IRJET-V8I686.pdf

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 08 Issue: 06 | June 2021 www.irjet.net p-ISSN: 2395-0072
© 2021, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 457
Autonomous Vehicle Using Machine Learning and Computer Vision.
Gaurav Kolhe1, Atharva Pawar2, Bhargav Kakade3 , Soham Lambate4,
Dr. Mohini Sardey5
1-4B.E. (students), 5Head of Department, Dept. of Electronics & Telecommunication Engineering,All India Shri
Shivaji Memorial Society’s Institute of Information Technology, Pune, Maharashtra, India.
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract -In this proposedwork, aprototypeofautonomous
vehicle is designed and developed which uses raspberry pi as
the core functioning and uses OpenCV and machine learning
technology. The vehicle uses the core processing unit as
Raspberry pi, which is interfaced with the Pi camera module,
which feeds the captured images to the raspberry pi for image
processing. Based on which detection like road lanes, traffic
signals, obstacles are done and commands are sent to arduino
to operate the car.
Key Words: OpenCV, Pi camera, Raspberry Pi, Arduino,
Machine Learning
1. INTRODUCTION
Road Safety is one of the major issues faced globally. Every
year roughly 5 million people meet with road accidents and
around 1.3 million lose their lives worldwide. These deaths
are majorly caused by human negligence and human error.
The most common reasons for road accidents are over
speeding, drink & driving, distractions to driver, red light
jumping, lane changing and avoiding safety gears like seat
belts. Traditional and manual cars require complete human
control and attentionoverthemwhiletravelling,wheremost
of the people make mistakes which eventually results into
road accidents. To overcome thisissueAutonomousvehicles
technology come into picture. Autonomous vehicles are
those vehicles which are capable of sensing its environment
and move safely with no human input. Various high-end
sensors are situated in different parts of the car which
detects its surrounding and sends the data to the High-
performance computing devices which then instructs the
vehicle to follow a particular path and take decisions.
Majorly with the help of video cameras, vehicle detects the
traffic lights, road signs, pedestrians and other cars on the
road and sends those real time images to the processor.
Machine learning is mostlyusedforperceptionanddecision-
making in real-time. The algorithms need to make the
decisions within a fraction of a second for applying a break
or to make a turn right or left. The main goals of project is to
a design and develop a low-cost prototype of autonomous
vehicle which uses raspberry pi as the core functioning unit
and uses OpenCV and machine learning technology. The
vehicle uses the core processing unit as Raspberry pi, which
is interfaced with the 8MP Pi camera module, which feeds
the captured images to the raspberrypiforimageprocessing
.Based on which detection like road lanes, traffic signals,
obstacles are done and commands are send to arduino to
operate the car and follow designed path on our self-made
track, which helps car to travel to specified place cautiously
and timely.
2. LITERTURE SURVEY
1. Truong-Dong Do, Minh-Thien Duong [1] proposeda
monocular vision based in this paper, a monocular
vision-based autonomous car prototype using Deep
learning on Raspberry Pi. Convolutional Neural
Networks (CNNs) have shown great results in this
domain. The main reason behind these great results
is their ability to learnmillionsofparametersusinga
large data. In this work, authors have focused on
finding a model which directly maps raw input
images to a predict steering angle as output using a
neural network. The CNN model parameters were
trained byusingdatacollectedfromvehicleplatform
built with RC car, Raspberry Pi 3 Model B computer
and front facing camera.
2. “Convolutional Neural Network based Working
Model of Self Driving Car - A Study” [2] authors of
this paper proposed that to achieve good results
then technologies like Deep learning techniques
namely Convolution Neural Networks, YOLO
algorithm, Hough Transform Algorithms, Canny
Edge Detection algorithm. Software components
such as Arduino IDE, Raspberry Pi Cam Interface,
Open CV and hardware components such as
Raspberry Pi 3, Arduino UNO, Pi Camera, sensors
should be used.
3. “Traffic Light Detection and Recognition for Self-
Driving Cars using DeepLearning”[3]authorsofthis
paper proposed deep learning-based model for
detection and recognition of traffic lights using
transfer learning. The method uses faster region
based convolutional network (R-CNN) Inception V2
model in TensorFlow for transfer learning. The
model was trained on dataset containing different
images of traffic signals in according to the Indian
Traffic Signals which are distinguished in five types
of classes. The model was able to detect traffic light
with correct type.
4. “A Lane Detection Approach for Driver Assistance”
[4] authors of this paper proposed a general
technique for lane detection that combines filters
developedbyFreemanandAdelsonwithCannyedge
detection algorithm.Thesteerablefiltersusedinthis
paper are based on second derivatives of two-

