DATA CLASSIFICATION, VISUALIZATION, AND ENHANCEMENT USING N-DIMENSIONAL PROBABILITY DENSITY-FUNCTIONS (NPDF) - AVIRIS, TIMS, TM, AND GEOPHYSICAL APPLICATIONS

The n-Dimensional Probability Density Functions (nPDF) approach is a u ser-interactive image analysis technique which overcomes many of the i nherent limitations of traditional classifiers. In this paper we illus trate the applications of nPDF analysis in three broad areas: data vis ualization, enhancement, and classification. For data visualization, n PDF provides a method for transforming multiple bands of data in a pre dictable and scene-independent way. These transformations may be desig ned so as to enhance a particular cover type, or to give the best visu al representation of the multi-band image data. These approaches are i llustrated with the enhancement of hydrothermally altered areas in The matic Mapper (Tm) data, and the display of a false-color composite of six bands of Thermal Infrared Multispectral Scanner (TIMS) imagery. Sp ectral frequency plots of the nPDF components give a multispectral vie w of data distribution that can be used to investigate the number and distribution of spectral classes in a high dimensional data set. In ad dition, these plots are used in a non-parametric classification of the image for discrimination of discrete classes, as well as for classes that are mixtures at the sub-pixel scale. In a mixed deciduous and con iferous forest, an nPDF Deciduous Forest Index shows a high correlatio n with percent deciduous vegetation determined from field surveys. A c lassification of TIMS imagery of Death Valley results in excellent dis crimination of 13 discrete rock types. Classification of TM data, as w ell as classification of combined geophysical data, is used to illustr ate the power and variety of complex applications. The procedure is th e opposite of a ''black box'' approach: nPDF transformations and plots show graphical representations of the spectral and informational clas s distributions, and the user decides on the exact location of the spe ctral boundaries of each class in the classification. In comparisons w ith standard statistical classifiers, nPDF is extremely accurate and f ast, making it possible to analyze large data sets, such as full scene s of Advanced Visible/Infrared Imaging Spectrometer (AVIRIS) data, on a personal computer.