Exploitation of hyperspectral imagery and lidar for landuse classification

The goal of this research is the development of near-autonomous methods to remotely classify and characterize regions of military interest, in support of the Terrain Modeling Project Office (TMPO) of the National Imagery and Mapping Agency (NIMA). These methods exploit remotely sensed data sets including hyperspectral (HYDICE) imagery, near infrared and thermal infrared (Daedalus 3600), radar, in addition to terrain datasets. Data normalization and sub-pixel image registration are critical to the successful outcome of data fusion of these various types of imagery. While there has been a lot of attention toward potential contributions of hyperspectral imagery, very little has been published on the necessary pre-processing of hyperspectral data in order to extract information from imagery over large areas. This poster focuses on the methods and uncertainties associated with the necessary data processing of the hyperspectral datasets. In particular, the issues associated with mosaicking large numbers of frames of HYDICE data are discussed: first, the radiometric normalization and determination of apparent surface reflectance (atmospheric modeling versus direct calibration with ground reflectance panels), second, image registration (geometric correction), and third, hyperspectral band selection.