dimensional Gaussians. These filters are helpful in
many early vision and image processing tasks. The
filter results are then processed to remove noise
based on the road. Therealgorithmisabletoprovide
accurate extraction of lane edge markings under
different light and road situations.
5. “A Vision-based Method for Improving the Safety of
Self-driving” [5] in these authors have used
computer vision and deep learning to train existing
data sets. They have used efficient convolution
neural networks trained the data in Traffic Sign
Recognition Benchmark and KITTI respectively to
realize classification of traffic signs and detection of
vehicles, and functions of OpenCV are used to
identify and locate trafficidentificationlines.Toplan
and make decisions on the driving path for the car.
3. METHODOLOGY
Figure 1: Circuit Diagram
The raspberry-pi is the main core controller which will be
attached on the vehicle. The module of pi-camera will be
attached on the top of the prototype. The ln298 IC is used for
the movement of the prototype.Computervisionisutilizedto
find object in front of the prototype and take necessary
movements. When there is any object in front of the
prototype andisinameasurabledistancefromtheprototype.
The raspberry-pi gives instructions to the arduino uno and it
gives orders to the l298n integrated circuits to stop giving
power to the tires and therefor halts the movement of the
prototype contingent on the nearness of the object. The
displacement calculated is also showed on the output
window of the program. The following process is finding of
lanes and road traffic signs.
3.1 Hardware Used
3.1.1 Raspberry Pi
Raspberry Pi 3 B+ is the main processing unit of this project
and is mounted on the top of the car. Using raspberry pi 3 B+
version for image processing with the help of OpenCV
software. Pi camera is interfaced to the Raspberry Pi. A
machine learning algorithm is implemented and the images
are trained using neural network technology. All the function
like traffic sign detection and obstacle detectionisperformed
Raspberry Pi using machine learning.
Figure 2: Raspberry Pi 3 B+
3.1.2 Raspberry Pi Camera
The Pi-camera used is V2 version which is having
specifications of 8 mega pixels and supports1080p 30 fps
resolutions along with Sony IMX219 sensor with F2.9
aperture. This Pi cameracaptures the images andsendsitto
theraspberrypiforimageprocessingand machinelearning.
Figure 3: Raspberry Pi Camera V2 8 MP
3.1.3 Arduino Uno
Arduino Uno is ATmega329P based microcontroller board.
In our project Arduino is used to controlforward,backward,
stop,left and right movementofvehicle.Allthefunctionsare
programmed in Arduino using Arduino IDE.
Figure 4: Arduino Uno
3.1.4 Motor Driving Circuit L298N
It is a basic motor driver module used to drive dc motors as
well as stepper motors. H Bridge is used along with L298N
IC to drive motors

