Professional Documents
Culture Documents
Textbook Hyperspectral Imaging Remote Sensing Physics Sensors and Algorithms 1St Edition Dimitris G Manolakis Ebook All Chapter PDF
Textbook Hyperspectral Imaging Remote Sensing Physics Sensors and Algorithms 1St Edition Dimitris G Manolakis Ebook All Chapter PDF
https://textbookfull.com/product/hyperspectral-remote-sensing-
fundamentals-and-practices-1st-edition-ruiliang-pu/
https://textbookfull.com/product/hyperspectral-remote-sensing-of-
vegetation-volume-iv-advanced-applications-in-remote-sensing-of-
agricultural-crops-and-natural-vegetation-second-edition-huete/
https://textbookfull.com/product/polarization-remote-sensing-
physics-lei-yan/
https://textbookfull.com/product/remote-sensing-of-vegetation-
hyperspectral-indices-and-image-classifications-for-agriculture-
and-vegetation-second-edition-huete/
Fuzzy Machine Learning Algorithms for Remote Sensing
Image Classification 1st Edition Anil Kumar
https://textbookfull.com/product/fuzzy-machine-learning-
algorithms-for-remote-sensing-image-classification-1st-edition-
anil-kumar/
https://textbookfull.com/product/novel-sensors-and-sensing-first-
edition-jackson/
https://textbookfull.com/product/image-analysis-classification-
and-change-detection-in-remote-sensing-with-algorithms-for-
python-fourth-edition-canty/
https://textbookfull.com/product/lidar-remote-sensing-and-
applications-1st-edition-pinliang-dong/
https://textbookfull.com/product/lidar-remote-sensing-and-
applications-1st-edition-pinliang-dong-2/
Hyperspectral Imaging Remote Sensing
A practical and self-contained guide to the principles, techniques, models, and tools
of imaging spectroscopy. Bringing together material from essential physics and digital
signal processing, the book covers key topics such as sensor design and calibration,
atmospheric inversion and model techniques, and processing and exploitation algo-
rithms. Readers will learn how to apply the main algorithms to practical problems, how
to choose the best algorithm for a particular application, and how to process and inter-
pret hyperspectral imaging data. A wealth of additional materials accompany the book
online, including example projects and data for students, and problem solutions and
viewgraphs for instructors.
This is an essential text for senior undergraduate and graduate students looking
to learn the fundamentals of imaging spectroscopy, and an invaluable reference for
scientists and engineers working in the field.
DIMITRIS G. MANOLAKIS
Massachusetts Institute of Technology
Lincoln Laboratory
RONALD B. LOCKWOOD
Massachusetts Institute of Technology
Lincoln Laboratory
T H O M A S W. C O O L E Y
University Printing House, Cambridge CB2 8BS, United Kingdom
www.cambridge.org
Information on this title: www.cambridge.org/9781107083660
c Cambridge University Press 2016
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2016
Printed in the United Kingdom by TJ International Ltd. Padstow Cornwall
A catalogue record for this publication is available from the British Library
Library of Congress Cataloging-in-Publication Data
Names: Manolakis, Dimitris G., author. | Lockwood, Ronald B., author. |
Cooley, Thomas W., author.
Title: Hyperspectral imaging remote sensing : physics, sensors, and
algorithms / Dimitris G. Manolakis (Massachusetts Institute of Technology,
Lincoln Laboratory), Ronald B. Lockwood (Massachusetts Institute of
Technology, Lincoln Laboratory), Thomas W. Cooley
Description: Cambridge, United Kingdom : Cambridge University Press, 2016.
Identifiers: LCCN 2016025909 | ISBN 9781107083660 (hardback) |
ISBN 1107083664 (hardback)
Subjects: LCSH: Multispectral imaging. | Remote-sensing images. | Imaging
systems. | Image processing–Digital techniques.
Classification: LCC TR267.733.M85 M36 2016 | DDC 771–dc23
LC record available at https://lccn.loc.gov/2016025909
ISBN 978-1-107-08366-0 Hardback
Additional resources for this publication at www.cambridge.org/9781107083660
Cambridge University Press has no responsibility for the persistence or accuracy
of URLs for external or third-party internet websites referred to in this publication,
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
To Anna
DGM
To Susan
RBL
Preface page xi
1 Introduction 1
1.1 Introduction 1
1.2 Infrared Sensing Phenomenology 3
1.3 Hyperspectral Imaging Sensors 8
1.4 Data Preprocessing 15
1.5 Data Exploitation Algorithms 19
1.6 Applications of Imaging Spectroscopy 25
1.7 History of Spectral Remote Sensing 28
1.8 Summary and Further Reading 33
1.9 Book Organization 33
Bibliography 654
Index 678
Isocrates, 436-338 BC
Greek orator and rhetorician
Sir Isaac Newton invented the first monochromator with two slits and a prism in 1666,
and is largely credited with founding the field of spectroscopy based on his publica-
tion of Opticks in 1704. Over a century after Newton’s optics experiments, Joseph von
Fraunhofer made the significant advance in the field with the use of a diffraction grating
to replace the prism as the dispersion element. His ground-breaking studies of emissions
from the sun, in which he cataloged the series of absorption features in the solar spec-
trum that bear his name, established the discipline of spectral remote sensing. An early
demonstration of the power of spectroscopy for identification was when Joseph Fou-
cault showed that two of the Fraunhofer lines corresponded to sodium emission. The
precision afforded by the diffraction grating facilitated the field of spectroscopy becom-
ing a mainstay for the fields of chemistry, astronomy and, in recent decades, remote
sensing. Imaging spectroscopy is a relatively recent discipline which has been made
possible by progress in several technology areas including detectors, optical design,
optical components, atmospheric radiative transfer, and computer processing, to name
just a few. While spectroscopy and imaging techniques have advanced independently,
imaging spectroscopy is the merger of both, using spectroscopic techniques to form a
multi-band-imaged scene or “hyper-spectral” image.
