Remote Sensing of Geomorphology

,

• Publisher: Elsevier
• Year: 2020
• ISBN:
  9780444641786
• Topic: Physical Sciences
• Subtopic: Geology & Earth Sciences

Remote Sensing of Geomorphology, Volume 23, discusses the new range of remote-sensing techniques (lidar, structure from motion photogrammetry, advanced satellite platforms) that has led to a dramatic increase in terrain information, and as such provided new opportunities for a better understanding of surface morphology and related Earth surface processes. As several papers have been published (including paper reviews and special issues) on this topic, this book summarizes the major advances in remote sensing techniques for the analysis of Earth surface morphology and processes, also highlighting future challenges. Useful for MSc and PhD students, this book is also ideal for any scientists that want to have a single volume guideline to help them develop new ideas. In addition, technicians and private and public sectors working on remote sensing will find the information useful to their initiatives.

• Provides a useful guideline for MSc and PhD students, scientists, technicians, and land planners on the use of remote sensing in geomorphology
• Includes applications on specific case studies that highlight issues and benefits of one technique compared to others
• Presents future trends in remote sensing and geomorphology

Chapter 1

Structure from motion photogrammetric technique

Anette Eltner^a; Giulia Sofia^{b a} Institute of Photogrammetry and Remote Sensing, Technische Universität Dresden, Dresden, Germany
^b Department of Civil and Environmental Engineering, University of Connecticut, Storrs, CT, United States

Abstract

Structure from motion (SfM) with multiview stereo, a technique from photogrammetry and computer vision that uses overlapping images to reconstruct 3D surface models, is a valuable research tool in geomorphology and related disciplines. Images can be collected with standard consumer-grade cameras, making SfM a low-cost tool that compliments other 3D technologies, such as terrestrial and airborne laser scanning (lidar). The high level of automation of SfM processing offers an unprecedented occasion to describe earth surface processes, but this comes with strengths and challenges. Accordingly, this contribution seeks to give a guide in successfully applying SfM for a range of geomorphic studies. First, it offers an overview of the technique, history, evolution, and the reason behind its success. Second, it describes the method, with guidelines about suitable settings, accuracy, and georeferencing. Finally, including case studies that have been contributed by experts from around the world, it showcases the chances offered to reconstruct processes across spatial and temporal scales.

Keywords

Structure from motion; Multiview stereo; Remote sensing; Close-range photogrammetry; Smartphone technology; Low-cost surveying systems; Surface morphology; Time-lapse; Referencing

1 Introduction

Structure from motion (SfM) photogrammetry provides hyper-scale three-dimensional (3D) landform models using overlapping images acquired from different perspectives with standard compact cameras (including smartphone cameras) and geo-referencing information. As applied to the remote sensing of geomorphology, it is not so much a single technique, but rather a workflow employing multiple algorithms developed from computer vision, traditional photogrammetry, and more conventional survey techniques (Carrivick et al., 2016). Recent literature has provided reviews on the importance of SfM in geosciences (Carrivick et al., 2016; Eltner et al., 2016; Smith et al., 2016) or specific scientific contexts (Mancini et al., 2013; Dietrich, 2016; Entwistle et al., 2018). This contribution builds on the existing literature, to provide a showcase of the technology, relevant to the remote sensing of geomorphology.

1.1 Brief historical summary and state of the art

The roots of SfM lie in two key fields: photogrammetry and computer vision. When techniques from these fields are combined with both automation and precision, the result is a comprehensive tool (Pierrot-Deseilligny and Clery, 2011) for geomorphological applications. Photogrammetry is a relatively old technique (Slama et al., 1980). In this field, the reconstruction efforts of pioneers in the 1840s initially attempted using a pair of ground cameras separated by a fixed baseline and followed by applications using cameras for estimating the shape of the terrain from ground and aerial photographs (Maybank, 1993). With the introduction of aeroplanes and space photography, the development of photogrammetry flourished, with 2D photographs used to rectify images into appropriate coordinates, or mosaicking multiple frames to estimate structures or ground elevation. In a parallel effort, the computer vision community provided the first early algorithms for 3D scene reconstructions by stereo images (Marr and Poggio, 1976) or to pioneer work on motion-based reconstruction (Ullman, 1979).

The prime formalisms derived in these two communities provided the most important foundational theory for the SfM community. However, advances in SfM have been spurred mostly due to the wide range of modern applications. A search in the academic publications database Web of Sciences (WoS) for Structure from Motion (made in August 2018) delivered > 3000 records since the early 1980s (Fig. 1), covering as many as 125 fields of study.

