SIMULATION OF AUTONOMOUS VEHICLE AND RELATIVE DISTANCE CALCULATION
Chapter 1
INTRODUCTION
Many researchers and tech companies are competing to develop autonomous vehicles using
either traditional or Deep Learning approach. An autonomous vehicle which can receive its
environment such as other vehicles, pedestrian’s, traffic safety instructions, and move
automatically without any human interaction. The control system in the autonomous vehicle use
the sensory information to navigate the roads, avoid the obstacles and follow the traffic rules.
1.1 Aim
The main aim of the project is to detect the lane to identify the path in which the vehicle
should move, object detection, calculate the steering angle and the relative distance calculations
of the identified object in front of the vehicle. There are many autonomous vehicles that have
been developed for satisfying foreign road conditions. Our goal is to train the model that satisfies
the Indian roads.
1.2 Purpose
The purpose of this project is fewer the traffic accidents and congestion, improve fuel
efficiency, optimize fuel economy and help reduce harmful emissions. People with disabilities,
like the blind, are capable of self-sufficiency, and highly automated vehicles can help them live
the life they want. Autonomous vehicles maintain a safe and consistent distance between
vehicles, helping to reduce the number of stop-and-go waves that produce road congestion.
1.3 Overview
Autonomous vehicles can cause less traffic accidents compared to human drivers since they
are equipped with advanced sensors and technologies that can outperform humans in every
aspect. Autonomous vehicle are programmed to obey traffic laws such as speed limits of
distractions and therefore eliminate human error.
A Convolutional Neural Network is a Deep Learning algorithm which can atke in an input
image, assign importance to various aspects/objects in the image and be able to differentiate one
from the other. The goal of this project is to detect the lane by first taking the lane image and
DEPT OF CSE, NIEIT, MYSURU, 2019-20 Page 1
SIMULATION OF AUTONOMOUS VEHICLE AND RELATIVE DISTANCE CALCULATION
converting into grayscale. The grayscale image is given as the input to Canny Edge Detector.
Before detecting edge of the image using Guass smoothing filter to reduce noise. Hough
Transform (HT) can provide a significant solution to the problems associated with line detection
in an image. It replaces the original problem of detecting collinear points in an image space by a
simpler task of finding peak points in a parameter space.
Our system uses You Only Look Once (YOLO) at an image to predict what objects are
present and where they are. The algorithm applies a neural network to an entire image. The
network divides the image into a SxS grid and comes up with bounding boxes, which are drawn
around images and predicted probabilities for each of these regions.
The vehicle will be managed to drive autonomously by training a convolutional neural
network to predict steering wheel angles based on steering angle data and images captured by
three cameras (left, center, right) mounted in front of the car. The trained model will be able to
accurately steer the car using only the center camera. We will take the input image, the image is
cropped, normalized and resized using Tensorflow. We apply a series of convolutional layers.
Each convolutional layer is followed by a BatchNormalization operation. We flatten the input at
this stage and enter the fully connected phase. We apply a series of fully connected layers, of
gradually decreasing sizes. The output layer is obviously of size 1, since we predict only one
variable, the steering wheel angle.
The distance between our vehicle and the vehicle that have been detected is done by taking
the midpoint of our vehicle and the midpoint of the detected vehicle and applying the distance
formula.
DEPT OF CSE, NIEIT, MYSURU, 2019-20 Page 2
SIMULATION OF AUTONOMOUS VEHICLE AND RELATIVE DISTANCE CALCULATION
Chapter 2
LITERATURE SURVEY
A literature survey or a literature review in a project report is that section which shows the
various analyses and research made in the field of interest and the results already published,
taking into account the various parameters of the project and the extent of the project. It is the
most important part of report as it gives a direction in the area of our research. It helps to set a
goal for the analysis - thus giving the problem statement.
2.1 Survey Papers
A literature review or narrative review is a type of review article. A literature review is a
scholarly paper, which includes the current knowledge including substantive findings, as well as
theoretical and methodological contributions to a particular topic. Literature reviews are
secondary sources, and do not report new or original experimental work. We have surveyed the
following IEEE papers for our project.
