Remote Sensing of Landscapes with Spectral Images
Spectral images, especially those from satellites such as Landsat, are used worldwide for many purposes, ranging from monitoring environmental changes and evaluating natural resources to military operations. In a significant departure from standard remote-sensing texts, this book describes
how to process and interpret spectral images using physical models to bridge the gap between the engineering and theoretical sides of remote sensing and the world that we encounter when we put on our boots and venture outdoors.
Remote Sensing of Landscapes with Spectral Images is designed as a textbook and reference for graduate students and professionals in a variety of disciplines including ecology, forestry, geology, geography, urban planning, archeology, and civil engineering, who want to use spectral images to help solve problems in the field. The emphasis is on the practical use of images rather than on theory and mathematical derivations, although a knowledge of college-level physics is assumed. Examples are drawn from a variety of landscapes and interpretations are tested against the reality seen on the ground. The reader is led through analysis of real images (using figures and explanations), and the examples are chosen to illustrate important aspects of the analytic framework, rather than simply how specific algorithms work.
Preface of the Book
This book is about how to process and interpret spectral images of landscapes. It is designed for students and professionals in a variety of disciplines who want to do hands-on computer analysis of spectral images to solve problems in the field. The term ‘‘landscape’’ conveys the idea of a part of the land surface that we can see from a vantage point on the ground. It is at landscape scales, rather than at regional or global scales, that we are able to make the most convincing links between remote-sensing measurements and the materials and objects that we see on the ground. In this book, the approach to the study of spectral images at this scale is unconventional. Most other texts on spectral remote sensing emphasize radiative transfer, engineering aspects of remote-sensing systems, and algorithms for manipulating digital images. Indeed, these subjects already have been covered so well that we see no reason to replicate this work. We have taken a new approach that focuses on the intersection of photo interpretation and spectroscopic modeling.
There are practical reasons for shifting the emphasis to physically based models. People in a variety of disciplines use spectral images to obtain information that is important for their landscape-scale projects. For many of these projects it is essential that the extracted information is correct and that it makes sense in the given context. Investigators also want spectral images to give them new insights into materials and processes so that there is the possibility for discovery. However, text books that treat the fundamentals of remote-sensing science and engineering rarely take the next step to explain
how actually to go about interpreting spectral images. In our experience, students who have taken the basic remote-sensing courses, when faced with a practical problem that requires interpreting the data, often express the same frustration: ‘‘What do you actually DO?’’ they ask. They know how to invoke algorithms, but they do not yet have a clear understanding of how to tease the desired information out of the data. More importantly, many are not yet familiar with the spectral properties of materials on the ground.
Our objective is to bridge the gap between the more theoretical and engineering side of remote-sensing science and the world that we encounter when we put on our boots and venture into the field. A basic premise is that remotely sensed spectral images are a proxy for observing directly on the ground. Therefore, to interpret spectral images we need to understand the spectral behavior of natural materials, and we need to be able to think like field investigators. This is a tall order, and it may stretch the borders of most remote-sensing curricula. The payoff is a new ability to extract information from spectral images. By applying physical models to remote sensing at landscape scales, we also enhance our ability to ‘‘scale up’’ interpretations to regional and global scales.
To achieve our objectives, we have limited the scope of the book to spectral images in visible and near-infrared wavelengths, and, to a lesser extent, to thermal-infrared images. This emphasis reflects the reality that the most readily available spectral images to support field work are the ones that are being most widely used. Many of our examples are based on Landsat Thematic Mapper (TM) images of areas that we personally have studied on the ground. Landsat images happen to be ideal to illustrate the methods that we discuss, and they comprise a nearly continuous (and underutilized) 30-year global data base for on-going studies of environmental and land-use changes. The field of view of Landsat TM images is large enough to encompass most field-study areas, and the pixel footprint is small enough to recognize where you are on the ground. The six visible and near-infrared TM Bands provide useful spectral information, but the quantity of data is relatively easily managed. Nevertheless, the analytical and interpretive approach that we describe applies to all spectral imaging systems, including imaging spectrometers having hundreds of spectral channels.
In order to reach a broad, multidisciplinary audience, our explanations are light on equations and heavy on visual examples. The book is suitable as a text for classes in remote sensing and as a professional reference; however, we do assume that our readers already have some background in the
fundamentals of remote sensing. For those who need to brush up, there are several excellent basic texts in print. We elaborate on certain topics that only are treated briefly, or are omitted from most texts, and that we feel are essential for image interpretation. These topics include spectral contrast,
basic spectral components of scenes, spectral unmixing, the application of fraction images, detectability of targets, and physically based thematic mapping.
Content
About the authors page ix
Preface xi
Acknowledgments xiv
1 Extracting information from spectral images 1
1.1 Introduction 1
1.2 Field studies and spectral images 3
1.3 Photo interpretation of spectral images 7
1.4 Spectral analysis of images 19
1.5 Testing and validating results 27
1.6 Summary steps for extracting information 35
2 Spectroscopy of landscapes 39
2.1 Basics of spectroscopy for field investigators 39
2.2 Spectroscopy at landscape scales 52
2.3 Spectroscopy applied to images 60
3 Standard methods for analyzing spectral images 65
3.1 Initial evaluation 65
3.2 Calibration 70
3.3 Enhancement for photo interpretation 81
3.4 Data reconnaissance and organization 84
3.5 Physical modeling with spectral data 112
4 Spectral-mixture analysis 126
4.1 Endmembers, fractions, and residuals 128
4.2 Shade 135
4.3 Fraction images 137
4.4 Finding endmembers 145
4.5 Calibration feedback 159
4.6 Nonlinear mixing 164
4.7 Thermal-infrared images 165
5 Fraction images of landscapes 168
5.1 What to do with fraction images 168
5.2 Classification using endmember fractions 183
6 Target detection 192
6.1 Spectral contrast and target detection 192
6.2 Detection limits 224
6.3 Spectral contrast and spatial scale 237
7 Thematic mapping of landscapes 244
7.1 Field maps and image-derived maps 244
7.2 Thematic mapping with spectral images 250
8 Processes and change 298
8.1 Process pathways in spectral images 298
8.2 Reference pathways 312
8.3 Mapping changes in landscapes 324
Glossary 337
Reference 350
Index 357
Download Here
Posts
No comments:
Post a Comment