Ghosh, Jayanta Kumar 
However, the framework of knowledge based (KB) system does provide powerful and flexible methodologies to 
analyse the domain-specific knowledge and heuristics of visual image interpretation. There are many KB systems 
developed for satellite image interpretation by many researchers [Goldberg et al. (1985), Civco (1989), Blonda et al 
(1991) etc.]. It is found that nearly all the developers attempt to follow photointerpretation process without taking 
care of the inherent fuzziness of analyst’s cognition and reasoning processes and that of the satellite data. The 
approaches towards generalisation of some systems are also primitive. 
Ghosh (1996a) adopted a fuzzy knowledge based approach for interpretation of land cover. Segmentation and 
classification of pixels were done simultaneously by addressing the Linguistic variables in their fuzzy labels. But the 
parameterizations of the characteristic function were carried out without taking care of the uncertainty associated 
with it and inference mechanism was crude. 
The present paper discusses the development of an automated image interpretation system that follows a similar 
approach as adopted by Ghosh (1996a) but incorporates some new features to make it more logical and accurate. 
The proposed system is meant for mapping of the tea-gardens form satellite images. 
2. THE PROPOSED SYSTEM 
The broad objective of the proposed work is to develop an image interpretation system for mapping of tea gardens 
from satellite images emulating an expert image analyst. Experts, generally, use linguistic variables and their fuzzy 
labels to address facts and the domain knowledge of information classes. The reasoning for interpretation is also 
carried out by fuzzy inferencing (Zadeh, 1973). Thus, a fuzzy logic system has the capability to satisfy the objective 
of this work. The schematic representation of the proposed system of image interpretation is shown in Figure 1. 
  
CHARACTERISTIC 
FUNCTION 
  
  
  
  
  
  
Y 
PHYSICAL 
MEASUREMENT | PREPROCESSING LINGUISTIC FUZZY 
SYSTEM SPACE i VARIABLE LABLES 
  
v 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
S ub KNOWLEDGE INFERENCE 
BASE ENGINE 
à 
CLASSIFIER 1 
INTERPRETED 
IMAGE 
  
  
  
  
  
  
  
  
  
  
  
  
  
Figure 1 The Proposed Image Interpretation System 
Y 
| OUTPUT | 
2.1 Physical system 
The land covers within the study area represents the physical system module for the proposed system 
  
2.2 Measurement space 
Geocoded satellite data that represents the physical system module numerically, by spectral measurements at 
different bands, defines the measurement space module of the system. 
2. 3 Preprocessing 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 461