Spectral error characterization and database development

The authors are utilizing existing data as well as additional materials characterization to provide high resolution spectra and error statistics for a wide range of engineered and natural materials. Important aspects of the authors´ methodology involve determining the variability of these measurements between samples and among different samples belonging to broad material classes (e.g., soils, rocks, construction materials, paints, etc.). The definitions of these material classes, as well as the variability of spectra within a class or subclass, are key parameters required for quantitative remote sensing and exploitation algorithms. The authors have developed methodology for analyzing material classes and error statistics to provide spectral covariance matrices which facilitate the development of optimal classifiers. The product of their program is a database of error statistics suitable for a wide variety of remote sensing applications. The methodology is also directly applicable to the definition of new classes of materials when required. To date, they have assigned over 250 material samples into 12 main categories and 55 subcategories, based upon their physical and spectral characteristics. They have identified statistical factors for soils and a variety of construction materials and fabrics. They have begun a measurement and statistical extraction program to quantify estimates of uncertainty for measured spectra. They have found that comprehensive physical and chemical descriptions are important for defining factors