Project topics and materials
Get Complete Project Material File(s) Now! »
CHAPTER 3 – RESEARCH METHODOLOGY
INTRODUCTION
This study made use of an extended literature study and empirical research to solve a specific problem. The torment to orthorectify satellite imagery when there are limited GCPs available that are irregularly distributed, encouraged the investigation of this problem and develop a methodological approach to follow for improving the geometric accuracy of VHR satellite imagery when there is a lack of quality GCPs available. In addition to the literature study (Chapter 2) that was conducted to provide insight into the evolution and modernisation of satellites, how they operate, their characteristics, inherent distortions and the role auxiliary data play during the process of rectification, the empirical research was conducted to answer the research questions identified for this study. According to Niaz (2008), the research problem that needs to be resolved will determine the research methodology to utilise. Research methods (quantitative and qualitative approaches) describe the research strategy and empirical techniques used to resolve specific research problems. Both research approaches are used to devise, investigate and resolve research problems. However, the nature of reality, knowledge and the principles that inspire scientific research, will lead to the preferred and specific research method to be used. This is normally based on the relevance of this method in the specific methodological orientation (Onwuegbuzie et al., 2009).
In this chapter, a brief statement of the research problem and motivation, the research questions and hypotheses and the aim and objectives are included, as it was already discussed in detail in Chapter 1. Thereafter, a detailed breakdown is provided of the methodology and methods that this research is based on.
PROBLEM STATEMENT AND MOTIVATION
Motivation for the research study
It is inevitable that remote sensed imagery will inherit geometric distortions during data capturing, due to many influential factors that affect the positional accuracy of satellite imagery. Factors such as acquisition geometry, topographic properties of the image area, optical fidelity of the sensor and positional steadiness all play a vital role in the extent of geometric errors imbedded in remotely sensed imagery (Exelis VIS, 2013). Orthorectification is the process that eliminates the geometric distortions introduced during image acquisition. It produces a planimetric image that has a consistent image scale and is accurately registered to real-world map projections and ground coordinate systems.
Traditionally, orthorectification was a semi-automated process that required user inputs regarding the sensor platform, GCPs and terrain elevation to process the image data accurately using commercial image processing software. However, recently with the development of newly designed sensor systems this traditional approach has changed dramatically (Petrat and Eloff, 2014; Hoja et al., 2008). Automated orthorectification of imagery is now possible based on the comprehensive metadata embedded in remotely sensed data and utilising new and improved sensor models and algorithms to process the image data. As was mentioned in Chapter 2 (Paragraph 2.2.4), the modernisation of satellite systems brought about a new and improved dimension to the pointing accuracies of current and future generations of satellite systems (Petrat and Eloff, 2014). These days, orthorectification are more and more performed by using RPCs, elevation data and optional GCPs to achieve highly accurate ortho-images, due to the fact that not all 3rd party image processing software have extended sensor model libraries to include all rigorous sensor models (Dial and Grodecki, 2005; Toutin, 2006). As were discussed in Chapter 2 (Paragraph 2.3.2), this method of using RPCs (non-parametric approach) are simpler empirical mathematical models compared to using rigorous sensor models (parametric approach) with complicated mathematical modelling (Dial and Grodecki, 2005). The non-parametric approach is usually followed due to the lack of suitable auxiliary data such as the non-availability of sensor specific parameters. However, when highly accurate ortho-images are required and auxiliary data are readily available, then the use of rigorous sensor models will be the most suitable option. Most ortho-image applications require very high registration accuracy. For instance, a registration error of less than 1/5 of a pixel will produce a change detection error of less than 10% and for measurement accuracies of less than 1 m (e.g. measurements of ice flow and cosmic ground deformation) even better registration accuracies are required (Leprince et al., 2007). In practice, the acquisition of raw image data with detailed sensor information and sufficient elevation data to achieve high registration accuracies are not problematic.
However, the collection of ground control points poses a significant problem when performing single frame orthorectification, as were discussed in Chapter 1 (Paragraph 1.2.1). In such cases, the only available source for extracting GCPs is vector layers. Therefore, the following question arises: How accurate will an ortho-image be when GCPs are used that were extracted from a vector layer? Various orthorectification experiments were conducted during this study to determine the effect of such GCPs that are irregularly distributed, covering an entire image scene.
Research hypotheses and questions
In Chapter 1 (Paragraph 1.2.2), a detailed description of the research hypotheses and questions were formulated. The specific experiments conducted during the empirical research (Chapter 4) enabled the testing of the research hypotheses and answer the research questions which were formulated. The experiments conducted during stages 1, 2 and 3 tested the research hypotheses and answered research questions 1, 2 and 3 (these are discussed in detail below in Paragraphs 3.7.1 – 3.7.3). To answer research questions 4 and 5, two separate independent orthorectification experiments were conducted during stage 2. Firstly, utilising the TerraSAR-X-based GCPs acquired from Airbus Defence and Space and the 2 m DTM to create an ortho-image and determine if it is possible to create accurate ortho-images without manually collecting GCPs. Secondly, create an ortho-image by utilising only the geometric sensor model and an elevation source (i.e. 2 m DTM) – without the use of GCPs – and determine if an accurate ortho-image can be produced.
