wte Sensing, 58
H. and Jacopsen,
Jrientation Up to
hop on Mapping
ging, pp.3.2-108
and roads from
yrmation. In: GIS
settlements from
't oriented image
outh Africa). In:
of Urban Areas!
(FRegensburger
D.
s from IKONOS
In: RSPS 2001,
rmation Society,
1 supported by
Zentre, Germany.
or his help during
B3. Istanbul 2004
APPLICATIONS OF ARTIFICIAL IMMUNE SYSETMS IN
REMOTE SENSING IMAGE CLASSIFICATION
Liangpei ZHANG, Yanfei ZHONG, Pingxiang LI
State Key Laboratory of information Engineering in Surveying Mapping & Remote Sensing, Wuhan Univerity 129
Luoyu Road, Wuhan, Hubei, 430079, China - zIp62@public.wh.hb.cn
KEY WORDS: Remote sensing , artificial immune system, pattern recognition , classification, immune algorithms
ABSTRACT:
In this paper, some initial investigations are conducted to apply Artificial immune system(AIS) for classification of remotely sensed
images. As a novel branch of computational intelligence, AIS has strong capabilities of pattern recognition, learning and associative
memory, hence it is natural to view AIS as a powerful information processing and problem-solving paradigm in both the scientific
and engineering fields. Artificial immune system posses nonlinear classification properties along with the biological properties such
as self/nonself identification, positive and negative selection, clonal selection. Therefore, AIS, like genetic algorithms and neural
nets, is a tool for adaptive pattern recognition. However, few papers concern applications of AIS in feature extraction/classification
of aerial or high resolution satellite image and how to apply it to remote sensing imagery classification is very difficult because of its
characteristics of huge volume data. Remote sensing imagery classification task by Artificial immune system is attempted and the
preliminary results are provided. The experiment is consisted of two steps: Firstly, the classification task employs the property of
clonal selection of immune system. The clonal selection proposes a description of the way the immune systems copes with the
pathogens to mount an adaptive immune response. Secondly, classification results are evaluated by three known algorithm:
Parallelepiped: Minimum Distance and Maximum Likelihood. It is demonstrated that our method is superior to the three traditional
algorithms, and its overall accuracy and Kappa coefficient reach 89.80% and 0.8725 respectively.
1. INTRODUCTION
Drawing inspiration from the vertebrate immune system, a new
research field of Artificial Immune Systems(AIS) is springing
up. The vertebrate immune system is a rich source of theories
and acts as an inspiration for computer-based solutions Over the
last few years there has been an increasing interest in the area of
artificial immune system. AIS uses ideas gleaned from
immunology in order to develop systems capable of performing
tasks in various engineering applications(de Castro and Von
Zuben 2002). Although AIS has demonstrated its great values,
few applications are reported in remote sensing. Therefore, in
this paper, our aim is to employ AIS, a new tool of information
analysis for remote sensing image classification.
In remote sensing image classification, an key issue is to
improve classification accuracy. Conventional statistical
classifier, such as maximum likelihood (ML), has been applied
for remote sensing image classification for many years.
However, these conventional multivariate statistical methods
require nonsingular and class-specific covariance matrices for
all classes. Because of the complexity of ground matters and the
diversity of disturbance, these traditional classification methods
often have the drawback of low precision. In order to overcome
the shortcoming of conventional classifiérs, artificial immune
systems are applied to remote sensing image classification.
Compared to the conventional statistical classifier, AIS
classifier has the capacity of self-learning and high robust and
the advantages of artificial immune systems lies in the
following theoretical aspects. First, AIS are data driven self-
adaptive methods in that they can adjust themselves to the data
without any explicit specification of functional or distributional
form for the underlying model. Second, they are universal
functional approximators in that AIS can approximate any
function with arbitrary accuracy. Third, AIS are nonlinear
models, which makes them flexible in modeling real world
complex relationships. By Experiment, it shows that AIS
classification algorithm has high classification precision and
can be used in remote sensing image classification.
The remainder of the paper is organized as follows: Section 2
overviews the human immune system and Section 3 deals with
the property of clonal selection of AIS. Section 4 explains the
image classification algorithm using AIS in detail. In section 5,
the experimental results are provided. Finally, the conclusion is
given in section 6.
2. THE HUMMAN IMMUNE SYSTEM
The human immune system is a complex system of cells,
molecules and organs that represent an identification
mechanism capable of perceiving and combating dysfunction
from our own cells and the action of exogenous infectious
microorganisms. The human immune system protects our
bodies from infectious agents such as viruses, bacteria., fungi
and other parasites. Any molecule that can be recognized by the
adaptive immune system is known as an antigen(Ag).The basic
component of the immune system is the lymphocytes or the
white blood cells. Lymphocytes exist in two forms, B cells and
T cells. These two types of cells are rather similar, but differ
with relation to how they recognize antigens and by their
functional roles, B-cells are capable of recognizing antigens
free in solution, while T cells require antigens to be presented
by other accessory cells. Each of this has distinct chemical
structures and produces many Y shaped antibodies form its
surfaces to kill the antigens. Ab's are molecules attached
primarily to the surface of B cells whose aim is to recognize
and bind to Ag's(N.K. Jerne,1973).
