A PC BASED PACKAGE FOR SUPERVISED CLASSIFICATION OF IMAGES

Abstract

Since times immemorial nature has been a fascinating subject for man. It has always created a strong urge in the mind of man to explore his surroundings and to quench his quest for knowledge by knowing about things unknown to him. In an attempt to reach the unfathomable depths of knowledge he interacts with his surroundings and extracts the information resulting from these interactions. In order to know of something yet unknown, he categorises on the basis of his previous knowledge about the object as well as its properties. People working in different fields, classify objects according to their own interests. A botanist classifies plants and an astronomer classifies stars and planets. Similarly a remote sensing scientist identifies objects and judges their significance without coming in physical contact (Wolf,1983)with the help of art and science of techniques known to him(Jensen,1986) and extracts information according to his requirement. In order to extract information, from a remotely sensed image, manual and computer-assisted methods are used (Jensen,1986). Manual methods analyse the image received from remotely sensed data, detect and identify important phenomena in the scene and the information thus extracted is used in problem solving (Jensen,1986). Computer assisted methods,on the other hand, form a mathematical representation of an image (Dougherty and Giardina, 1987). Various operations are done on the image thus obtained. It is restored (cleaned up of noise), enhanced(features of special interest are made more distinctive),segmented(subdivided into mut