409
VEGETATION ENDMEMBER EXTRACTION IN HYPERION IMAGES
M. Heidari Mozaffar a , M.J. Valadan Zoej b , M.R. Sahebi c , Y. Rezaei d
a Remote Sensing Department, KNToosi University of Technology, Mirdamad Cross, Valiasr Av., Tehran, Iran,
Morteza_Heidarymzaffar@yahoo.com
b Remote Sensing Department, KNToosi University of Technology, valadanzouj@kntu.ac.ir
c Remote Sensing Department, KNToosi University of Technology, sahebi@kntu.ac.ir
d Remote Sensing Department, KNToosi University of Technology, y.rezaei@gmail.com
Commission WG VII/3
KEY WORDS: Hyperspectral imaging, Endmember Extraction, Hyperion Images
ABSTRACT:
Hyperspectral imaging sensors on environmental applications have high spectral resolution and low spatial resolution so that
numerous disparate substances can contribute to the spectrum measured from a single pixel or in the field of view of the sensor. An
important problem in Hyperspectral imaging processing is to decompose the mixed pixels into the materials that contribute to the
pixel, endmember, and a set of corresponding fractions of the spectral signature in the pixel, abundances, and this problem is known
as the unmixing problem. According to the definition, an endmember is an idealized pure signature of a class. Endmember extraction
is one of the fundamental and crucial tasks in hyperspectral data exploitation. It has received considerable interest in recent years,
with many researchers devoting their effort to develop algorithms for endmember extraction from hyperspectral data. An ultimate
goal of an Endmember Extraction Algorithm (EEA) is to find the purest form of each spectrally distinct material on a scene.
Endmember extraction tendency to the type of endmembers being derived, and the number of endmembers, estimated by an
algorithm, with respect to the number of spectral bands, and the number of pixels being processed, also the required input data, and
the kind of noise, if any, in the signal model surveying done. Identifying endmembers that satisfy both physical and mathematical
imperatives is a considerable challenge, making autonomous endmember determination the hardest part of the unmixing problem. Of
three stages that comprise unmixing, endmember determination is the most closely aligned with the material identification
capabilities of unmixing. Non-statistical algorithms or Geometrical approach essentially assume the endmembers are deterministic
quantities, whereas statistical approaches view endmembers as either deterministic, with an associated degree of uncertainty, or as
fully stochastic, with random variables having probability density functions. In addition, specific features concerning the outputs,
inputs, and noise models used by these algorithms are included according to the model specifically distinguishing the properties of
endmember-determination algorithms.
In this paper, Endmember Extraction Algorithms (EEAs) applied on a Hyperion image of southern of Tehran, IRAN. A large number
of endmembers were suggested to enhance the classification accuracy while the seasonal variation in the spectral response should be
taken into account in vegetation classification. We compare results of Geometrical approach in vegetation endmember extraction
assistance with vegetation indices.
1. INTRODUCTION
1.1. OVERVIEW
Hyperspectral imaging sensors on environmental applications
have high spectral resolution and low spatial resolution so that
numerous disparate substances can contribute to the spectrum
measured from a single pixel or in the field of view of the
sensor. An important problem in Hyperspectral imaging
processing is to decompose the mixed pixels into the materials
that contribute to the pixel, endmember, and a set of
corresponding fractions of the spectral signature in the pixel,
abundances, and this problem is known as the unmixing
problem.
Availability of new generation of hyperspectral sensors such as
the Hyperion has lead to the new challenges in the area of crop
type mapping and agricultural management. The sensor’s 242
Spectral band between 400 and 2500 nm (level 1B1 ) and spatial
resolution of 30m bear high potentials for agricultural crop
discrimination and detailed land use classification[ 1 ].
Advantages of this technology include both the qualitative
benefits derived from a visual overview, and more importantly,
the quantitative abilities for systematic assessment and
monitoring. In every remotely sensed image, a considerable
number of mixed pixels is present. A mixed pixel is a picture
element representing an area occupied by more than one ground
cover type.Several research objectives were accomplished:
(1) Select optimal bands in hyperspectral images those are most
useful in vegetation classification,
(2) Identify optimal endmember, signature spectrum that
represents a certain class, for vegetation classification, and
(3) Test effective Endmember Extraction Algorithms for
classification of vegetation type.
First pre-processing step which iclude radiometric correction
and removing atmospheric effect involves calibration and
atmospheric correction. We used Fast Line-of-sight
Atmospheric Analysis of Spectral Hypercube (FLAASH) in
ENVI software for atmospheric correction. In this research,
capabilities of Hyperion hyperspectral imagery acquired from
an agricultural area located in southern parts of Tehran, Iran ,