The hallmark of imaging spectroscopy is the utilization of spectroscopic techniques in
which measurements can be used for unambiguous direct and indirect identification of
surface materials and atmospheric trace gases. The measurement of the relative concen-
trations of materials and the ability to distinguish between subtle absorption features of
differing materials for a two-dimensional scene are examples of imaging spectroscopy.
The main objective of this book is to present the fundamentals of remote sensing
imaging spectroscopy from the perspective of the user. There is a strong interconnection
between foundational phenomenology, sensing technology, atmospheric propagation
effects, and exploitation algorithms. Successful application of remote sensing imag-
ing spectroscopy requires a working knowledge and fundamental understanding of all
xii Preface
these aspects. Providing the reader this foundation is the objective of this book. Imag-
ing spectroscopy is also used by meteorologists and atmospheric scientists to directly
characterize the properties of the atmosphere. Our focus is on the use of imaging
spectrometer data to identify attributes of the surface.
The book covers material from different disciplines ranging from physics, optics,
spectroscopy, linear algebra, statistics, and signal processing. Integrating material from
such disparate areas is not easy, but we have tried to provide smooth and coherent cover-
age. The reader is expected to have a basic background in physics, chemistry, electrical
engineering, or a related field. We attempt to proceed by introducing concepts from first
principles to provide both foundational ideas and consistent nomenclature throughout
the text.
We have tried to provide balanced coverage of topics of general interest. However, the
selection and emphasis has been, we hope only slightly, biased by our own experience
and backgrounds. We have selected material and references to reflect fundamental work
and provide useful and practical techniques, insights, and applications. We apologize
for any omissions or oversights of areas which may be sought by the reader, but the
foundations found throughout the book can be applied to a variety of applications. We
provide seminal bibliographic citations to provide historical context for the evolution
of the technology both from our direct experience in the field and through judicious
examination of the literature.
Special thanks go to Eric Truslow and Mike Piper for their cautious and thorough
review, helping with figures, and preparing the solutions manual. Michael Chrisp, Steve
Miller, and Kurt Thome lent their considerable expertise in reviewing several chapters.
Their contributions are gratefully acknowledged.
Also, RLB was privileged to study atomic physics under the guidance of Chun C.
Lin and in close collaboration with L. Wilmer Anderson and James Lawler at the Uni-
versity of Wisconsin at Madison. The presentation in Chapter 4 is extensively based
on the course in applied optics that was part of the curriculum there in the late 1980s,
co-developed by Edward Miller and Frederick Roesler. In particular, the instrumental
spectroscopy formalism here parallels Roesler’s notes and has been adapted for the mod-
est performance of an imaging spectrometer in comparison to the instruments used in,
for example, atomic and molecular physics or astrophysics. Fred Roesler collaborated
with Pierre Jacquinot early in his career, and his thinking, and therefore this presentation,
was heavily influenced by the French spectroscopy school.
RBL would also like to thank his parents, Robert and Sharon Lockwood. Both are
educators and farmers who instilled a keen interest in the natural world in each of their
children, leading to careers in science and mathematics. Having raised us on a family
farm in West Texas, they also ensured that we all have healthy work ethics! This book
is a result of their support and encouragement.
Most importantly, we must thank our families for their patience and enduring support
for this project. No undertaking of this magnitude and duration is accomplished without
a foundation of support to see it through. Anna, Susan, and Mary know the effort and
time invested in this undertaking, and each is a pillar of forbearance for the time we have
taken to complete this text. We are grateful for their wisdom, support, and endurance.
Preface xiii
This undertaking required significant encouragement from our families, friends, and
colleagues who recognized the opportunity to provide this compilation for the broader
community of scientists, researchers, and students entering the field. We are grateful for
the many voices of encouragement throughout this project. We wish also to thank the
team at CUP, including Phil Meyler and Heather Brolly, who have patiently awaited the
completed manuscript and provided valuable comments and guidance throughout. We
hope you find benefit from our enterprise.
Acknowledgments
We have been fortunate to build upon the work of many other researchers, and to have
access and help from many colleagues leading up to and through this project. We would
like to thank the many scientists and researchers who have contributed directly and
indirectly to this undertaking, and specifically want to thank: Gail Anderson, Peter Arm-
strong, Gregory Berthiaume, Stuart F. Biggar, William Blumberg, Rebecca N. Breiding,
Michael Brueggeman, Thomas G. Chrien, James Dodd, David Driscoll, Richard Dunn,
Janet Fender, James Gardner, Mary Ann Glennon, Steven Golowich, Peggy Grigsby,
Lynn Hatfield, David Hardy, John Jacobson, Angela Kim, Steven Lipson, Joaquin
Mason, Karis W. Miles, Dan Mooney, Richard Nadile, Sidi Niu, Martin Pilati, Ryan
M. Pospisal, C. Richard Quade, Frederick L. Roesler, Stephanie Schieffer, John Schum-
mers, Vittala Shettigara, Krista Steenbergen, Stanley Straight, Johnny L. Touchstone,
Klaes Wandland, Christopher J. Ward, Andrew Weisner, Benjamin Worker, and Randall
Zywicki.
