Abstract
Parking space detection is a major challenge in our cities and drivers waste time when moving from one place to another in search of a free parking space. The current parking space detection systems available are based on sensors which are costly to install and maintain. The sensor systems cannot also be used outdoor environments such as in cities as the sensors can be stolen or vandalized. This study compared the performance of M-RCNN and YOLO algorithm which are a deep learning algorithms used to classify images. YOLO was seen to be the best to use in this study because it was able to run under low comping resources and give accurate predictions. It was thus used to develop a prototype that was used for detecting the status of parking slots as either empty or occupied.
The solution was verified by feeding it with a parking area video stream that had vehicles coming and leaving the parking area and monitoring how well its able to identify the vehicle objects from other objects and how well it is able to predict the status of a parking area as either vacant or occupied. The model achieved an accuracy of 92.6% in parking space status detection. Our experiments showed that the model proposed can be used to achieve automated parking space status detection in any marked parking area.
Definition of Terms
Parking Slot: An area designated for parking vehicles.
Deep Learning: Form of machine learning where human brain functions are mimicked in terms of processing information and drawing patters for decision making.
Automated Parking: Process devoid of human control.
CNNN: Convolution Neural Network
M-RCNN: Mask Region Based Convolution Neural Network
YOLO: You Only Look Once
List of Figures
Figure 1: Counter Based Systems for Parking Slots 18
Figure 2: Sensor Based Parking System 19
Figure 3: Vision Based Parking Systems 20
Figure 4: Gray Scale Image representation in a Matrix 23
Figure 5: Conversion of a Gray Scale image to a Binary Image 24
Figure 6: Perceptron Neural Network 25
Figure 7 Convolution Neural Network 25
Figure 8 Region Based CNN 26
Figure 9 Faster R-CNN 27
Figure 10 Yolo 28
Figure 11: Conceptual Model for Automated Parking Spot Detection 30
Figure 12: Model Classification Process 31
Figure 13: YOLO Performance in Object detection 43
Figure 14: Image of the Parking Area 38
Figure 15: Parking areas Marked 38
Figure 16: Parking Slot Coordinates 39
Figure 17: Model Flow Chart 41
Figure 18: Video Stream of the Parking Area 41
Figure 19: Image of Video Stream Output of detected parking spots, vacant spots and parking count 41
Figure 20: Yolo Car Detection 43
Figure 21: Mask-RCNN Vehicle Detection 46
Figure 22: Yolo Image detection 46
Figure 23:Original Parking spot Image 47
Figure 24: Sobel Edge detection 47
List of Tables
Table 1: Model Performance with Motion Detection Results 44
Table 2: Model Performance with No Motion Detection Results 45
Table of Contents
Declaration 2
Dedication 3
Acknowledgement 4
Abstract 5
Definition of Terms 6
List of Figures 7
List of Tables 9
Chapter One: Introduction
1.1 Background 13
1.2 Problem Statement 14
1.3 Research Objectives 15
1.4 Research Questions 15
1.5 Scope 16
1.6 Significance of the Study 16
Chapter Two: Literature Review
2.1 Background 17
2.1.0 Counter Based Systems 17
2.1.1 Sensor Based Systems 18
2.1.2 Vision Based Systems 19
2.2 Related Research 20
2.3 Computer Vision and Machine Learning Algorithms for Automated Parking Space Detection 22
2.3.1 Neural Networks 24
2.3.2 Convolution Neural Networks 25
2.3.3 Region Based CNN (R-CNN) 25
2.3.4 Faster R-CNN 26
2.3.5 Mask RCNN 27
2.3.6 YOLO 27
2.4 Research Gap 28
2.5 Process Model 29
Chapter Three: Methodology
3.0 Introduction 32
3.1 Research Design 32
3.2 Business Understanding 33
3.3 Data Understanding 33
3.4 Data Collection and Preparation 33
3.5 Modelling 34
3.6 Evaluation 34
3.7 Deployment 35
CHAPTER 4: ANALYSIS, DESIGN AND IMPLEMENTATION
4.1 Introduction 36
4.2 Attribute Selection 36
4.3 Analysis of Testing Data 36
3.5 Modelling 37
4.6 Implementation 41
4.7 Prototype Evaluation 41
CHAPTER 5: RESULTS AND DISCUSSIONS
5.0 Introduction 43
5.1 Model Performance 43
5.2 Discussions 45
References 52
Chapter One
Introduction
1.1 Background
Identification of parking spots in cities is a challenge especially during peak hours when people are reporting to work. An average of 7-8 minutes is spent by people cruising for a parking spot in the city which contributes to 30% of the traffic in the cities (D.Azshwanth, 2019). Most Motorists who park their vehicles in Nairobi Central Business District are forced to wake up early so that they can find uncopied parking spots (Musulin, 2017). To alleviate this problem, suggestions on how to solve city parking spots have been given but majority have not been effective. An example, in Nairobi City, it had been suggested that parking fee should be doubled so as to make people leave their cars at home but the proposal was rejected by the courts (Mediamax Network Limited, 2019). Other suggestions such as increasing the number of parking spots have also been suggested but the solution is not sustainable due to the increasing urbanization.
