AN ALGORITHM FOR REMOTE SENSING IMAGE CLASSIFICATION BASED ON 
ARTIFICIAL IMMUNE B -CELL NETWORK 
Shizhen Xu a ’ *, Yundong Wu b,c 
a Insitute of Surveying and Mapping, Information Engineering University - 66 Middle Longhai Road, Zhangzhou, 
Henan 450052 China - xshizhen@163.com 
b College of Sciences, Jimei University, 361021 Amoy, China - wuyundong@yahoo.com.cn 
institute of Automation, Chinese Academy of Sciences (CASIA), 100080 Beijing, China 
Youth Forum 
KEY WORDS: Pattern recognition, Classifier, Image Understanding, Feature Recognition, Land Cover 
ABSTRACT: 
In this paper a novel supervised classification algorithm (AIBN) for remote sensing image classification based on artificial immune 
B-Cell network is proposed. Even though some effective classifiers have been proposed in the field of Artificial Immune System, 
there are still some deficiencies in them. In AIBN, clone selection and immune network theories are used as mechanism for data 
training in order to gain reduction form of the data. After training the distribution and density information of B cells is learned by 
AIBN and B cells can directly be used as a criterion for classification. Then, the classification task is carried out. The experiment 
results show AIBN is superior to Maximum Likelihood classifier and Artificial Immune Recognition System. Some analyses of user 
defined parameters are made. AIBN provide an alternative way to perform remote sensing image classification task. 
1. INTRODUTION 
The human immune system is a complex of cells, molecules 
and organs, which keeps foreign invaders such as viruses, 
bacteria far away from us. Without immune system, humans 
can hardly survive. Like other biologically-motivated intelligent 
computation, Artificial Immune System (AIS) which takes 
inspiration form natural immune system has been applied to 
solve many problems, including search and optimization, 
classification and clustering, anomaly detection, automatic 
control and so on. 
Through the effort of researchers in the field of remote sensing, 
we could see AIS have been used as an alternative tool to solve 
remote sensing problems, such as Remote Sensing Image 
Classification(Zhong, 2006), image registration!Yin, 2003.) and 
image segmentation(McCoy, 1997). 
In this paper we propose a novel algorithm for supervised 
remote sensing image classification based on immune B-Cell 
network, we call it AIBN for short. AIBN uses mechanisms of 
clonal expansion and immune network suppression together to 
train data set and use K-Nearest- Neighbours (KNN) algorithm 
to classify data set. Unlike other supervised classification 
algorithms in the fields of AIS, AIBN takes relationship among 
antigens into consideration and could reserve the density 
information of the data set by calculating adaptive radius of 
each antigen. 
The rest of the paper is organized as follows. In section 2, we 
present an overview of the natural immune systems. In section 3, 
we review some of the current classification algorithms based 
on AIS. In section 4, we describe our AIBN algorithm. In 
section 5, we present its performance on remote sensing image 
classification problems and make some analysis. Finally, in 
section 6, we present our conclusions. 
2. THE NATURAL IMMUNE SYSTEM 
The immune system could perform many tasks, such as learning, 
memory, pattern recognition, optimization, noise tolerance, 
generalization, and distribution detection. The basic of those 
performances are built on the ability of immune system that can 
recognize all cells within the body and categorize them as self 
or non-self. Then it removes those non-self substances through 
some immune mechanisms and immune processes. There are 
two type of immunity, innate and adaptive. Only adaptive one 
is normally concerned with. The adaptive immune response is 
mainly based on the behaviours of two types of lymphocytes: B 
cells and T cells. 
When a pathogen invades the body, some B cells recognize the 
antigens on the surface of antigen presenting cells with different 
affinity. With the help of T cells, those B cells begin to 
proliferate and mutate to produce antibodies for matching the 
antigen better. It is said memory cells transiting from high 
affinity B cells retain in the body for a long periods of time 
after the response. And these memory cells will response more 
rapidly and powerfully to a similar pathogen in the future 
response. 
Above mentioned content is from the viewpoint of clonal 
selection. There is a controversial theory, immune network 
theory, which holds that B cells are interconnected through their 
idiotopes and parotopes. Then the recognition between B cells 
could results positive or negative affect to the response to the 
Corresponding author. This is useful to know for communication with the appropriate person in cases with more than one author.