KNOWLEDGE BASED VEGETATION CLASSIFICATION

Abstract

Classification is the process in which the processed satellite image is divided into categories depending upon the spectral signature of the individual classes. Supervised classification is a classification type in which the user initially selects training data that is later used for classification process. During the process of supervised classification, ambiguities may arise on the selection of training pixels and the accuracy assessment pixels, that depend upon individual to individual. So, the approach is to tend towards knowledge based classification. It is a GIS based process where a database of the area is prepared by the digitization process of a high resolution satellite data of the concerned area. A shapefile is prepared as an output that contains the ground classes present. Separate shapefiles are generated by splitting the original shapefile into the number of ground classes present. The permanent non-vegetation related features as Residential Areas, waterbodies, roads etc. in the study area are masked ie. separated from the process of classification. The shapefile holds the spatial information about the prevalence of ground classes. The shapefiles are imported in the image to be classified. ROIs are created for the vegetation classes from the shapefile and the image is classified based upon these ROIs. The accurate digitization of this data will provide a way for the automatic classification of the images, as any image of the area can be classified provided the database has been prepared for that image acquisition time. The results are more accurate as compared to the traditional methods because ground truth data is also used in the process of classification. Keywords:. Classification, Supervised, ROI vi