Structure from Motion in the Geosciences

Jonathan L. Carrivick, Mark W. Smith, Duncan J. Quincey

• Publisher: Wiley-Blackwell
• Year: 2016
• Print ISBN: 9781118895849
• eBook ISBN: 9781118895825
• Edition: 1
• Topic: Tecnologia e ingegneria
• Subtopic: Ingegneria civile

Structure from Motion with Multi View Stereo provides hyperscale landform models using images acquired from standard compact cameras and a network of ground control points. The technique is not limited in temporal frequency and can provide point cloud data comparable in density and accuracy to those generated by terrestrial and airborne laser scanning at a fraction of the cost. It therefore offers exciting opportunities to characterise surface topography in unprecedented detail and, with multi-temporal data, to detect elevation, position and volumetric changes that are symptomatic of earth surface processes. This book firstly places Structure from Motion in the context of other digital surveying methods and details the Structure from Motion workflow including available software packages and assessments of uncertainty and accuracy. It then critically reviews current usage of Structure from Motion in the geosciences, provides a synthesis of recent validation studies and looks to the future by highlighting opportunities arising from developments in allied disciplines. This book will appeal to academics, students and industry professionals because it balances technical knowledge of the Structure from Motion workflow with practical guidelines for image acquisition, image processing and data quality assessment and includes case studies that have been contributed by experts from around the world.

1
Introduction to Structure from Motion in the Geosciences

Abstract

Structure from Motion (SfM) is a topographic survey technique that has emerged from advances in computer vision and traditional photogrammetry. It can produce high-quality, dense, three-dimensional (3D) point clouds of a landform for minimal financial cost. As a topographic survey technique, SfM has only been applied to the geosciences relatively recently. Its flexibility, particularly in terms of the range of scales it can be applied to, makes it well suited to a field as diverse as the geosciences. This book is designed to act as a primer for scientists and environmental consultants working within the geosciences who are interested in using SfM or are seeking to understand more about the technique and its limitations. The early chapters consider SfM as a method within the context of other digital surveying techniques, and detail the SfM workflow, from both theoretical and practical standpoints. Later chapters focus on data quality and how to measure it using independent validation before looking in depth at the range of studies that have used SfM for geoscience applications to date. This book concludes with an outward look towards where the greatest areas of potential development are for SfM, summarising the main outstanding areas of research.

Keywords

geosciences; Structure from Motion; multi view stereo; GIS; landform

1.1 The Geosciences and Related Disciplines

Geoscience is a term that encompasses many disciplines of research and industry, particularly environmental consultancy. It is an umbrella term for climate, water and biogeochemical cycles, and planetary tectonics, which are the three basic processes that shape the Earth’s surface. These are complex natural systems in space and time. For example, process responses and interactions occur on spatial scales spanning hundreds of kilometres to microns, such as river catchments and abrasion marks on fluvially-transported grains, respectively. Process responses and interactions occur on timescales ranging from picoseconds for chemical reactions to millions of years for plate tectonics and biological evolution, respectively.

Whether academic or applied, whether large scale or small scale, the geosciences seek to understand the forces and factors that shape our world and the environments in which we live. Reasons for requiring understanding of these forces and factors span many remits: exploitation for the hydrocarbon and renewable energy sectors, managing natural hazards, managing a resource-consuming and dynamic society, mitigating effects of climate change, and academic interest and enquiry, for example.

In seeking ever-refined understanding for application to real-world problems, the geosciences now transcend “traditional” earth science disciplines (Fig. 1.1). The multidisciplinary nature of the geosciences is partly due to it having become particularly adept at pursuing interactions between the biological, chemical, and physical sciences. Analysis across these traditional boundaries is critical to understanding systems in an integrated and holistic manner.

Figure 1.1 A word cloud of geoscience sub-disciplines. The font size of each word does not indicate anything.

Furthermore, the geosciences are now established as being notable for embracing emerging and novel technologies and innovations. Indeed the revolution brought about by spatial analysis software such as geographical information systems (GIS) has been argued as a new paradigm in the discipline. Many technologies in the geosciences have been adapted from the military, from the petroleum industries, and more recently from computer science.

No matter what particular specialism to which they affiliate themselves, many geoscience disciplines will generally recognise three key tasks:

1. Recognition of spatial patterns
2. Documentation of transient landforms
3. Linking processes to products

Here it is important to note that in this book we use the term landform independent of any scale; the term as used in this book is considered to encompass landscape, terrain, feature, surface, and texture, for example.