1 Introduction
1.1 Introduction
The term “remote sensing” has several valid definitions. In the broadest sense, accord-
ing to Webster’s dictionary, remote sensing is “the acquisition of information about a
distant object without coming into physical contact with it.” For our purposes, remote
sensing deals with the acquisition, processing, and interpretation of images, and related
data, obtained from aircraft and satellites that record the interaction between matter and
electromagnetic radiation.
The detection of electromagnetic radiation via remote sensing has four broad compo-
nents: a source of radiation, interaction with the atmosphere, interaction with the earth’s
surface, and a sensor (see Figure 1.1). The link between the components of the system
is electromagnetic energy transferred by means of radiation.
Source The source of electromagnetic radiation may be natural, like the sun’s
reflected light or the earth’s emitted heat, or man-made, like microwave radar. This leads
to a classification of remote sensing systems into active and passive types. Active sys-
tems emit radiation and analyze the returned signal. Passive systems detect naturally
occurring radiation either emitted by the sun or thermal radiation emitted by all objects
with temperatures above absolute zero. With active systems, like microwave radar, it is
possible to determine the distance of a target from the sensor (range); passive systems
cannot provide range information.
Atmospheric interaction The characteristics of the electromagnetic radiation propa-
gating through the atmosphere are modified by various processes, including absorption
and scattering. This distortion is undesirable and requires correction if we wish to study
the earth’s surface, or desirable if we wish to study the atmosphere itself.
2 Introduction
Figure 1.1 Pictorial illustration of how the atmosphere, the solar illumination, and the spectral response of
the sensor affect the relationship between the observed radiance spectrum and the wanted
reflectance spectrum of the ground resolution cell. (A black and white version of this figure will
appear in some formats. For the color version, please refer to the plate section.)
Figure 1.2 Principle of hyperspectral imaging sensing. The resulting “data cube” can be viewed as a set of
co-registered images at multiple wavelength regions or as a set of spectra corresponding to each
pixel of the image. (A black and white version of this figure will appear in some formats. For the
color version, please refer to the plate section.)
ground element area. Each individual image can then be processed and analyzed using
established image processing techniques that exploit geometrical or shape informa-
tion. However, if we collect radiation measurements at different wavelength bands for
the same spatial sample, and arrange them in a vector, we can analyze the resulting
“spectral” data using multivariate statistical techniques.
In the remainder of this chapter, we discuss the basic principles underlying each
part of the remote sensing system, we outline the key ideas of data preprocessing and
exploitation techniques, and we summarize some common applications.
The main sources of radiation in passive remote sensing is the electromagnetic radiation
emitted by the sun and the self-emission of objects in the scene due to their temperature.
Electromagnetic radiation is the means by which electromagnetic energy is propagated
in the form of waves.
Electromagnetic waves are characterized by their location within the electromagnetic
spectrum. The division of the electromagnetic spectrum into different regions, such as
visible, infrared, or microwave, has more to do with the different methods used for
sensing it rather than the nature of the radiation itself. Figure 1.3 shows the regions used
in electro-optical remote sensing and the typical applications for each region.
Figure 1.3 Typical examples of spectral information present at different spectral regions of the
electromagnetic spectrum for different applications.
it to heat energy, and transforms the heat back into radiant energy at the maximum
possible rate allowed by thermodynamic laws, is a perfect thermal emitter known as a
blackbody. The spectral radiant emittance of a blackbody is given by Planck’s law.
For our present purpose, the spectral radiant exitance, the power per unit area emitted
by the sun, as seen from above the earth’s atmosphere, can be approximated by a black-
body curve at a temperature of 5800 K. Thus the sun, with a temperature of 5800 K,
will have a maximum emittance at a wavelength of 0.50 m. In contrast, the earth’s
ambient temperature on a warm day, due largely to heating by the sun, is about 300 K
or 27 ı C and the maximum spectral radiant emittance from earth’s features occurs at
a wavelength of about 9.7 m. This radiation is known as “thermal infrared” energy
because it is related to terrestrial heat.
In the visible and near infrared regions of the electromagnetic spectrum we study
the radiation from the sun as it is modified by the earth’s atmosphere and surface. In
the thermal infrared region we study the radiation emitted by the earth’s atmosphere and
surface. The dividing line between reflective and emissive wavelengths is approximately
4 m. Reflective energy predominates below this wavelength and emissive energy
predominates above it. This difference in the nature of radiation also has profound
effects on the sensors used in the reflective and emissive regions of the electromagnetic
spectrum.
