干旱区生态环境知识资源中心

Satellite imagery provides an advanced and cost-effective way to emphasize the lithological information of any poorly mapped area. In this study, a lithological identification method, based on the support vector machine (SVM) classifying algorithm, is proposed to discriminate the widely exposed lithological features around Wadi Atalla area, Central Eastern Desert of Egypt. The SVM classifier has been applied to a series of datasets derived from the Advanced Space-borne Thermal Emission and Reflection Radiometer (ASTER) imagery with an ASTER-derived Digital Elevation Model (DEM) in order to find the best set of data input for the optimal classification results. Combinations of various input datasets including; the Visible Near Infrared (VNIR) and Shortwave Infrared (SWIR) of the ASTER bands as well as some of its derivatives e.g. Principal Component Analysis (PCA), Independent Component Analysis (ICA), the stacking of both PCA and ICA data (PC/IC-Stack) as well as the ASTER generated DEM are tested for best classification accuracy. A combination of the ASTER-(PCAJICA and DEM stack) data input provided the highest overall classification accuracy of 95% for the independently validated samples of the Ethological classes using the SVM classifier. Results indicate that this particular dataset input can help producing a good lithological distribution map for any remote area that have some background information about its lithology. This new proposed method successfully differentiated between ophiolitic assemblage, highly deformed rocks of Meatiq group, intrusive rocks and Hammamat molasse sediments in the study area.