International Journal of Computational Cognition (http://www.YangSky.com/yangijcc.htm) Volume 2, Number 2, Pages 35–69, June 2004 Publisher Item Identifier S 1542-5908(04)10203-0/$20.00 Article electronically published on April 12, 2003 at http://www.YangSky.com/ijcc22.htm. Please cite this paper as: hRonei Marcos de Moraes, “An Analysis of the Fuzzy Expert Systems Architecture for Multispectral Image Classification Using Mathematical Morphology Operators(Invited Paper)”, Inter- national Journal of Computational Cognition (http://www.YangSky.com/yangijcc.htm), Volume 2, Number 2, Pages 35–69, June 2004i. AN ANALYSIS OF THE FUZZY EXPERT SYSTEMS ARCHITECTURE FOR MULTISPECTRAL IMAGE CLASSIFICATION USING MATHEMATICAL MORPHOLOGY OPERATORS(INVITED PAPER) RONEI MARCOS DE MORAES Abstract. A fuzzy expert-system architecture for image classifica- tion was proposed by Moraes, Banon and Sandri (1998, 2000, 2002), whose rules are implemented through translation invariant mathemat- ical morphology operators. In this paper, we analyze this architecture by an expert system that classifies an area of the Tapaj´os National Forest, in Brazil. We compare this classifier with others classical clas- sifiers and analyze also its performance in other areas with the same rules. We conclude that the expert systems for image classification are more accurate that others classical classifiers. However, the ex- pert systems are very dependent of the knowledge of an expert and they are not adaptive for classification of others areas. c 2003 Yang’s Scientific Research Institute, LLC. All rights reserved. 1. Introduction The use of mathematical morphology has produced many applications in several areas of digital image processing, and also in what regards pattern recognition in binary images [32]. Work on the classification of gray-level images using mathematical morphology is still in its early stages. Some results can be found on shape and texture classification using granulometry [1, 10, 11]. Moraes [22] proposed a classifier based on an intersection of an elementary morphological operators: an erosion and an anti-dilation [3]: the sup-generating operator which it is the base of classification process [22]. Received by the editors April 4, 2003 / final version received April 12, 2003. Key words and phrases. Fuzzy expert systems, multispectral image classification, mathematical morphology, classification algorithms, performance analysis. The author would like to thanks to Gerald Jean Francis Banon, Sandra Aparecida Sandri and Michelle Pontes Seixas for valuable contributions to this work. c 2003-2004 Yang’s Scientific Research Institute, LLC. All rights reserved. 35