Figure 1.4 shows the contrast between the solar spectral irradiance and the spectral
irradiance from a blackbody at 300 K. Irradiance will be discussed in more detail in
Chapter 2, but it describes the spectral energy per second that flows through a real or
1.2 Infrared Sensing Phenomenology 5
104 1
Irradiance (W/m2/μm)
0.8
Transmission
102
0.6
0.4
100
0.2
10−2 0
0.2 1 10 25
Wavelength (μm)
Figure 1.4 Irradiance and atmospheric absorption for the wavelength range 0 to 25 m. The black curve is
the exoatmospheric spectral irradiance and the dashed curve is the spectral irradiance from a
blackbody at a temperature of 300 K with both referenced to the left vertical scale. The gray
curve is the atmospheric transmission referenced to the right vertical scale. Note the presence of
atmospheric windows in the thermal wavelength regions 3 to 5 m and 8 to 14 m.
imaginary aperture of unit area. The two curves cross at slightly longer than the 4 m
dividing line defined above.
60
Fir tree
40
Dry, yellowed
20 grass
0
0.4 0.8 1.2 1.6 2.0 2.4
Wavelength (micrometers)
Figure 1.5 Typical spectral reflectance curves for different types of green and yellowed (dry) vegetation.
The spectral curves for green vegetation at different regions are shaped by the components
shown at the top.
1.2 Infrared Sensing Phenomenology 7
1
Grass
0.9
Emissivity
0.8
Granite
0.7
7 8 9 10 11 12 13
Wavelength ( m)
Figure 1.6 Typical emissivity spectra of grass, granite, and sandy loam soil materials in the thermal infrared
window from 7 to 13 m.
are called reflectance peaks. These spectral features provide valuable information for
the identification of different materials.
Real materials are not perfect blackbodies, but instead emit radiation in accor-
dance with their own characteristics. The ability of a material to emit radiation can
be expressed as the ratio of its radiant emittance at a given temperature to the radiant
emittance of a blackbody at the same temperature. This ratio is known as the spectral
emissivity of the material. The plot of emissivity as a function of wavelength is called
the emissivity spectrum of a material. Figure 1.6 shows examples of emissivity spectra
with unique spectral features; the data were obtained from the ASTER spectral library
described by Baldridge et al. (2009a).
Radiance in the thermal infrared, the mid-wave infrared and the long-wave infrared
(MWIR and LWIR) is determined by both the emittance and temperature of the
materials. Variations in temperature can cause both large radiance variations and
large spectral variations at the input of a remote sensor. Furthermore, it is not
possible to measure separately the spectral emissivity and temperature using hyper-
spectral sensing alone. Emissivities of most materials vary between 0.8 and 0.95
in the 8 to 14 m thermal IR window. Thermal emission is generally the dom-
inant term; the variations caused by variations in spectral emittance are small by
comparison. This causes additional variability in the statistics of target and back-
ground clutter, which makes detection and discrimination in the thermal infrared more
challenging.
In conclusion, the signal of interest in spectral remote sensing applications, that is,
the information-bearing signal, is the reflectance or emissivity spectrum of the imaged
material. The details and relationships between spectral reflectance, emittance, and
temperature will be fully developed in subsequent chapters.
8 Introduction
aa
a
l l a
l
ll
individual detector elements. The data are stored in a data structure, known as a data
cube, consisting of two spatial dimensions and one spectral dimension. Each datum in
the cube is associated with a particular spatial location of a certain size and its value
is proportional to the spectral radiant energy contained within that particular spectral
sample and area.
The order and procedure by which each datum in the cube is populated are deter-
mined by the details of the imaging spectrometer. The challenge faced by the optical
designer is to record a three-dimensional data set using the signal generated by either a
two-dimensional array of detectors, a line array of detectors, or a single detector. Addi-
tionally, there is a limited number of ways that light can be separated into its spectral
components.
As described above, the fore optic presents an image of a scene location at the input of
a spectrometer. The spectrometer input is often a spatial mask of some sort. For example,
a slit spectrometer employs a two-dimensional opening that is long and narrow. One can
think of this imaging process in the reverse sense with the spectrometer input mask
being imaged onto the scene. The spatial dimensions of the cube are then assembled as
this projected mask is scanned across the scene of interest.
Figure 1.8 illustrates some of the different methods by which the data cube is popu-
lated depending upon the type of imaging spectrometer. The data subset marked with an
“A” is from a slit spectrometer, which relies on a dispersing element to spatially separate
the light into the different wavelength samples which are recorded by a two-dimensional
detector array. A scanning mechanism is utilized to move the projected slit image across
Figure 1.8 Three-dimensional illustration of a data cube showing the spatial dimension .x, y/ and the
spectral dimension . The numbers of spatial and spectral samples will depend upon the details
of the sensor design. The regions, labeled A, B, and C, illustrate the portion of the data cube that
is acquired during one acquisition cycle as described in the text.
