IMAGE CLASSIFICATION BASED ON FUZZY LOGIC 
I. Nedeljkovic 
MapSoft Ltd, Zahumska 26 11000 Belgrade, Serbia and Montenegro 
igor.n@sezampro.yu 
Commission VI, WG VI/1-3 
KEY WORDS: fuzzy logic, classification, if-then rules, digital, imagery, remote sensing, land cover 
ABSTRACT: 
Fuzzy logic is relatively young theory. Major advantage of this theory is that it allows the natural description, in linguistic terms, of 
problems that should be solved rather than in terms of relationships between precise numerical values. This advantage, dealing with 
the complicated systems in simple way, is the main reason why fuzzy logic theory is widely applied in technique. It is also possible 
to classify the remotely sensed image (as well as any other digital imagery), in such a way that certain land cover classes are clearly 
represented in the resulting image. /f that’s so, can we use fuzzy logic technique to diminish the influence of person dealing with 
supervised classification? Can we eliminate the prejudice? These questions were the light motive for this paper. In this paper, a 
priori knowledge about spectral information for certain land cover classes is used in order to classify SPOT image in fuzzy logic 
classification procedure. Basic idea was to perform the classification procedure first in the supervised and then in fuzzy logic 
manner. The later was done with ©Matlab’s Fuzzy Logic Toolbox. Some information, needed for membership function definition, 
was taken from supervised maximum likelihood classification. Also, the idea for result comparison came from OPCI's ImageWorks 
used for supervised procedure. Results of two procedures, both based on pixel-by-pixel technique, were compared and certain 
encouraging conclusion remarks come out. 
1. INTRODUCTION =» input (image channels) and output variables (land 
classes) are introduced in Matlab's environment, 
1.1 About fuzzy logic =» membership functions are defined using results from 
; supervised classification, 
Over the past few decades, fuzzy logic has been used in a wide =» Matlab’s Fuzzy Logic Toolbox was used in 
range of problem domains. Although the fuzzy logic is definition of fuzzy logic inference rules, 
relatively young theory, the areas of applications are very wide: > these rules are tested and verified through the 
process control, management and decision making, operations simulation of classification procedure at random 
research, economies and, fot this paper the most important, sample areas and at the end, 
pattern. recognition and classification. Dealing with simple =» SPOT image classification was conducted. 
‘black’ and ‘white’ answers is no longer satisfactory enough; a 
degree of membership (suggested by Prof. Zadeh in 1965) 2. SUPERVISED CLASSIFICATION 
became a new way of solving the problems. A fuzzy set is a set 
whose elements have degrees of membership. A element of a 2.1 Input data 
fuzzy set can be full member (100% membership) or a partial 
member (between 0% and 100% membership). That is, the The procedure of supervised image classification was 
membership value assigned to an element is no longer restricted conducted with PCI ImageWorks software. As the source for 
to just two values, but can be 0, | or any value in-between. classification procedure, SPOT Image recorded in "XS" 
Mathematical function which defines the degree of an element's multispectral mode was used. This image contains three 
membership in a fuzzy set is called membership function. The channels recorded in following bands: 
natural description of problems, in linguistic terms, rather than =» band Bl covering 0.50 to 0.59 pm (green), 
in terms of relationships between precise numerical values is =» band B2 covering 0.61 to 0.68 um (red) and 
the major advantage of this theory. =» band B3 covering 0.79 to 0.89 pm (near infrared). 
An idea to solve the problem of image classification in fuzzy In order to use them further in different software (PCI 
logic manner as well as comparison of the results of supervised ImageWorks, Matlab), SPOT image channels (named 701, 702, 
and fuzzy classification was the main motivation of this work. 703) are first converted from original SPOT format into ri/, and 
Behind this idea was also the question if the possible promising then exported from tif into pix format in Geomatica Focus 
results can give the answer to the question of diminishing the module (Figure 1.). The images were taken over the city of 
influence of person dealing with supervised classification. Cologne. The size of images is 3593x2990 pixels. 
1.2 Algorithm 
In this paper, a priori knowledge about spectral information for 
certain land cover classes is used in order to classify SPOT 
image in fuzzy logic manner. More specifically, 
83