In the paper Canny edge detection based on Open CV by Zhao Xu1,
Xu Baojie, Wu Guoxin1, in 2017, they presented , a approach for a traditional edge detection
algorithm The edge is the most basic feature of an image, which refers to the set of pixels that
have a sudden change in the grey level. Edge detection is a basic method to recognize and
segment the edges of images based on grey discontinuous points. The Canny operator is a
multiply-scale edge detection algorithm proposed by John F.Canny, the goal in our project is to
find an optimal edge detection algorithm, which is widely used in the field of image processing,
and are constantly improved and innovated. The Canny operator chooses the edge points by
threshold approach,
so, the threshold value has an important influence on the results of image edge detection. How to
obtain an accurate threshold and determine the optimal threshold is a key problem, however the
influence of different thresholds cannot be visualized by numbers on the detection results. Open
DEPT OF CSE, NIEIT, MYSURU, 2019-20 Page 3
SIMULATION OF AUTONOMOUS VEHICLE AND RELATIVE DISTANCE CALCULATION
CV is a great tool for image processing with strong function, it has a sufficient computer vision
library to deal with graphic problems. Programming with Open CV algorithm library can display
different results under different thresholds, which is convenient for finding the optimal threshold,
and is propitious to image edge detection.
The following algorithm used in Canny edge detection algorithm are:
Step1:The gray-scaled image is sent as an input to the gaussian smoothing filter to reduce the
noise in the image
Step 2: Calculate the gradient amplitude and direction, usually the gradient direction takes the 4
angles - 0, 45, 90, and 135 degrees
Step 3: Non maximum suppression ˈ which eliminates non edge pixels, leaving only a few fine
lines.
Step 4: Select the hysteresis threshold, hysteresis threshold needs two thresholds which retain or
exclude pixels to select the edge. If the amplitude of the pixel position is higher
than the high threshold, the pixel is reserved as an edge pixel.
If the amplitude of the pixel position is less than the high threshold, the pixel is excluded.
If the amplitude of the pixel position is between the two thresholds, the pixel is reserved only
when connected to a pixel higher than the high threshold.
In the paper Implementation of Hough Transform for Image Processing Applications by L.
Chandrasekar and G. Durga, they present concept of hough transformation for Road Lane
Detection to track the lane from video’s frames and to provide additional information that can be
helpful for the decision-making process of the self-driving car. Hough Transform (HT) can
provide a significant solution to the problems associated with line detection in an image. Line
detection is used to detect the presence of lines in an image, at a particular orientation. The
importance of line detection is used for detecting sharp changes in image brightness. HT is a
popular technique in computer vision. The HT for line detection replaces the original problem of
detecting collinear points in an image space by a simpler task of finding peak points in a
parameter space. So the basic concept is from an extracted points in an image(aligned points) we
can be able to detect the lines present in the image Hough transform is a method of representing
DEPT OF CSE, NIEIT, MYSURU, 2019-20 Page 4
SIMULATION OF AUTONOMOUS VEHICLE AND RELATIVE DISTANCE CALCULATION
any straight line over Hough space. Every straight line is represented by the equation y = ax + b,
i.e. for constant values of a and b the graph will represent the value of y for every variable value
of x. So, if we have a plane represented by (a,b) then the above specified line can be represented
only by a dot over the (a,b) plane which is known as Hough plane. Any straight line can also be
represented in polar equation i.e xcosθ+ ysinθ=ρ, where ρ is the perpendicular distance from
the origin to the line and θ being the angle between x axis and perpendicular distance ρ . So an
edge detected image is given as input to hough transformation, for every point present on the
image the algorithm starts to iterate over each possible value of (θ, ρ).The occurrences are
defined into the Hough space for each point. Each occurrence adds weightage to the Hough
space corresponding to the values of (θ, ρ). The max occurrence of any point on Hough space
would result in conclusion that there is a straight line represented by the Hough space point
acquired by the equation: xcosθ+ ysinθ=ρ.
Chapter 3
Conclusion
DEPT OF CSE, NIEIT, MYSURU, 2019-20 Page 5
