligent accident-detection

and ambulance-rescue system,‖ PULSE, vol. 450, no. 16, pp. 2, Jun.
2014.
8. G. Liang, “Automatic traffic accident detection based on the Internet
Things and Support Vector Machine”, International Journal of Smart
Home (IJSH), vol. 9, no. 4, pp. 97-106, Apr. 2015.

9. C.Vidya Lakshmi, J.R.Balakrishnan, ―Automatic Accident Detection

via Embedded GSM message interface with Sensor Technology
,International Journal of Scientific and Research Publications, vol. 2, no.
4, Apr 2012.

45
10. X. Tong, Z. Hong, S. Liu, X. Zhang, H. Xie, Z. Li, S. Yang, W.
Wang and F. Bao, ―Building-damage detection using pre-and post-
seismic high-resolution satellite stereo imagery: a case study of the May
2008 Wenchuan earthquake,‖ ISPRS Journal of Photogrammetry and
Remote.Sensingg, vol. 68, pp.13-27, Mar. 2012.

46