Project File Details


Original Author (Copyright Owner): 

3,000.00

Instant Download

Download your project material immediately after online payment.

100% Money Back Guarantee

File Type: MS Word (DOC) & PDF
File Size: 1,833KB
Number of Pages:51

ABSTRACT

Over the ongoing years, innovatively propelled nations have kept on joining the quest for growing completely self-sufficient driven vehicles. This Autonomous vehicles intend to address issues of driver profitability and effectiveness. Dependable traffic light discovery is a vital segment for self-sufficient driving. Recognizing the traffic lights amidst everything is a standout amongst the most significant errand. The focus of this research is to develop and find the optimal parameters for an efficient Neural Network Architecture to aid a hardware engineer to implement on a hardware for the autonomous vehicle. This is done by designing an Artificial Neural Network (ANN) that would be capable of detecting and correctly classifying any traffic light within the city of Abuja, Nigeria. This study first attempts to develop a reliable traffic sign detector by constructing MLP, training using BP, and tuning various Convolutional Neural Networks (CNN). Images for training are obtained from Abuja city metropolis
Keywords: Backpropagation (BP), neural networks, CNN, MLP, ANN

TABLE OF CONTENTS

Contents
CERTIFICATION ……………………………………………………………………………………………………………………………. 4
SIGNATURE PAGE ………………………………………………………………………………………………………………………… 5
ABSTRACT …………………………………………………………………………………………………………………………………… 6
DEDICATION ……………………………………………………………………………………………………………………………….. 7
ACKNOWLEDGMENTS ………………………………………………………………………………………………………………….. 8
LIST OF TABLES ………………………………………………………………………………………………………………………….. 11
LIST OF FIGURES ………………………………………………………………………………………………………………………… 12
CHAPTER ONE …………………………………………………………………………………………………………………………… 14
INTRODUCTION …………………………………………………………………………………………………………………………. 14
1.1 Research Background …………………………………………………………………………………………………… 14
1.2 Artificial Neural Networks …………………………………………………………………………………………….. 15
1.3 Neural Network Architectures ………………………………………………………………………………………. 15
1.4 Problem Statement ……………………………………………………………………………………………………… 16
1.5 Research Aim and Objectives: ………………………………………………………………………………………. 16
1.6 Limitation of the Study…………………………………………………………………………………………………. 17
Chapter Two ……………………………………………………………………………………………………………………………… 18
Literature Review ………………………………………………………………………………………………………………………. 18
2.1 Artificial Neural Network …………………………………………………………………………………………………… 18
2.1.1 Perceptron ………………………………………………………………………………………………………….. 20
2.1.2 The Neuron Model (Single-Input Neuron) ……………………………………………………………….. 20
2.1.3 Activation function ……………………………………………………………………………………………….. 22
2.1.4 Cost function ……………………………………………………………………………………………………….. 22
2.1.5 Forward propagation ……………………………………………………………………………………………. 23
2.2.6 Backward propagation ………………………………………………………………………………………….. 24
2.2 Generalization and overfitting ………………………………………………………………………………………. 24
2.3 Autonomous Vehicles ………………………………………………………………………………………………….. 25
2.4 Artificial Neural Network in Autonomous Vehicles ………………………………………………………….. 25
Chapter Three …………………………………………………………………………………………………………………………… 26
Research Methodology ………………………………………………………………………………………………………………. 26