Fig. 1 “Structure from motion” search in academic databases: first 25 results and number of records per discipline (as of August 2018).

Computer science and artificial intelligence is the category with the most counts of that phrase. Engineering is ranked second, remote sensing is fourth, and geosciences is currently ranked sixth. This wide range of applications of SfM results in research with different goals, hence emphasizing multiple ways of addressing SfM problems in space and time. The computer vision field features much older publications than other fields, with the first papers published in the 1980s (Bolles et al., 1987) introducing a technique for building a 3D description of a static scene from a dense sequence of images, and the latest (Zhu et al., 2018) discussing new methods for bundle adjustment (the optimization method needed to simultaneously retrieve the image pose parameters from overlapping images considering corresponding image points). Notably, the geosciences have only started producing publications incorporating SfM photogrammetry in the past decade, but with improvements in the technique moving at an incredible speed: note that a similar search in 2015 by Carrivick et al. (2016) ranked Geosciences in the ninth position. In this field, the first work was published (according to WoS) by Heimsath and Farid (2002). Here, results from three unconstrained photographs characterized hillslope topography, and yield to an estimated surface with errors of the order of 1 m. In comparison, one of the last papers published in the field at the time of the search (Smith and Warburton, 2018) illustrates that topographic data from SfM photogrammetry (with errors on the scale of < 1 mm) inherits enough information to analyze the relationship between geomorphological process and form, at the microscale (few millimeters).

These few examples show an evolution of SfM photogrammetry in time and topics. In the computer vision field, the emphasis remains on methods for obtaining information from images, whereas the evolution of SfM photogrammetry is different in geosciences. Early SfM photogrammetry studies in geosciences emphasized the accuracy of reconstruction, whereas modern geosciences applications focus more on the information that can be retrieved from such analyses.

1.2 Reasons for success in geomorphological surveys

For geomorphological studies, the availability of a high-resolution topographic dataset is fundamental, particularly so for those systems characterized by a complex morphology. We find four main reasons for the success of SfM photogrammetry in geomorphology: (i) spatial accuracy and temporal frequency, (ii) cost; (iii) speed and ease of use. A further reason for SfM's success, although it is still in its exploratory phase, is and (iv) the possibility of involving citizens in science. These points are intrinsically interrelated and build on each other to determine the success of the technique in geosciences.

In geosciences, SfM photogrammetry is a workflow that is virtually independent of spatial scale (Carrivick et al., 2016), it allows potentially unlimited temporal frequency (Carrivick et al., 2016) 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 (Westoby et al., 2012). It offers therefore exciting opportunities to characterize surface topography in unprecedented detail, allowing workers to detect elevation, position, and volumetric or areal changes that are symptomatic of earth surface processes across spatial (see Section 3) and temporal (see Section 4) scales.

When speaking about the costs of a SfM photogrammetry application, they can vary depending on sensors, survey design, and ground control points (GCPs)—when present. SfM photogrammetry sensors are based on consumer-grade cameras, or even smartphones (Micheletti et al., 2014; Prosdocimi et al., 2016; Sofia et al., 2017), which can be handheld or mounted on UAV systems. The sensors, mounting systems or cameras can vary substantially in price and complexity, but the trade-offs between these and the quali...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• Contributors
• Foreword
• Introduction to remote sensing of geomorphology
• Chapter 1: Structure from motion photogrammetric technique
• Chapter 2: Topo-bathymetric airborne LiDAR for fluvial-geomorphology analysis
• Chapter 3: Ground-based remote sensing of the shallow subsurface: Geophysical methods for environmental applications
• Chapter 4: Topographic data from satellites
• Chapter 5: Linking life and landscape with remote sensing
• Chapter 6: SfM photogrammetry for GeoArchaeology
• Chapter 7: Landslide analysis using laser scanners
• Chapter 8: Terrestrial laser scanner applied to fluvial geomorphology
• Chapter 9: Remote sensing for the analysis of anthropogenic geomorphology: Potential responses to sediment dynamics in the agricultural landscapes
• Chapter 10: Using UAV and LiDAR data for gully geomorphic changes monitoring
• Chapter 11: Zero to a trillion: Advancing Earth surface process studies with open access to high-resolution topography
• Chapter 12: Reproducible topographic analysis
• Index