Computer Vision and deep learning algorithms can be applied to solve the parking spot detection in cities and shopping malls. This will involve training a model to be able to recognize parking spaces from other spaces, and vehicle detection. When parking spots and vehicles are detected, classification can then be applied to distinguish between a parking spot that has been occupied by a vehicle from one that is not occupied. This information can then be mapped to a user interface which drivers can use to check for parking spots in the city and in shopping malls. Drivers with thus be empowered with knowledge on where to locate a vacant parking area thus saving on time and preventing traffic congestion in the city.
Suggestions to use Bluetooth technologies and sensors have been recommended, however this is an expensive as it requires the existing parking slots be revamped and it is not applicable in open parking areas due to vandalism of the materials. This thus makes computer vision a good approach as it involves use of a camera which is strategically position to view the parking spots. The training and recognition of empty and occupied parking spots is done using machine learning algorithms such as convolution neural network which has enhanced object detection capabilities (Girshick R, 2015).
Computer Vision Parking spot detection System using deep learning can be based on two approaches, i.e. Object detection and Image classification (Rivano & Mouël, 2017). Image classification, an image if first captured by a camera and parking spaces are identified and segmented into individual parking slots where Convolution Neural Network (CNN) is used to predict each individual parking spot as empty or occupied. With Object detection technique, a neural network functions as a car detector which keeps a count of all cars that have been detected.
1.2 Problem Statement
With the growing economy and immigration of people from rural areas to urban areas, the number of vehicles coming in the cities has increased. According to (Musulin, 2017), most drivers get frustrated when trying to look for parking space in major cities in the world. In a survey conducted by IBM, Nairobi City was listed as the eight most difficult city for one to find parking space (Armonk, 2011). This is due to the time that one has to spend trying to look for parking space which is always not guaranteed. The time one spends looking for parking space does not only affect the person alone but it also leads to increased traffic in the city.
Suggestions such as increasing parking fees in cities and malls have been proposed as the solution to end the parking problem but this has not worked (Musulin, 2017). This was seen as a measure to force people to use public transport and thus leave their cars at home, however the proposal was not adopted by the court. Increasing the number of parking spaces in the city could help solve the parking space problem, however as a city develops this measure becomes unsustainable. To solve the parking problem in our cities, computer vision and deep learning technologies can be adopted to provide drivers with information about where to find free parking spaces in the city.
1.3 Research Objectives
The objectives of this research are:
i. To investigate the application of computer vision technologies in parking space detection.
ii. To investigate features of machine learning algorithms such as Region Based Convolution Neural Network, Faster R-CNN and YOLO algorithms on accurate detection of cars and parking spaces.
iii. Design a model for detecting parking spaces based on an optimized machine learning algorithm.
iv. Implement a prototype based on the above model for automated and real time parking spot detection.
v. Evaluate the performance and accuracy of the model using independent data.
1.4 Research Questions
The research questions that guided this study are:
i. How can parking spaces be detected using computer vision and machine learning?
ii. How to distinguish a parking space that has been occupied from one that has not been occupied?
iii. How to distinguish between cars parked in a parking space and other objects.
1.5 Scope
The scope of this study was limited to development of an efficient machine learning model that can detect and distinguish vehicles from other objects in a parking area under different lighting areas and weather conditions.
1.6 Significance of the Study
Application of an efficient machine learning and computer vision in developing an automated parking space will aid:
i. Better planning
This being a real time tool, drivers will be able to check in advance the state of parking spaces the city or malls and decide whether they will leave their vehicles or they can look for parking in areas that the system will show are less busy. City and mall managers will also have data to guide them when increasing or reducing parking spaces.
ii. Better Traffic Control
(Musulin, 2017), says that vehicles that stop to look for parking spaces are the major causes of traffic in the cities. Through having a parking space detection system, drivers will not be moving round the city looking for parking space thus aiding in traffic control.
iii. Improved Revenue Projection
Through the parking space detection system, approximation about the number of vehicles which park in the city can easily and accurately be achieved. This can then be used to project the expected revenue based on the number of cars parking and the time spent by each car. Accurate revenue projection will lead to increased developments in the cities.
Buyers has the right to create
dispute within seven (7) days of purchase for 100% refund request when
you experience issue with the file received.
Dispute can only be created when
you receive a corrupt file, a wrong file or irregularities in the table of
contents and content of the file you received.
ProjectShelve.com shall either
provide the appropriate file within 48hrs or
send refund excluding your bank transaction charges. Term and
Conditions are applied.
Buyers are expected to confirm
that the material you are paying for is available on our website
ProjectShelve.com and you have selected the right material, you have also gone
through the preliminary pages and it interests you before payment. DO NOT MAKE
BANK PAYMENT IF YOUR TOPIC IS NOT ON THE WEBSITE.
In case of payment for a
material not available on ProjectShelve.com, the management of
ProjectShelve.com has the right to keep your money until you send a topic that
is available on our website within 48 hours.
You cannot change topic after
receiving material of the topic you ordered and paid for.
Login To Comment