CSEIT1722312 | Received : 16 April 2017 | Accepted : 24 April 2017 | March-April-2017 [(2)2: 1000-1004]
International Journal of Scientific Research in Computer Science, Engineering and Information Technology
© 2017 IJSRCSEIT | Volume 2 | Issue 2 | ISSN : 2456-3307
1000
Comparative Study of Algorithms for Hyper Spectral Image
Classification
Pallavi Juyal, Manish Jain, Varsha Nemade
Department Computer Engineering, Mukesh Patel School of Technology Management and Engineering, Mumbai,
Maharashtra, India
ABSTRACT
Now days the use of remote sensing imagery has increased drastically, ranging from the applications in farms to that
in the field of defense. These images range from hyper spectral, multispectral to ultra spectral. One of the major
applications of the remote sensing image is that of classification of the image. Several algorithms have been found
for classification of images based on various factors. The algorithms used for classification are either supervised or
unsupervised. In this paper we study different algorithms used for classification of the hyper spectral images. An
analysis of the output obtained by the implementation of various algorithms has also been done. The analysis has
been done based on the number of pixels classified.
Keywords : Remote sensing, hyperspectral image, classification, classifier, spectrum, electromagnetic spectrum,
Euclidian distance.
I. INTRODUCTION
Remote sensing refers to the collection and analysis of
data that is acquired by a satellite type instrument.With
an example, remote sensing can be explained as
follows, if we take the picture of a school building and
in the picture we see that it is composed of doors,
windows and roof all of which appear to be of different
colour, this is remote sensing.
Remote sensing deals with these three components,
a platform is needed to hold the instrument,a
target object to be acquired and then observed and an
instrument or a detector to sense and observe the target.
The image shown below is an explanation of how
remote sensing is actually performed.
Figure 1. Acquisition of Satellite Image
A hyper spectral imaging like any other multispec
image deals essentially with the collection and
processing of information by making use of the
electromagnetic spectrum which is the range of all
types of EM radiation. The naked human eye colour
sees only colour in 3 bands, the RGB whereas a
spectrum image separate the image into multiple bands.
This technique of dividing the image into various
different spectrums is far beyond the visible. Below is
shown a hyper spectral image which will be used as a
input in the implementation of the classification
algorithm.
Figure 2. The Image to be used as input