CHAPTER ONE

INTRODUCTION
1.1 Research Background
Computer vision takes root in signal processing; wherein the effect of a system on a signal is studied, and frameworks for exploring this effect are examined. The black box is an excellent illustration of the basic premise of signal processing. In general, a signal is input into the black box, propagates through the unknown system therein, and another signal is output. The usual question we tend to ask is, “what is in the black box?” Typically, we may look at the input and output in terms of specific characteristics called metrics to infer some quality of the process undergone inside the black box. A research engineer may even explore different metrics that seem intuitive based on other information about the process and describe the unknown system in a novel way. The issue with the discovery attitude is that no measurement or set of measurements can show with sureness the substance of the black box in light of the fact that numerous interesting capacities exist with a similar arrangement. While overseeing multifaceted nature, utilizing presumptions and approximations may do the trick for some different fields of building. The information in Machine vision is excessively entangled and voluminous for this methodology.
Conventional traffic light detection methods often suffer from false positives in an urban environment because of complex backgrounds. To overcome such limitation, Deep Neural Network is emphasized, which is fast, but weak to false positives (Lee & Park, 2017). To realize autonomous vehicles, image recognition with high accuracy and high speed is necessary for the vehicle environment. ANNs are recently used in many machine
15
learning applications, from speech recognition and natural language processing to computer vision, and image recognition.
Conventional traffic light serves as the input to the black box, which by design, using the neural network will produce the desired output.
1.2 Artificial Neural Networks
An ANN is based on a collection of connected units or nodes called artificial neurons, which loosely model the neurons in a biological brain. Each connection, like the synapses in a biological brain, can transmit a signal from one artificial neuron to another. An artificial neuron that receives a signal can process it and then signal additional artificial neurons connected to it, thus producing the desired output, which is very useful for decision making in autonomous vehicle systems. The neural network itself is not an algorithm, but rather a framework for many different machine learning algorithms to work together and process complex data inputs. These systems “learn” to perform tasks by considering examples, generally without being programmed with any task-specific rules (Marcel van Gerven et al., 2017).
1.3 Neural Network Architectures
There are many neural network architectures; their different layers of neurons regularly organize them. These layers comprise of input, hidden, and output layers. Two metrics are frequently used to measure the neural network size, the number of neurons and the number of parameters. It is essential to mention that the network size plays an integral part in designing a neural network.
16
1.4 Problem Statement
Efficiency, accuracy, and quick decision making is a common challenge in the field of Autonomous Vehicles. Rapid processing of data helps in speeding up any operation to be performed on such data. Many frameworks for traffic light detection using machine learning have been proposed in the past but were rather time-in-efficient in either the reduction of the dimension or in terms of the efficiency of the machine learning algorithms used. The need for better approaches that can improve the computational problems associated with image processing or classification is in high demand and cannot be overemphasized.
This thesis aims to explain the design of a Neural Network architecture that will be used for image classification and in this case, a traffic light detection in autonomous vehicles. It describes the theory behind the neural network and Autonomous Vehicles, and traffic light dataset as its only input that can be designed to test and evaluate the algorithm’s capabilities. The thesis will show that the Artificial Neural Network can, with an image resolution of 64 × 64 and a training set with 55 images, make decisions with a 0.54 confidence level.
1.5 Research Aim and Objectives:
Reliable traffic light detection is a crucial component for autonomous driving. One of the main tasks that such a vehicle must perform well is the task of following the rules of the road. Identifying the traffic lights amid everything is one of the most critical tasks.
The main objectives of the research are:
i. Design a neural network for image classification.
17
ii. Develop and find the optimal parameters for an efficient Neural Network Architecture to aid a hardware engineer to implement on hardware for the autonomous vehicle to improve driver productivity, enhance transportation efficiency, and increase safety.
1.6 Limitation of the Study
This research study is limited to the design of neural networks and classification of traffic light images for autonomous vehicles only. The study highlights the significance of neural networks in autonomous vehicle decision making.

GET THE FULL WORK

DISCLAIMER: All project works, files and documents posted on this website, projects.ng are the property/copyright of their respective owners. They are for research reference/guidance purposes only and the works are crowd-sourced. Please don’t submit someone’s work as your own to avoid plagiarism and its consequences. Most of the project works are provided by the schools' libraries to help in guiding students on their research. Use it as a guidance purpose only and not copy the work word for word (verbatim). If you see your work posted here, and you want it to be removed/credited, please call us on +2348157165603 or send us a mail together with the web address link to the work, to hello@projects.ng. We will reply to and honor every request. Please notice it may take up to 24 or 48 hours to process your request.