ABSTRACT
Early detection of oil spills and quick interventions are key elements in reducing the menace caused by oil spills in our environment. In this research work, oil spills classification system based on laser fluorosensor spectra data was modeled and simulated. This is to distinguish oil spills from various backgrounds and to classify oil spills into different products. Artificial Neural Network (ANN) toolbox in Matlab/Simulink with MLP (multi-layer perceptron) based supervised architecture was used for the simulation. Using the data in form of 90-channel spectra as inputs, the ANN presents the analysis and estimation results of oil products and various backgrounds as outputs. The network was trained to understand numerous spectra data of laser fluorosensor for different oil spill products (light oil, medium oil, and heavy oil) and various backgrounds (water, sand and stone). The trained network was tested using data set to the network. A back propagation learning algorithm with an optimizer based on gradient descent method was used during the training of the network. The effect of different number of hidden layers (2, 3, and 4) and the different number of neurons (7,7), (7,7,7), (7,7,7,7), (7,8), (9,7), (8,10) and (10,10) in hidden layers was checked on the overall performance and accuracy of the MLP artificial neural network and the results were compared. It was found that the ANN with MLP based supervised architecture performed well when the number of neurons in hidden layers is the same ((7,7), (7,7,7), (7,7,7,7), (10,10)) and an average of 100% classification result was achieved. The network however performed badly when it was trained with different number of neurons ((7,8), (9,7), (8,10), in the hidden layers and an average of 18.3% classification result was achieved. It was also found that laser fluorosensors must be operated at wavelengths between 308 nm and 340 nm to produce well distinguished fluorescence spectra. Finally, it has been noted from this research work that laser fluorosensors are currently the most effective remote sensors as they can detect oil spills in almost all backgrounds (land, water, ice, snow etc.).
TABLE OF CONTENTS
Title Page i
Declaration ii
Certification iii
Dedication iv
Acknowledgement v
Table of Contents vi
List of Tables xi
List of Figures xii
Symbols and Abbreviations xvi
Abstract xviii
CHAPTER 1: INTRODUCTION
1.1 Background of the Study 1
1.2 Problem Statement 3
1.3 Aim and Objectives 4
1.4 Significance of Study 4
1.5 Scope of the Study 5
1.6 Justification of the Study 5
1.7 Thesis Organization 5
CHAPTER 2: LITERATURE REVIEW
2.1 Historical Background of a Remote Sensing System 7
2.2 Theoretical Background of a Remote Sensing System 10
2.2.1 Remote sensors 13
2.2.1.1 Active remote sensors 14
2.2.1.2 Passive remote sensors 14
2.2.2 Types of remote sensors and their applications 14
2.2.2.1 Laser fluorosensors 15
2.2.2.2 Optical remote sensing 15
2.2.2.3 Microwave sensors 16
2.2.2.4 Visible sensors 17
2.2.2.5 Infrared sensors 17
2.2.2.6 Ultraviolet (UV) sensors 18
2.2.2.7 Radar 18
2.2.3 Airborne technology 19
2.2.4 Spaceborne technology 19
2.2.5 Oil spill detection using laser fluorosensors 20
2.2.6 Oil spill detection and classification methods 22
2.3 Review of related works on remote sensing system to
detect and monitor oil spills 27
2.4 Research gap 38
CHAPTER 3: MATERIALS AND METHODS
3.1 Materials 39
3.2 Methods 39
3.2.1 Artificial neural network model 40
3.2.2 Multi-layer Perceptron (MLP) Model 40
CHAPTER 4: RESULTS AND DISCUSSIONS
4.1 Computer Simulation 54
4.2 Results and Discussion 61
4.2.1 Data Visualization 61
4.2.2 Result obtained when the network was trained with two hidden layers and each hidden layer having seven neurons 65
4.2.3 Result obtained when the network was trained with three hidden layers and each hidden layer having seven neurons 69
4.2.4 Result obtained when the network was trained with four hidden layers and each hidden layer having seven neurons 73
4.2.5 Result obtained when the network was trained with two hidden layers and one of the hidden layers having seven neurons and the other having eight neurons 77
4.2.6 Result obtained when the network was trained with two hidden layers and one of the hidden layers having seven neurons and the other having nine neurons 81
4.2.7 Result obtained when the network was trained with two hidden layers and one of the hidden layers having eight neurons and the other having ten neurons 85
4.2.8 Result obtained when the network was trained with two hidden layers and each hidden layer having ten neurons 89
CHAPTER 5: CONCLUSION AND RECOMMENDATIONS
5.1 Conclusion 94
5.2 Recommendations 95
5.3 Contribution to Knowledge 95
REFERENCES 97
APPENDIX 102
LIST OF TABLES
4.1 Laser fluorosensor wavelength/corresponding relative intensities of all the substances 55
