International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 06 Issue: 01 | Jan 2019 www.irjet.net p-ISSN: 2395-0072

Open CV Python Autonomous Car

Professor Chetan S More1, Suanuska Debbarma2, Nidhi Kandpal3, Vibhuti Singh4
1Professor, Dept of Electronics and Telecommunication Engineering, Bharati Vidyapeeth College of Engineering,
Maharashtra , India
2,3,4Student, Department of Electronics and Telecommunication Engineering, Bharati Vidyapeeth College of

Engineering, Maharashtra, India

---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - Numerous kinds of autonomous cars for lurking competitive. In short, they will be a lot, a lot safer than we
motives are being created. The most challenging part of are.
developing an autonomous car is to achieve low cost, low
power and consists of many other missions. The objective of 2. Congestion – There are more people and cars on this little
this project is to develop an autonomous car, which is capable island than ever before, and they’re all in a traffic jam
of both navigating itself through the track and overtaking any between where you are and where you need to be. Driverless
obstacle. Our project is an open CV python autonomous car cars can travel in convoys, inches apart, without any
which is designed to give the human driver a calm and stress- needless dabs of the brake pedal to filter backward through
free driving. When it comes on to designing an automobile traffic and create mysterious, pointless hold-ups.
various aspects have been taken into consideration to make a
vehicle automated. In this project, we have concentrated on 3. Cost – Buying, insuring and maintaining a car is expensive.
the application of an automated car, in which the car will So why not let someone else do it! The future of driverless
detect the obstacle, traffic signal, and stop signs. It can also cars is likely to include sharing schemes, and affordable
drive itself on the track, control, and look out the environment leasing options – cars will be more productive rather than
with the help of Ultrasonic sensors. The input is taken from the spending 99 percent of their lives motionless outside your
camera and ultrasonic sensors, with the help of Haar Cascade house.
Classifiers it analyzes the objects and reports it to the
4. Parking - No longer will it be our problem to find a space –
raspberry pi. To test the car, some obstacles are placed on the
our car will handle it. No more parking tickets, no more
track. The car will navigate itself and overtake all the
dented bumpers, no more endlessly driving in circles waiting
obstacles on the track; it will also change the track when it
for a neighbor to leave. This will unencumber urban road
detects any kind of traffic signs (left, right, and stop). If
area for wider pavements.
anything goes wrong in autonomous mode, user able to switch
to manual mode to control the car. User will also be able to see 1.1 Project Objective
the live view from the car through a webcam.
This project aims to produce an autonomous car using
Key Words: Autonomous Cars, Open CV Python, computer vision technique. This car will be able to navigate
Driverless Car, Raspberry Pi, Object Detection, Traffic itself through the track with the help of computer vision and
Light Detection, Haar Cascade Classifier ultrasonic sensor. User will be able to watch the live view
from the webcam that will be attached to the car. A two lane
1. INTRODUCTION
track will be build and when the car detects the obstacle, it
An autonomous car (also known as a driverless car, self- will move to another lane to overtake the obstacle. The sub-
driving car, robotic car) is a robotic vehicle that is designed objectives are as below:
to travel between destinations without any human
intervention. About 1.24 million People have killed inroads  To produce low cost autonomous car
every year throughout the world. According to BUET
 To produce low power consumption autonomous
accident research centre, the death toll every year is 10-12
car.
thousand and a countless number of people are injured or
become disabled destroying so many lives and families. The  The car is able to navigate itself through the track
autonomous system is a solution to this problem. Here are and overtake any obstacles by computer vision
some ways through which driverless car will change the technique.
world as follows:
 The main goal of self driving autonomous car is to
1. Safety - No matter what we like to believe, humans are no
avoid accidents.
good at driving. The 1.24 million people killed every year on
roads worldwide are proof of that. Unlike us, autonomous
cars will never get drunk, frustrated, dozed off or

© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 595
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 01 | Jan 2019 www.irjet.net p-ISSN: 2395-0072