10 Introduction
2-D FPA
Spectral
Focusing
optics
Collimating Sp
optics ati
al
al
ati
Optics Sp
Dispersing
element
Slit
Scan
(a)
line
Spatial
2-D FPA
Spatial
Focusing Sp
ati
optics al
FFT al
e ctr
Optics
Sp Sp
ati
al
FTIS OPD
(b)
Figure 1.9 Illustration of two types of imaging spectrometers. Illustration (a) is a slit spectrometer with each
data frame composed of spatial and spectral dimensions and the second spatial dimension
generated through scanning the projection of the slit onto the scene. Illustration (b) is of a FTIS,
with each data frame corresponding to both spatial dimensions at a particular OPD. The spectral
information is recovered by taking the fast Fourier transform (FFT) of the interference data for
each spatial location. (A black and white version of this figure will appear in some formats. For
the color version, please refer to the plate section.)
the scene. The spatial image recorded by a single data frame has m 1 samples with m
being the number of spatial samples along the long axis of the slit. The spectral infor-
mation is recorded in the perpendicular direction by n detector elements. The entire
data cube is assembled by scanning the slit along the scene and reading out the array
of detector elements at an appropriate rate. This is often referred to as pushbroom scan-
ning, with the forward motion of the airborne or space-based platform providing the
slit scanning motion. The diagram in Figure 1.9(a) sketches this common and highly
successful design.
A second example illustrated in Figure 1.8 and labeled as “B” is a staring system that
utilizes a series of spectral filters mounted in a filter wheel. A system of this type “stares”
at an area of the scene and cycles through a series of filters. A two-dimensional array
records an area of the scene at the various filter wavelengths and the cube is populated
1.3 Hyperspectral Imaging Sensors 11
as the filter wheel cycles. Once the spectral data corresponding to the spectral bandpass
from each filter in the wheel has been recorded, the scanning mechanism transitions to
the next portion of the scene and the process is repeated until the full scene has been
imaged. This is not a commonly employed imaging spectrometer, but it does introduce
the concept of a staring sensor.
A staring imaging spectrometer that is employed, particularly for remote sensing of
the atmosphere, is a Michelson interferometer, or Fourier Transform Imaging Spectrom-
eter (FTIS), which is illustrated in the sketch of Figure 1.9(b). This spectrometer utilizes
interference of light to record the spectral information. The interference pattern is intro-
duced by splitting the light into two unequal paths and recombining it. The light will
constructively and destructively interfere after recombining depending upon the optical
path difference (OPD) and the wavelength. A moving mirror produces the necessary
OPD and the interference pattern is recorded by repeatedly reading out the data from
each detector element in a two-dimensional array at a particular mirror position. Each
element of the two-dimensional array records the spectral information for a particular
spatial sample. In this case, the spectrum is recovered by taking the Fourier transform
of the interference signal for each spatial location. Once the full mirror travel has been
completed the spatial scanning mechanism transitions to the next portion of the scene
and another area corresponding to the projection of the two-dimensional detector array
is recorded, as was done for the filter wheel system. The details of this type of imaging
spectrometer will be addressed in Chapter 4; it is mentioned here as a practical staring
system for some applications.
The final example, labeled as “C”, is an imaging spectrometer that measures a com-
plete spectrum for a single spatial sample. A system of this type employs either a line
array, as in the case of a slit spectrometer, or a single detector element, if the imaging
spectrometer is a Michelson interferometer. The full data cube is assembled by raster-
scanning the single spatial element until the full scene area has been collected. This
is often referred to as whiskbroom scanning, since the forward motion of the platform
generates one axis of the scanning motion and a rotating mirror “whisks” the projected
spatial sample in the perpendicular direction. An example slit imaging spectrometer
that utilizes this scanning method is the widely used Airborne Visible/Infrared Imaging
Spectrometer (AVIRIS) system built at the Jet Propulsion Laboratory (Vane et al., 1993).
Armed with this background in the collection of a full hyperspectral data cube, we
will now introduce the concepts critical to our discussion of the details of the sensor
systems. The spatial, spectral, and radiometric figures of merit are defined. These ideas
are critical in order to fully understand the details of each spatial-spectral datum in the
image cube.
Figure 1.10 Illustration of the ground resolution element and the geometrical figures of merit for an imaging
spectrometer. The IFOV is alternatively defined as the ratio of the pixel width p to the focal
length f .
from the ground surface area and focuses the intercepted rays onto the surface of a
detector element, as in the filter wheel case, or onto a spatial mask such as a slit. For
those systems with a spatial mask, the area is further segmented depending upon the
number of spatial samples as was described above for a slit spectrometer. Figure 1.10
illustrates this concept for a single spatial element from either case. The area of the
detector or segment of the spatial mask projected through the system’s optics onto the
earth’s surface coincides with the ground resolution element. Each ground resolution
element is mapped onto one picture element or pixel of the image collected by the sensor
as the surface is properly scanned.
The minimum solid angle through which radiant energy originating from the scene is
accepted by the optical system and reaches a single detector element or mask segment is
called the sensor’s instantaneous field of view (IFOV). The IFOV is a strictly geometric
factor used as a rule of thumb to describe the smallest spatial resolution element that
is sampled by the imaging spectrometer. The true spatial resolution is determined by
an analysis of the optical aberrations and the effects of diffraction. The angular field of
view is the full angular extent of all the IFOVs. This is often restricted to the cross track
dimension, as in the case of a slit spectrometer that is pushbroom scanned. The cross
track direction corresponds to the projection of the long axis of the slit onto the scene.
In this case, the along track dimension is accumulated by scanning as described above
and is arbitrary in its extent.
The ground sample distance or GSD is the simple geometrical representation of a
detector element or mask segment projected onto the scene. It is calculated by mul-
tiplying the IFOV by the range, the distance from the sensor to the scene location.