4.2 The target substances and their corresponding target values (actual values) 59
4.3 The network training parameters 60
4.4 Results obtained when the network was trained with two hidden layers and each hidden layer having seven neurons 65
4.5 Results obtained when the network was trained with three hidden layers and each hidden layer having seven neurons 69
4.6 Results obtained when the network was trained with four hidden layers and each hidden layer having seven neurons 734.7 Results obtained when the network was trained with two hidden layers and one of the hidden layers having seven neurons and the other eight neurons 77
4.8 Results obtained when the network was trained with two hidden layers and one of the hidden layers having seven neurons and the other nine neurons 81
4.9 Results obtained when the network was trained with two hidden layers and one of the hidden layers having eight neurons and the other ten neurons 85
4.10 Results obtained when the network was trained with two hidden layers and each hidden layer having ten neurons 89
LIST OF FIGURES
2.1 Region of Electromagnetic Spectrum 11
2.2 Interaction of radiation with earth surface features 13
2.3 Fluorescence spectra of Light 21
2.4 Fluorescence spectra of Background materials 21
2.5 Perceptron model of an artificial neural network 25
2.6 Multi-layer Perceptron model of an artificial neural network 25
2.7 Self-organizing map (SOM) artificial neural network 26
2.8 A typical artificial neural network model 26
3.1 Multi-Layer Perceptron (MLP) model with one input layer, two hidden layers and one output layer 41
3.2 Artificial Neural Network Model (Multi-Layer Perceptron supervised based architecture); with six inputs, two hidden layers and six outputs. 45
3.3 Multi-Layer Perceptron (MLP) implementation block diagram 52
3.4 Multi-Layer Perceptron (MLP) implementation flow chart/classification scheme 53
4.1 Fluorescence spectra of oil products and background materials 61
4.2 Fluorescence spectra of light oil 61
4.3 Fluorescence spectra of medium crude 62
4.4 Fluorescence spectra of water 62
4.5 Fluorescence spectra of heavy oil 62
4.6 Fluorescence spectra of sand 63
4.7 Fluorescence spectra of stone 63
4.8 Performance curve obtained when the network was trained with two hidden layers and each hidden layer having seven neurons 66
4.9 Regression curve obtained when the network was trained with two hidden Layers and each hidden layer having seven neurons 67
4.10 Receiver Operating Characteristics (ROC) curve obtained when the network was trained with two hidden layers and each hidden layer having seven neurons 68
4.11 Performance curve obtained when the network was trained with three hidden layers and each hidden layer having seven neurons 70
4.12 Regression curve obtained when the network was trained with three hidden layers and each hidden layer having seven neurons 71
4.13 Receiver Operating Characteristics (ROC) curve obtained when the network was trained with three hidden layers and each hidden layer having seven neurons 72
4.14 Performance curve obtained when the network was trained with four hidden layers and each hidden layer having seven neurons 74
4.15 Regression curve obtained when the network was trained with four hidden Layers and each hidden layer having seven neurons 75
4.16 Receiver Operating Characteristics (ROC) curve obtained when the network was trained with four hidden layers and each hidden layer having seven neurons 76
4.17 Performance curve obtained when the network was trained with two hidden layers and one of the hidden layers having seven neurons and the other eight neurons 78
4.18 Regression curve obtained when the network was trained with two hidden layers and one of the hidden layers having seven neurons and the other eight neurons 79
4.19 Receiver Operating Characteristics (ROC) curve obtained when the network was trained with two hidden layer sand one of the hidden layers having seven neurons and the other eight neurons 80
4.20 Performance curve obtained when the network was trained with two hidden layers and one of the hidden layers having seven neurons and the other nine neurons 82
4.21 Regression curve obtained when the network was trained with two hidden layers and one of the hidden layers having seven neurons and the other nine neurons 83
4.22 Receiver Operating Characteristics (ROC) curve obtained when the network was trained with two hidden layers and one of the hidden layers having seven neurons and the other nine neurons 84
4.23 Performance curve obtained when the network was trained with two hidden layers and one of the hidden layers having eight neurons and the other ten neurons 86
4.24 Regression curve obtained when the network was trained with two hidden layers and one of the hidden layers having seven eight and the other ten neurons 87