Figure 5: L298N
3.2 Software Used
3.2.1 Arduino IDE
The Arduino Integrated Development Environment (IDE)is
a software application (for Windows, macOS, Linux). The
Arduino IDE has the languages C and C++ usingrulesofcode
structuring. The maincode, created on the IDE platformwill
generate a Hex File which is then uploaded in the controller
on the board. The IDE environment mainly contains two
major parts: Editor and Compiler where editor is used for
writing the required codeandcompilerisusedforcompiling
and uploading the code into the Arduino Module.
3.2.2 OpenCV
OpenCV (Open Source Computer Vision Library) is an open
source computer vision and machine learning software
library developed by Intel. OpenCV is used to provide a
common architecture for computer vision applications and
to accelerate the use of machine learning in the commercial
products. The library has more than2500algorithms,which
includes a set of both classic and state-of-the-art computer
vision and machine learning algorithms. These algorithms
are used to detect and recognize faces, identify objects,
classify human actions in videos, track camera movements,
track moving objects. Computer Vision tasks performed in
this project are:
 Region of Interest
 Perspective Transformation (Bird Eye View)
 Grayscale Conversion and Thresholding
 Canny Edge Detection
 Gaussian Blurring
3.3.3 Haar Cascade
Haar cascade classifier is a machine learning object
detection software that identifies objects in an image and
video. The algorithm can be state in four phases: 1.
Calculation of Haar Features, 2. Creating Integral Images
using Adaboost, 3. Implementing CascadingClassifiers.This
algorithm requires a lot of positive images samples and
negative images samples of its surroundings to train the
machine learning model.
4. TEST RESULTS
Figure 6: Lane detection
Figure 7: Stop Sign Detection
5. CONCLUSION
A prototype of an autonomous vehicle model isdesigned and
developed. With the help of Image Processing and Machine
Learning techniques a successful model is developed which
worked as per our expectation. Image processingtechniques
like calculating region of interest(ROI),conversionofimages
from RGB into gray scale, applying Gaussian filter for
removing noise, canny edge detection for extracting lane
edges are successfully performed for detection of lanes as
well as machine learning model is successfully trained in
HAAR cascade software by using positive and negative
samples of Stop Sign, Red Light, Green light and Obstaclesfor
traffic sign detection. At last, the car is able to follow lane
efficiently and the traffic signs are detected properly as well
as followed and decisions are made accurately.
6. FUTURE SCOPE
Further improvements can be made in this project by using
GPU based high speed processing system and a high
definition camera. In addition, computer vision techniques
such as camera calibration, structure from motion, etc. can
have very profound impact on the accuracy of navigation.

7. AKNOWLEDGEMENT
We wish to express our deep sense of gratitude and honor
towards our respected guide Dr. Mohini Sardey for her
guidance and constant encouragement. Along with that, we
gratefully acknowledge and express heartfelt regard to all
the people, who helped us in making the idea of theproject,a
reality.
REFERENCES
[1] T. Do, M. Duong, Q. Dang and M. Le, "Real-Time Self-
Driving Car Navigation Using Deep Neural
Network," 2018 4th International Conference on
Green Technology and Sustainable Development
(GTSD), 2018, pp. 7-12, doi:
10.1109/GTSD.2018.8595590
[2] P. G. Chaitra, V. Deepthi, S. Gautami, H. M. Suraj and
N. Kumar, "Convolutional Neural Network based
Working Model of Self Driving Car - a Study," 2020
International Conference on Electronics and
Sustainable Communication Systems(ICESC),2020,
pp.645650,doi:10.1109/ICESC48915.2020.9155826
[3] R. Kulkarni, S. Dhavalikar and S. Bangar, "Traffic
Light Detection and Recognition for Self Driving
Cars Using Deep Learning," 2018 Fourth
International Conference on Computing
Communication Control and Automation
(ICCUBEA), 2018, pp. 1-4, doi:
10.1109/ICCUBEA.2018.8697819.
[4] H. Wu, Y. Liu and J. Rong, "A Lane Detection
International Conference on Information
Engineering and Computer Science, 2009, pp. 1-4,
doi: 10.1109/ICIECS.2009.5365404.Approach for
Driver Assistance," 2009
[5] D. Dong, X. Li and X. Sun, "A Vision-Based Method
for Improving the Safety of Self-Driving,"201812th
International Conference on Reliability,
Maintainability, and Safety (ICRMS), 2018, pp. 167-
171, doi: 10.1109/ICRMS.2018.00040.
[6] https://www.arduino.cc/en/Reference/Board?fro
m=Guide.Board
[7] https://www.raspberrypi.org/documentation/confi
guration/
[8] https://www.raspberrypi.org/downloads/raspbian
/
[9] https://www.raspberrypi.org/documentation/hard
ware/camera/
[10] www.opencv.org

IRJET-V8I686.pdf

More Related Content

What's hot (20)

Similar to IRJET-V8I686.pdf (20)

Recently uploaded (20)

IRJET-V8I686.pdf