However, the true spatial sampling is somewhat different from this idealization. As
mentioned above, the light from a given scene point is modified due to optical aber-
rations and diffraction as it is sampled by the imaging optics. This smearing effect is
described by a weighting function, known as the point spread function of the sensor.
1.3 Hyperspectral Imaging Sensors 13
The spatial resolution is further smeared by transmission through the atmosphere and
the scanning process. The resulting spatial resolution is known as the ground resolution
distance or GRD. The GSD defines the smallest area viewed by the sensor and estab-
lishes a lower limit for the level of spatial detail that can be represented in a digital
image.
Although scanning mechanisms vary according to sensor type, after geolocation pro-
cessing, each measurement corresponds to a location in the scene and is presented as
a pixel in an image. A sensible geometrical correspondence exists between scene loca-
tions and image pixels, such that the image can be considered as a projection of the
earth’s surface onto an image plane.
15
10
5
Radiance( W/cm2sr/nm)
0
500 1000 1500 2000 2500
15
10
0
500 1000 1500 2000 2500
Wavelength(nm)
Detectors are classified on the basis of the physical mechanisms that cause the
conversion from radiant to electrical energy. The two main classes are thermal detec-
tors and quantum (or photon) detectors with imaging spectrometers employing the
later type.
The line or area arrays employed in an imaging spectrometer are composed of individ-
ual photoconductors. These semiconductor devices work quantum-mechanically with
photons being absorbed to produce free charge carriers. Photoconductors function over a
broad spectral range and are employed throughout our spectral region of 0.350 to 14 m.
The wavelength range is determined by the details of the semiconductor quantum struc-
ture with the quantum efficiency, the ratio of a quantum of light input to the device to a
charge carrier generated within the device, varying as a function of wavelength.
The resulting current generated by applying a voltage across the detector element
is collected and stored in a capacitor for measurement. This voltage is converted to a
digital number (DN) using an analog-to-digital converter with a set number of bits B,
typically 12 or 14. A B-bit converter records 2B levels. The number of bits determines
the radiometric resolution. For example, if the maximum range of the electrical signal
is R, then the radiometric resolution is R 2B .
Another random document with
no related content on Scribd:
the next evening, if the weather permitted. This programme had
been drawn up by my husband and myself two or three weeks
before Mr. Quinton’s arrival, but it has since come to light that the
Jubraj suspected us of treachery in asking him to arrange and be
present at these nautches.
We returned to the office in about an hour, after I had seen that all
the preparations for lunch were made. The cook had departed, but
the bearer and I between us managed to get things ready in a
fashion. I took a book to read with me, and busied myself in that
manner until, about one o’clock, Colonel Skene and some of his
officers, with my husband, returned from the scene of action. Our
first inquiries were for Mr. Brackenbury, and then it became evident
that something serious had happened to him, and all our fears were
aroused. After that, things seemed to assume quite a different
aspect, for the officers were all talking so gravely together, and did
not seem quite satisfied with the way things were going.
However, we went back to the Residency to get something to eat.
All had returned with the exception of Mr. Simpson and Captain
Butcher, who were still at the Jubraj’s house, and Mr. Brackenbury,
whose exact whereabouts were unknown. We had commenced
lunch, when my husband asked me if I would give orders that some
food should be sent to the two officers who were not able to leave
their posts, and I went away to a little room adjoining the dining-room
and commenced cutting sandwiches for them, as the servants had
disappeared, and one had to get everything for one’s self or go
without.
I had been busily engaged for about ten minutes, when I heard a
sound which filled me with alarm, and a bullet crashed through the
window above my head. It frightened me more than the one at the
telegraph-office had done, and I dropped my knife, left the
sandwiches as they were, and rushed into the dining-room. All the
officers meanwhile had gone out, and had found that the Manipuris
had crept round to the back of the Residency and commenced an
attack upon us, using as cover the Naga village which lay between
our grounds and the river. This was a clever move on their part, and
it was some time before the troops could drive them back, as most of
our men were engaged in holding the posts inside the palace
captured early in the morning, and this left only a small guard for the
Residency, treasury offices, and Sepoys’ camp. Eventually our party
set fire to the Naga village, and drove the Manipuris out. Bullets had
made their way through the window-panes and doors of the dining-
room, and had smashed some of the breakfast-things and the glass
on the sideboard. It was difficult to find out the most secure place in
the house, as the firing was hot in the front of the Residency by this
time, and the walls, being only lath and plaster, were little or no
protection.
My husband suggested my descending to the cellars, which were
under the house and built of stone; but I did not like the idea, and
remembered how scornful I had been when we had talked over
matters weeks before, and he had joked about the snug corner he
would make ready for me in the basement of the house. So I made
up my mind to remain above-board, so to speak, until the worst
came to the worst. It was heart-rending to see the work of
destruction which was proceeding in the different rooms meanwhile.
The windows were broken, and every now and then bullets crashed
into the rooms, smashing different things—first a picture, then a
vase, then a photograph. All my beloved household gods seemed
coming to grief under my very eyes, and I was powerless to save
them. We did try to collect some of the most valuable of our
belongings together and put them away in a heap in the durbar
room, which at that time had escaped with only one broken pane; but
it was dangerous work going into the front rooms to remove them, for
as the afternoon went on the firing became hotter, and bullets rained
into the house at every second.