4.25 Receiver Operating Characteristics (ROC) curve obtained when the network was trained with two hidden layers and one of the hidden layers having eight neurons and the other ten neurons 88
4.26 Performance curve obtained when the network was trained with two hidden layers and each having ten neurons 90
4.27 Regression curve obtained when the network was trained with two hidden layers and each having ten neurons 91
4.28 Receiver Operating Characteristics (ROC) curve obtained when the network was trained with two hidden layers and each having ten neurons 92
LIST OF SYMBOLS AND ABBREVIATIONS USED
ABBREVIATION DEFINITION
ANN Artificial Neural Network
AVHRR Advance Very High Resolution Radiometer
CGDA Conjugate Gradient Descent Algorithm
CZCS Coastal Zone Colour Scanner
DSS Decision Support System
FAI Floating Algae Index
GDA Gradient Descent Algorithm
GOCI Geo-stationary Ocean Colour Imager
GPR Grand Penetration Radar
HCMM Heat Capacity Mapping Mission
HDSAR High Definition Synthetic Aperture Radar
IR Infrared
IR/UVLS Infrared/Ultraviolet Line Scan
LF Laser Fluorosensor
MLP Multi-layer Perceptron
MSE Mean Square Error
MWR Microwave Radiometer
NMR Nuclear Magnetic Resonance
NOSDRA National Oil Spills Detection and Response Agency
OSRCAP Oil Spills Response and Contingency Action Plan
PCA Principal Component Analysis
RBF Radial Basic Function
RMSE Root Mean Square Error
ROC Receiver Operating Characteristics
SAR Synthetic Aperture Radar
SLAR Slide Looking Airborne Radar
SLEAF Scanning Laser Environmental Airborne Fluoro-sensors
SOM Self-organizing Map
TIR Thermal Infrared Region
UAS Unnamed Aircraft System
UAV Unnamed Aerial Vehicles
UV Ultraviolet
CHAPTER 1
INTRODUCTION
1.1 BACKGROUND OF THE STUDY
A remote sensing system plays an important role in continuous detection and classification of oil spills on oceans. This has helped the emergency and monitoring team to take a quick and proactive action in order to reduce pollution that happen as a result of oil spills in our environment. In major cases of discharged oil to land and seas, the detection and the classification of the oil spills have critical importance for rapid emergency response activities. Five major issues are to be covered in detection and classification of oil spills namely, (i) Prevention (ii) Alarm, (iii) Monitoring (iv) Damage quantification and (v) deployment of right instruments for cleanup. (Bava et al., 2002). Oil spill is among the major sources of pollution to the sea which may happen deliberately or accidentally. Some ships discharges to seas can be through leakages, ship accident and tank washing residues. Also, fuel oil sludge, engine room wastes and foul bilge water produced by all type of ships also end up in the sea (Konstantinos, 2014).
The effect of the oil spill on the sea ecosystem has claimed the lives of so many aquatic animals. Whatever the source of oil spill is, oil spill pollution will continue to occur, therefore, in order to lessen its effect, the improvement of its detection and continuous monitoring are the most important issues to effectively plan countermeasure responses (Akkartal and Sunar, 2008).
Vessels, airplanes, and satellites are the major classes of tools used to convey the remote sensing systems that are used to detect and classify oil spills. If Vessels are equipped with specialized radars, they can detect oil at sea but they can cover a very limited area. However, if oil sampling is required, the vessel remains necessary. Airplanes and Satellites equipped with Synthetic Aperture Radars (SAR) are the main systems to monitor sea-based oil pollution. SAR is an active microwave sensor, which captures two dimensional images (Konstantinos, 2014).
Nowadays as remote sensing technologies advance, parties that are potentially responsible for oil spill incident can easily be identified and even minor spills can be immediately detected before they cause wide spread damage. For an oil spill detection and decision support system (DSS), remote sensing data can be a very useful input. Visual detection of an oil spill isn't reliable as oil is often confused with other substances, e.g. sea weeds and fish sperm. Moreover, oil on the surface cannot be observed clearly through fog and darkness (Fingas, 2011). This is the reason why remote sensing technologies are very useful in detection and classification of oil spills as they can easily distinguish oil spill from other substances. The following information can also be provided by remote sensors: the location and spread of an oil spill over a large area, the thickness of the spilled oil, the distribution of an oil spill to estimate the quantity of spilled oil, and the classification of the oil type. All these information are necessary in order to estimate environmental damage, take appropriate response activities, and to assist in clean-up operations (Maya, et al., 2014).