A common requirement for each of these three key tasks is for the geosciences to have topographic information. The primary function of topographic information in digital format is to quantify landform variability and more specifically three-dimensional (3D) structure.

Topographic information can be used to identify landforms and landform properties. Landforms can include natural and artificial features and thus form part of the description of a specific place. When landforms are observed to change, the processes causing those changes are often inferred conceptually, and perhaps also tested by numerical models. New methods of acquiring topographic data with a fine spatial resolution are to be welcomed because they expose greater detail about landform morphology. They also provide an opportunity to match the scale of topographic data with the spatiotemporal scale of the landform or processes under investigation.

On the basis of the multidisciplinary nature of the geosciences and of the widespread academic and applied need for topographic survey data, we consider that this book, which will focus specifically on one specific method for generating topographic data, has relevance for all the geosciences (Fig. 1.1) and for related disciplines. Related disciplines requiring topographic information include architecture, archaeology, civil engineering and subdisciplines associated with built structures, objects, and artefacts, and biology and medicine where concerns range from vegetation to anatomical surveys.

1.2 Aim and Scope of this Book

The aim of this book is to describe an emerging survey method and workflow that is better established in related disciplines such as archaeology (e.g. De Reu et al. 2013) and cultural heritage (Koutsoudis et al. 2014) and is now finding widespread uptake in the geosciences, namely, “Structure from Motion” (SfM). This book is designed to act as a primer for geoscientists who are interested in using SfM or are seeking to understand more about the technique and its limitations.

This book is designed to appeal to students, professional academics, and industry practitioners, particularly environmental consultants. Whilst existing texts dealing with SfM are often heavily mathematical, originating commonly from computer vision literature, this book is designed to be fully accessible by an interested geoscience audience that may not necessarily be fully conversant in complex mathematical operations involved in SfM. Thus, the workflow of SfM is described in a predominantly qualitative manner, and the reader is referred elsewhere for further technical details. Important terms are in bold at first use, a list of abbreviations is provided, and emphases of particularly important properties are in italics. This book is designed to balance the conceptual discussion of application, theory, and technical details of analytical methods. Thereby this book serves as a synoptic reference to both inform and educate. In educating, we emphasise the discussion and development of a critical understanding of the application of SfM in the geosciences to date. In terms of informing, we build on this critical understanding to stimulate ideas for carefully considered future developments of the SfM workflow by the geosciences.

1.3 The Time and the Place

This book is timely and of immediate relevance because of (i) the emergence of an affordable, user-friendly software; (ii) rapid developments in unmanned aerial vehicles (UAVs) or drones and other potential SfM survey platforms; and (iii) a dearth of textbooks on SfM in the geosciences. Notwithstanding that, of course, the pace of technological change in hardware and software is incredibly rapid, and for that reason the forward-looking chapters of this book do not dwell on specific hypothetical applications but rather on major themes and concepts.

At present, the use of SfM can only really be evaluated in academic literature since technical and industry reports tend not to be listed on public databases. A search in the academic publications database Web of Knowledge for Structure from Motion (made in April 2015) delivered approximately 1000 records since the early 1980s (Fig. 1.2). Computer science was the category with the most counts of that phrase. Engineering was ranked 2nd and geosciences was ranked 9th. Notably, the geosciences have only started producing publications incorporating SfM in the past decade (Fig. 1.2).

Figure 1.2 Count of citations to “Structure from Motion” in the academic literature.

The impact of SfM is arguably going to be greater than that associated with the advent of airborne laser scanning (ALS) or airborne light detection and ranging (LiDAR), not least because SfM workflows democratise data collection and the development of fine-resolution 3D models at all scales of landscapes, landforms, surfaces, and textures. Moreover, to produce such advanced data products, very little input data are required: as little as a photograph set fro...

Table of contents

• Cover
• Title Page
• Table of Contents
• Abbreviations
• About the Companion Website
• 1 Introduction to Structure from Motion in the Geosciences
• 2 The Place of Structure from Motion
• 3 Background to Structure from Motion
• 4 Structure from Motion in Practice
• 5 Quality Assessment
• 6 Current Applications of Structure from Motion in the Geosciences
• 7 Developing Structure from Motion in the Geosciences
• 8 Concluding Recommendations
• Index
• End User License Agreement