IMPROVEMENT IN SUPERVISED CLASSIFICATION ACCURACY USING SYNTHETICALLY GENERATED IMAGES DERIVED FROM SATELLITE DATA

Abstract

Techniques for estimating crop conditions and vegetal pattern in any region are based upon the available topographical or thematic maps. These maps are prepared from the information collected through field surveys and statistical data. Due to the disadvantages of conventional techniques, people have been for an alternative approach for monitoring the land use, especially vegetation. Remotely sensed data becomes ideal in many of these cases. Classification is an important step for remote sensing images. Supervised classification mostly adopted when there is previous knowledge of study area. It is closely controlled by the analyst. In this process, the user selects pixels that represent patterns or land cover features that represent patterns or land cover features that they recognize, or that they can identify with help from 'other sources, such as aerial photos, ground truth data or maps. There are two primary motivations for assessing the accuracy of a map. The first relates to understanding the errors in the map. Both producers of users of thematic maps are interested in this kind of information. Supervised classification accuracy can be improved using synthetically generated images derived from satellite data. Synthetically generated images are the images generated using some arithmetical operator such as ratio, addition, and subtraction or through some transformation such as Principal component analysis and Tassel cap. Feature selection techniques are used to select the best band combinations for the classification. Using synthetically generated images and raw bands in the classification accuracy can be improved.