Many authors have reviewed remote sensing technologies for oil spill surveillance. It has been noted by Goodman (1994) that operational use of remote sensing for oil spill contingency planning is limited, however, simple systems such as infrared (IR) and radar have been used to some extent in responding to oil spills. It has also been noted by (Brown et al., 2011) that no single sensor can give all the information required for oil spill contingency planning. A large number of remote sensors are currently available for oil spill surveillance and due to this; there is need for a comprehensive overview and comparison of the existing sensors remote sensors. A better knowledge of the oil spill surveillance sensor characteristics will help improve the effectiveness and operational use of these sensors for oil spill response and contingency planning.
Classification of oil spills into various oil types is important because it helps the response team to know the type of instrument to be deployed during clean-up processes. Oil spills are often classify into light, medium and heavy oil because the instruments use to clean up all light oils are the same and the instruments used to clean up all medium crudes are the same and also the instruments used to clean up all heavy oils are the same. Since the major reason for oil spill classification is to know the type of instruments to be used during clean up, it is therefore reasonable to limit the classification of oil spills to light, medium and heavy oils.
In this research work, the remote sensing system that will be modeled and be simulated will base its oil spill detection and classification on light oil, medium crude, heavy oil, water, and sand spectra data of laser fluorosensor.
Laser Fluorosensors have been noted recently to be the most effective remote sensors as they can detect oil under the water surface and on various backgrounds including snow or ice (Brown et al., 2014).
1.2 PROBLEM STATEMENT
The slow response and intervention by the oil spill monitoring team in Nigerian over the years is due to the fact that oil spills are often detected very late and also the difficulty in making decision on the type of instruments to be deployed during clean-up. This has led to death of many aquatic animals, reduction in the yield level of farm lands, and reduction in the quality of natural water available for human beings.
Therefore, distinguishing oil spills from various backgrounds and classifying oil spills into different products will go a long way in addressing the problem of late detection of oil spills and the slow decision making on the instruments to be deployed during clean-up.
1.3 AIM AND OBJECTIVES
The aim of this research work is to analyse and simulate a remote sensing system for detection and classification of oil spills using laser fluorosensor.
Objectives:
i. To review existing literature on remote sensing technologies for oil spill detection and classification
ii. To develop a scheme for oil spill classification system using laser fluorosensor spectra data
iii. To develop a mathematical model of an oil spill classification system using laser fluorosensor spectra data
iv. To simulate an oil spill classification system using laser fluorosensor spectra data
v. To validate results.
1.4 SIGNIFICANCE OF STUDY
This research work will help the oil spill monitoring team to respond fast to oil spills incidences, hence, preventing the spreading of oil spills and the damages that it can cause to our environment. The research work will also help the oil producing companies and communities to avoid the pollution caused by oil spills in their environments to a large extent. Finally, the research work will help Nigerian government to cut down the amount budgeted annually for oil spills clean up in the affected oil producing communities in Nigeria.
1.5 SCOPE OF THE STUDY
This research work covers the modeling and the simulation of an oil spill classification system that will distinguish oil spills from other backgrounds and also classify oil spills into different products.
1.6 JUSTIFICATION OF THE STUDY
The purpose of this research work is to analyse and simulate a remote sensing system that will detect oil spills and classify them into different products. Many authors have researched on this topic and have used MLP neural network to simulate an oil spills classification system. Their work did not cover the aspect of training a multi-layer perceptron (MLP) artificial neural network with laser fluorosensor spectra data. None of them has also checked the effect of different number of hidden layers and different number of neurons in hidden layers on the performance and accuracy of the MLP network.
In this research work, a classification system was therefore simulated by training an MLP neural network with laser fluorosensor spectra data. The effect of different number of hidden layers and different number of neurons in hidden layers was also checked on the performance and accuracy of the network.
1.7 THESIS ORGANISATION
Chapter 1: This is the introduction of the report that discussed the background study, problem statement, aim and objectives, project scope, justification of the study and the significance of the project.
Chapter 2: This gives a well detailed literature review and some of the related works done by other authors.
Chapter 3: This explained in detail the materials and the method used to achieve the proposed research work.
Chapter 4: This chapter is all about result and discussion; the result of the modeling and the simulation done in this research work was discussed extensively in this chapter.
Chapter 5: This chapter gives the summary of what was achieved in this research work and some recommendations that were made.
Click “DOWNLOAD NOW” below to get the complete Projects
FOR QUICK HELP CHAT WITH US NOW!
+(234) 0814 780 1594
Login To Comment