A COMPARATIVE STUDY ANALYZING THE METHODS OF CLASSIFYING HYPERSPECTRAL DATA FOR OUTCROP CHARACTERIZATION

Recently, the use of remote sensing to characterize and classify outcrops has grown an exponential amount to become a vial tool in studying the surface. Remote sensing technology provides a new outlook on the methods of acquiring and analyzing spatial, spectral and temporal resolutions. With given high spectral resolution, objects and materials can be identified by hyperspectral imaging sensors through the collection of reflectance values from wavelengths in visible infrared and shortwave infrared portions of the spectrum. In this study, ground-based hyperspectral data for four outcrops are collected, processed and analyzed to compare and determine the best method of classification. Although advances have been made on several classifying techniques, few studies go into great depth to explain the many variances obtained from the different methods of classifications. Classification methods such as Spectral Angle Mapper (SAM), Support Vector Machine (SVM), and Neural Network (NNC) will be focused on as well as a brief look at Mixture Tuned Match Filtering (MTMF). Ground truth will be established by a geochemical analysis on samples obtained from the outcrops using an X-Ray Fluoresce machine. This serves as both a guide through processing and analyzing hyperspectral data as well as an in-depth comparative study of what classification method yields the most accurate result.