 People are free to concentrate on other tasks or to 2. Methodology

rest during their journeys.
A Raspberry Pi board, attached with a web camera module
 The present location of the vehicle can be known and an HC-SR04 ultrasonic sensor is used to collect input
using GPS data. The processing unit (computer) handles multiple tasks:
receiving data from Raspberry Pi, object detection (stop sign
1.2 Literature Survey and traffic light), and sending instructions to the car through
a Wifi connection.
 Parth Verma, a talented man did a research with the
title “The Google Autonomous car”. The Google In this project we have used the Haar feature-based cascade
autonomous car was invented by Sebastian Thrun classifiers for object detection. The Haar classifiers use the
who was the co-inventor of the Street view mapping shape-based approach. Since each object requires its own
service. The advantage of this research was that it classifier and follows the same process in training and
decreases human error while driving and result in detection, this project only focused on the stop sign and
reduced possibility of accident occurring. Better fuel traffic light detection.
efficiency is achieved when more autonomous car
on the road which change the driving habits. OpenCV provides a trainer as well as a detector. We collected
Besides that, safety features were included in the the positive samples (samples which contain a target object)
autonomous car. Destination can be reached in less using a cell phone and these samples were cropped such that
time which the car will be able to locate the shortest only the desired object is visible. However, the negative
route. The limitations of this research are when the samples (samples which do not have a target object), were
car is traveling at high speed which over 100mph, collected randomly. In particular, traffic light positive
the car will hardly be able to differentiate the samples contain an equal number of red traffic lights and
objects on the roads. green traffic light.

 A research was done by Michal Ruzicka and Petr Thus initially the car will ask the user to enter the
Masek with the title “Real Time Visual Marker destination. Then the main task of the car is to reach the
Detector and Tracker Based on Computer Vision for destination autonomously through the path while stopping
Semi-autonomous Convoy Purpose”. This research whenever a stop sign is detected, also if a traffic light is in the
is based on controlling the semi-autonomous way the car should wait for a specified period of time if the
convoy by designing the computer vision method. traffic light is red and if the traffic light is green the car
The advantages of this research is low power should move continuously.
consumption and efficient product cost. The
limitations of this research are using 320 x 240 low
resolution of captured frames which will cause the
result inaccurate. Higher resolution cannot be used
because of the low cost power device which does
not do well in real time processing due to low frame
rate. But this problem can be solve by OpenCL
optimization.

 Another research was done with the title

“Autonomous Three-Wheeled Robot with Computer
Vision System”. The research is based on three
wheel robot with computer vision. The advantage of
this research is the product has many features such Fig -1: Block Diagram
as GPS, accelerometer and compass which can use
for data collections purpose. Higher accuracy result
is achieved with higher resolution and frequency.
The weakness of this research is having a bad robot
control method which can change to Wi-Fi control
that can support higher control range. Other than
that, the robot cannot operate at dark surroundings.
The robot may have problem with the movement
which cannot avoid the obstacles with accurate
turning angle of the robot.

© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 596
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 01 | Jan 2019 www.irjet.net p-ISSN: 2395-0072

[3] J M.A. Alvarez, A.M. Lopez & R. Baldrich, Illuminant

Invariant Model-Based Road Segmentation. Intelligent
Transportation Systems, IEEE Transactions on, 12, 2008, pp
184–193.

[4] A. Bar Hillel, R. Lerner, D. Levi, & G. Raz. Recent progress

in road and lane detection: a survey. Machine Vision and
Applications, Feb. 2012, pp. 727–745

Fig -2: Workflow Model

3. CONCLUSION

Therefore, we conclude that in this project we have made the

use of different components and a basic programming
language to create an autonomous car which will provide us
live feed besides detecting obstacles on the track so as to
represent a short model of a vehicle with less probability of
accidents. This autonomous car is currently prepared in
small scale but in future it can be brought to a broader scale
if the project meets success. The speed of the car is however
not controlled but it can be, by bringing about modifications
and changes while developing it in large scale in future. This
project took us through various phases of learning and
project development, also we have gained a lot of new
technical and non-technical knowledge.

REFERENCES

[1] OpenCV Documentation – Haar Feature based Cascade

Classifier for Object Detection

[2] OpenCV Documentation - Cascade Classifier Training

© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 597