Supplementing hyperspectral data with digital elevation

This paper describes an experimental study where using a fusion of two essentially different types of data proves significantly superior to the individual use of either one or the other. The task is to identify and accurately delineate building roof-tops in a flightline of hyperspectral data of the Washington D.C. Mall. There are 210 channels of spectral data available, supplemented with a channel containing digital elevation map (DEM) data for each pixel of the scene. Experiments using gradient-based algorithms on the DEM data show that its use alone is not sufficient to sharply delineate building boundaries. A spectral classifier does not have these problems. However, building roof-tops in this urban scene are constructed of different materials and are in various states of condition and illumination. This and the fact that, in some cases, the material used in roof-tops is spectrally similar to that used in streets and parking areas make this a challenging classification problem, even for hyperspectral data. It is shown in this paper that combining hyperspectral and DEM data can substantially sharpen the identification of building boundaries, reduce classification error, and lessen dependence on the analyst for classifier construction