A FUZZY IMAGE CLASSIFICATION PACKAGE FOR REMOTELY SENSED DATA

Abstract

Map preparation is one of the major jobs that is being computerized over the years. The process of preparing maps has under gone constant changes as one gets to unveil the complications involved in it. There have been tremendous improvement in the data collection, processing, analysis and presentation techniques. This has been made possible with the advancements in computer technology. The availability of digital remote sensing data has also revolutionized the map preparation process. The technique used to produced good quality maps from Remote Sensing Data is called digital image classification. Due to the limitations of earlier techniques, a new generation of techniques such as fuzzy, neural networks and knowledge based approaches are being applied to produces maps showing the spatial distribution of a category on ground per map. Since their is large heterogeneity in Remote Sensing Data due to mixing of various features on the earth's surface, fuzzy approaches seem to be more appropriate to represent the activities on the earth surface. However, these fuzzy techniques have been recently adopted by the Remote Sensing community, therefore these are not available in commercially available digital image processing softwares. Therefore and attempt has been made to develop a software for some of the widely used fuzzy classification algorithms. The package has been named as FUZZICLASS and has been written in VC++. It incorporates three fuzzy classifiers namely, Maximum Likelihood Classifier (MLC), Fuzzy c- Means Classifiers (FCM) and Linear Mixture Modelling (LMM) different options for these classifiers have been incorporated in the package. There is a separate display and accuracy assessment module to test the performance of the classified image both qualitatively and quantitatively