It must have been about half-past four that the big guns began to
be played against us. It had been found necessary to concentrate
the whole of our force on the Residency and out-buildings, such as
the treasury and offices, and this entailed abandoning all the
positions captured in the early part of the day inside the outer wall of
the palace, and bringing all the men together. The wounded had to
be recovered from all directions and conveyed to the hospital, which
was some distance from the Residency.
Lieutenant Brackenbury had been discovered lying on the bank of
the river which flowed north of the palace, where he had fallen
shortly after the attack was made early in the morning. He had
mistaken the direction, having got the wrong side of the wall near the
Jubraj’s house, from which point he had been exposed to a heavy
fire from the enemy. It was only a marvel that he was still alive when
eventually discovered, for he had remained where he fell the whole
of that day, and the Manipuris had never ceased firing at him as he
lay. When his exact whereabouts did become known, it was a difficult
and dangerous task to remove him. Efforts had been made by some
of the Sepoys to drag him away, and a native officer had been
mortally wounded in the attempt. At last, about four o’clock in the
afternoon, he was rescued and brought into the hospital, and it was
found that he had received terrible injuries, being wounded in several
places.
The sound of the first shell which whizzed over the Residency
made me speechless with terror. I had heard the boom of the guns in
the morning, and knew that they had been used to try and drive
Captain Butcher’s party out of the Jubraj’s house, which had been
captured; but they had sounded some distance off, and I had not
realized how terrible they could be until they were turned against our
own house.
The cellars were by this time unavoidable. My husband told me
that we should have to make some sort of rough hospital in one of
them, as the Residency hospital, where the wounded had been
taken, was built of plaster and would not be bullet-proof; so we set to
work to get blankets and sheets down from the house, and
everything we thought might be useful.
Meanwhile shells were doing dreadful damage over our heads,
and we were afraid they might set fire to the thatch and force us out
of our temporary shelter. Luckily most of them went over the house
into the garden at the back, where they could not do such serious
damage; but the noise the guns made, added to the other firing,
which had never ceased, was deafening.
There was not the slightest doubt by this time that our position was
about as bad as it could very well be. I seemed paralyzed with fear,
and it was only by forcing myself to do something, and never thinking
or imagining for one moment what the end of it all might be, that I
kept my senses sufficiently to be able to make an effort to help the
rest. I heard that the wounded were to be brought up to the house
immediately, as the hospital was getting too hot for them to remain in
it. Poor fellows! they had endured so much as it was in getting there,
that it seemed very hard to be obliged to move them again so soon,
and take them up to the Residency.
There were a good many of us in the cellar by this time—Mr.
Quinton, Colonel Skene, my husband and myself, Mr. Cossins, and
Mr. Gurdon. It was about seven o’clock, and a lovely evening. The
sun was just setting, and the red glow of the sky seemed to
illuminate the landscape around and the faces of the colonel and my
husband as they stood in the doorway talking together in low tones.
It was no difficult matter to read what was written on both their faces,
and I did not dare ask what was going to happen.
At last my husband came and told me that we were to leave the
Residency, and try and find our way to Cachar. It seemed worse to
me to think of going out of the house than to remain there; but
whatever was to take place had to be at once, and there was no time
to spend in giving way to the terrible fear which possessed me.
However, a further consultation was held, and it was decided to
make a truce with the regent, and put an end to hostilities by coming
to some terms with him. A letter was written, which the Chief
Commissioner signed. It ran as follows:
‘On what condition will you cease firing on us, and give us
time to communicate with the Viceroy, and repair the
telegraph?’
While this letter was being written, the colonel had ordered our
buglers to sound the ‘cease fire,’ which they did at once; but it was
some time before the Manipuris followed suit. At last their guns
ceased, and all was quiet. Then my husband went out with the letter,
and called a Manipuri off the wall to take it to the Jubraj. The man
went away with it, and my husband returned to the Residency.
I remained where he had left me, alone for some minutes, though
some of the officers were standing just outside the door of the cellar
where I was sitting. It seemed so hard that I could not go with my
husband. I feared being left alone without him, and felt very lonely
and broken-hearted among so many men, mostly strangers to me. I
knew, too, that they would look upon me as an extra burden, and
wish me very far away.
I was roused to action by the doctor, who had taken advantage of
the truce to get his wounded brought up from the hospital to the
house, and had come up first to see what kind of a place could be
got ready. I showed him the cellars, for there were several, which all
communicated with each other, and formed the entire basement of
the house.
Shortly afterwards the Kahars[16] arrived, carrying poor Mr.
Brackenbury on a mattress, and the others followed fast, so that the
small cellar was very soon quite full of men lying side by side on the
stone floor. The blankets and sheets that we had already collected
were very useful, and I made several journeys up to the house, and
gathered up every kind of covering from every direction, and all the
pillows I could find. A little cooking-stove proved of great service. I
fixed it securely upon a table in one corner which I reserved for
cooking operations.
The soup we had made on the previous day was in great request.
Fortunately there was a large quantity of it, to which I added the
contents of five or six tins which I found in the store-room. Milk was
the difficulty. All the cows were out in the grounds, and many of them
had strayed away altogether and we could not get any milk from
them, so were obliged to fall back on condensed milk, of which we
also had several tins.