RESEARCH AIM AND OBJECTIVES
The aim of this study was to investigate and compare the positional accuracies of orthoimages under various orthorectification scenarios and provide improved geometric accuracies of VHR satellite imagery when diverse ground control and elevation data sources are available. Considering the aim, a methodological approach was developed for improving the geometrical accuracy of VHR imagery when there are inadequate GCPs available that are irregularly distributed in an entire image scene. The parametric approach was followed to conduct all orthorectification experiments. Sufficient auxiliary data were available to create a highly accurate ortho-image (e.g. master image), which were used to measure the accuracy of ortho-images created by utilising GCPs extracted from a vector layer. In order to achieve the aim of this study, specific objectives were formulated (Chapter 1, Paragraph 1.3). Achieving the objectives and ultimately the aim of this study contributed to the formulation of the methodological approach (see Paragraph 4.9, Chapter 4), which are described in terms of the procedure to follow, with specific reference to the:
a) number of GCPs necessary;
b) distribution and placement of GCPs; and
c) effect of the elevation data and quality DEM necessary.
This approach highlighted the precise optimum data and reference sources necessary to use when performing orthorectification on VHR satellite imagery. It indicated expected location accuracy limitations when:
a) utilising various quality elevation data sources and
b) using a limited number of GCPs that are irregularly distributed.
RESEARCH METHODOLOGY
Methodology is defined as the rules, principles and formal conditions that govern scientific research for the purpose of organising and broadening ones knowledge of the phenomena that is being researched (Gelo et al., 2008). Gelo et al. (2008) specifically refers to methodologies as the relationship between the researcher’s views, theory, research questions, hypotheses and research methods. Two types of research approaches exist, namely quantitative and qualitative approaches. Quantitative approaches dominated science research until the 1960’s. Since the 1960’s,researchers
started to criticise the use of pure quantitative approaches and proposed a naturalistic, contextual and holistic method – this came to known as qualitative research (Gelo et al., 2008). Quantitative research (also known as traditional or experimental approaches) is a positivistic research paradigm that promotes the status of experimental research and the quantitative methods of analysis (Cohen et al., 2004; Creswell, 2012; Doll, 1970). Gelo et al. (2008, p 267) cites that quantitative research describes the “how much of an entity there is”, which means that quantitative research consists of calculating the frequency of events and the volume or the size of associations between variables (Maree, 2007).
Controversially, qualitative research is based in the post–positivism that promotes powerful descriptions of the phenomena being investigated (Gelo et al., 2008). Gelo et al. (2008, p 267) state that qualitative research implies “describing the constituent properties of an entity”, meaning that qualitative research focus rather on the clarification of phenomena. Addition to the foregoing differences between quantitative and qualitative research, both these approaches also differ regarding methodological assumptions and research methods:
a) Quantitative research: according to Gelo et al. (2008), quantitative research requires a nomographical methodology. Nomography refers to the science of common law that manage generalisation. This means that facts are collected, confirmed and processed with the purpose to generalise. The methods to collect quantitative data are directly from the source of the data (primary data) or indirectly by using personal or official documents and archive material (secondary data). Qualitative data collection methods can be used to collect quantitative data, as long as the data are statistically analysed by awarding numerical values to the collected data. Therefore, quantitative data collection requires redirecting phenomena to numerical values for statistical analysis, while qualitative data collection constitutes non-numerical representations (texts, pictures, photos, videos, etc.). Quantitative research stresses meanings, concepts, characteristics, metaphors, symbols and descriptions of phenomena (Berg, 2004). The researcher’s role during quantitative data collection is one of objectivity and is limited to the collection of data for confirming research questions and hypotheses and focuses on the validity of what is being observed (Johnson and Christensen, 2008).
b) Qualitative research: ideology (qualitative research), in contrast, refers to the complete representation of a particular event with the aim to record and develop an understanding of the event. Qualitative data collection also entails the collection of primary and secondary data, but uses different methods to collect data (Gelo et al., 2008). According to Howe (2003), the procedure for qualitative data collection is not as strictly defined as with quantitative data collection. The range is boundless and results are provided in descriptive or narrative form. Quantitative and qualitative research methods differ in several respects and Dreyer (1998) warns that “Whether one conducts quantitative or qualitative research, one has to be both insider and outsider, engaged participant and detached observer.”
The difference between the two approaches is also evident in the dichotomy descriptive and concept formation. Quantitative approaches tend to be more descriptive, in that phenomena and their relationships are described, to confirm predictions made by theory. Qualitative approaches, in contrast, refer to concept formation, meaning personal perspectives, experiences and understandings of phenomena. Therefore, quantitative and qualitative research approaches do not need to be mutually exclusive. The one approach can complement the other. Some researchers even prefer to combine both research methods, which are known as mixed method research (Bergman, 2008; Strydom, 2009). According to Denzin as cited in (Keeves, 1988), reality is better understood when mixed research methods are used and it is therefore deemed as the ideal research approach to interpret reality. For this study, mixed research methods were utilised to collect and analyse data.
