LITHOLOGICAL DISCRIMINATION USING BAND SELECTION & FEATURE EXTRACTION BY ASTER DATA, UM NAR AREA. EGYPT
Paper ID : 1159-ICRSSSA (R1)
Authors
Nehal mohamed abdelrahman soliman *1, Sahar Ayoub2
1Geology Department, NARSS
2Biological and geological department Education Collage Ain shams University
Abstract
Eastern Desert of Egypt has several types of igneous and metamorphic rocks, that were formed during collision between east and west Gondwana and the closure of Mozambique ocean 600 Ma . The study area as a part of the Central Eastern Desert is almost built up of ophiolitic mélange and associated rocks, together with subordinate molasse-type sediments and late tectonic volcanics and granitoid intrusives. Remote Sensing technique achieve great advantage over the normal survey methods due to their ability to map inaccessible areas and also for their time and money saving. Conventional band selection and feature extraction methods largely fall into .two categories in terms of criteria and objectives: OIF, SI, CI attempt to keep maximum image variability (regarded as information) and they are primarily used in preliminary exploration and image compression; the four class separability measurements attempt to maximize the between-class separability and they are primarily used for image classification. A new ratio 9/7-7/1-4/7 is extracted from the higher bands. Discrimination of geological lithology is a challenging subject to many scientists. ASTER data was used to extract the effectively display for visual interpretation. The three RGB channels should be selected from the image under the criteria of maximum information and maximum between-class separability. Three band selection methods, Optimum Index Factor (OIF), a Sheffield index (SI) and an alternative index (CI) have been used. Four separability measures: Divergence, transformed divergence, Bhattacharyya and Jeffreys - Matusita distance were applied to determine the best band subset of multispectral bands
Keywords
ASTER, Mapping Lithology, Um Nar area, Egypt
Status: Conditional Accept (Poster)