Some of the men were terribly wounded, but poor Mr. Brackenbury
was by far the worst. His legs and arms were all broken, and he had
several other injuries besides. It seemed a marvel that he was still
alive and fully conscious to all that was going on around him. The
doctor attended to him first of all, and had bound up his broken
limbs, and done as much as possible to alleviate his sufferings; but it
was a terrible sight to see the poor lad in such agony, and be so
powerless to lessen it in any way. He was very thirsty, and drank a
good deal of soup and milk, but we could not get him into a
comfortable position. One minute he would lie down, and the next
beg to be lifted up; and every now and then his ankle would
commence bleeding, and cause him agony to have it bound up
afresh. His face was gray and drawn, and damp dews collected on
his forehead from the great pain he was suffering.
That scene will never be forgotten—the little cellar with a low roof,
and the faces of the wounded lying together on the floor. We did not
dare have a bright light for fear of attracting attention to that
particular spot, and the doctor did his work with one dim lantern.
Such work as it was, too! Every now and then he asked me to go
outside for a few moments while the dead were removed to give a
little more space for the living.
There were some terrible sights in the cellar that night—I pray I
may never see any more like them; but being able to help the doctor
was a great blessing to me, as it occupied my attention, and gave
me no time to think of all the terrible events of the day, and the wreck
of our pretty home. I was very weary, too—in fact, we all were—and
when at about half-past ten I asked everybody to come and get
some sort of a dinner, they seemed much more inclined to go to
sleep, and no one ate much.
The dinner was not inviting, but it was the best that could be got
under the circumstances, for I had had to do it all myself. One or two
of the servants still remained, but they cowered down in corners of
the house, and refused to move out or help me at all. Perhaps had
we known that it was our last meal for nearly forty-eight hours, we
should have taken care to make the most of it; but no thought of
what was coming entered our minds, and long before the melancholy
meal was ended most of the officers were dozing, and I felt as
though I could sleep for a week without waking.
We all separated after dinner about the house. I went back to the
hospital for a little, and found the doctor wanted more milk, so I
returned to the dining-room, where I was joined by Captain Boileau,
and we sat there for some time mixing the condensed milk with
water, and filling bottles with it, which I took downstairs. It was
quieter there than it had been. The wounded had all been attended
to, and most of them had fallen asleep. Even Mr. Brackenbury was
dozing, and seemed a little easier, and only one man was crying out
and moaning, and he was mortally wounded in the head. So finding I
could do no more there, I went upstairs again, resolving, if possible,
to go to my room and lie down for a little while and sleep, for I was
very tired.
I went sorrowfully through our once pretty drawing-room, where
everything was now in the wildest confusion, and saw all the
destruction which had overtaken my most cherished possessions.
There are those who imagine that in a case like this a woman’s
resource would be tears; but I felt I could not weep then. I was
overwhelmed at the terrible fate which had come upon us, and too
stunned to realize and bewail our misfortunes.
Perhaps the great weariness which overcame me may have
helped me to look passively on my surroundings, and I walked
through the house as one in a dream, longing only to get to some
haven of rest, where I could forget the misery of it all in sleep.
I wended my way to the bedroom through a small office of my
husband’s, but when I reached the door I found it would not open,
and discovered that part of the roof had fallen in, caused by the
bursting of a shell. So I gave up the idea of seeking rest there, and
retired to the veranda.
I went down the steps and stood outside in the moonlight for a few
minutes. It was a lovely night, clear and bright as day! One could
scarcely imagine a more peaceful scene. The house had been
greatly damaged, but that was not apparent in the moonlight, and the
front had escaped the shells which had gone through the roof and
burst all round at the back. The roses and heliotrope smelt heavy in
the night air, and a cricket or two chirped merrily as usual in the
creepers on the walls.
I thought of the night before, and of how my husband and I had
walked together up and down in the moonlight, talking of what the
day was to bring, and how little he had thought of such a terrible
ending; and I remembered that poor lad lying wounded in the cellar
below now, who only twenty-four hours ago had been the life and
soul of the party, singing comic songs with his banjo, and looking
forward eagerly to the chances of fighting that might be his when the
morning came.
I wondered where my husband was, and why they had been away
so long. They would be hungry and tired, I thought, and might have
waited to arrange matters till the next day, as they had apparently
been successful in restoring peace. I had an idea of wandering as far
as the gate to see whether the party was visible, but on second
thoughts I went back into the veranda, and resolved to wait there
until my husband should return.
There was one of the officers asleep in a chair close to me, and I
was about to follow his example, when Captain Boileau came out,
and I went to him and asked him if he would mind going down to the
gate and finding out whether he could hear or see anything of the
Chief Commissioner’s party, and if he came across any of them to
say I wanted my husband. He went off at once, and I fell into a doze
in the chair.
It was about twelve o’clock at this time. I do not know how long I
had been asleep, when I was awaked suddenly by hearing the
deafening boom of the big guns again, and knew then that it was not
to be peace.
For a few seconds I could not stir. Terror seemed to have seized
hold of me, and my limbs refused to move; but in a minute I
recovered, and ran through the house down to the cellar again,