DECLARATION
SUMMARY OF RESEARCH
ACKNOWLEDGEMENTS
LIST OF FIGURES
LIST OF TABLES
LIST OF EQUATIONS
LIST OF ACRONYMS
CHAPTER 1 – INTRODUCTION AND PROBLEM STATEMENT
1.1 INTRODUCTION
1.2 PROBLEM STATEMENT AND MOTIVATION
1.2.1 Motivation for the research study
1.2.2 Research hypotheses and questions
1.3 RESEARCH AIM AND OBJECTIVES
1.4 RESEARCH METHODOLOGY
1.4.1 Literature study
1.4.2 Empirical research
1.4.2.1 Data collection
1.4.2.2 Data analysis
1.4.2.3 Research overview
1.5 ETHICAL ASPECTS
1.6 CONTRIBUTION OF THE STUDY
1.7 CHAPTER SUMMARY
CHAPTER 2 – LITERATURE STUDY: THE ART AND SCIENCE OF CONDUCTING ORTHORECTIFICATION WITH AN OVERVIEW ON REMOTE SENSING (RS)
2.1 INTRODUCTION
2.2 REMOTE SENSING PLATFORMS
2.2.1 Evolution of satellite platforms
2.2.2 Properties of satellite systems
2.2.2.1 Regions of the electromagnetic spectrum
2.2.2.2 Converting recorded digital data into images
2.2.3 Characteristics of satellite systems
2.2.4 Modernisation of optical satellite systems
2.2.5 Definition of resolution capabilities
2.2.6 Satellite image distortions
2.2.6.1 Radiometric distortions
2.2.6.2 Geometric distortions
2.3 WAYS AND MEANS TO IMPROVE THE GEOMETRICAL ACCURACY OF SATELLITE IMAGERY
2.3.1 The application of image processing systems
2.3.2 Geometric correction methods
2.3.2.1 Parametric approach
2.3.2.2 Non-parametric approach
2.4 THE REQUIREMENTS AND ACQUISITION OF GCPS AND DEMS To CREATE ACCURATE ORTHO-IMAGES.
2.4.1 The application of GCPs
2.4.2 The application of DEMs
2.5 ASSESSMENT OF AN ORTHO-IMAGE
2.5.1 Calculating RMSE
2.5.2 Utilising the ERDAS IMAGINE® 2015 MAA tool
2.5.3 Performing visual inspections
2.6 CHAPTER SUMMARY
CHAPTER 3 – RESEARCH METHODOLOGY
3.1 INTRODUCTION
3.2 PROBLEM STATEMENT AND MOTIVATION
3.3 RESEARCH AIM AND OBJECTIVES
3.4 RESEARCH METHODOLOGY
3.5 RESEARCH DESIGN
3.6 DATA ACQUISITION AND COLLECTION
3.7 DESIGN OF EXPERIMENTS
3.8 DATA ANALYSIS
3.9 CHAPTER SUMMARY
CHAPTER 4 – EMPIRICAL RESEARCH: ORTHORECTIFICATION EXPERIMENTS AND METHODOLOGICAL CONTRIBUTION
4.1 INTRODUCTION
4.2 PROCESS FOLLOWED TO PERFORM ORTHORECTIFICATION
4.3 ORTHORECTIFICATION TESTS: STAGE 1 EXPERIMENTS
4.4 ANALYSIS OF STAGE 1 EXPERIMENTS
4.5 ORTHORECTIFICATION TESTS: STAGE 2 EXPERIMENTS
4.6 ANALYSIS OF STAGE 2 EXPERIMENTS
4.7 ORTHORECTIFICATION TESTS: STAGE 3 EXPERIMENTS
4.8 ANALYSIS OF STAGE 3 EXPERIMENTS
4.9 STUDY CONTRIBUTIONS
4.10 FINAL OUTCOME
4.11 REALISING THE NEED FOR DEVELOPING AN AUTOMATIC GCP EXTRACTION SCRIPT (A-GCP-ES)
4.12 CHAPTER SUMMARY
CHAPTER 5 – DEVELOPMENT OF AN AUTOMATIC GCP EXTRACTION SCRIPT
5.1 INTRODUCTION
5.2 MEASURES FOLLOWED TO VERIFY PLACEMENT OF GCPS
5.3 BACKGROUND CONCEPT OF THE A-GCP-ES
5.4 DEVELOPMENT OF THE A-GCP-ES
5.5 TESTING AND EVALUATING THE A-GCP-ES
5.6 CHAPTER SUMMARY
CHAPTER 6 – CONCLUSIONS AND RECOMMENDATIONS
6.1 INTRODUCTION
6.2 STUDY REVIEW
6.3 ACHIEVEMENT OF STUDY OBJECTIVES AND AIM
6.4 CONCLUSIONS
6.5 RECOMMENDATIONS
6.6 FUTURE RESEARCH
6.7 CHAPTER SUMMARY
REFERENCES
APPENDICES
GET THE COMPLETE PROJECT
