DBN-based classifcation of spatial-spectral hyperspectral data

In this paper, present situation because of the high spectral image spectral information and spatial information leading to an increase in the increasing demand for new classification method of depth in recent years confidence in network feature extraction process large amounts of data, such as the Chinese information extraction, and other aspects of cancer determine the success of reality Combine. Introducing depth belief networks classify hyperspectral images, in excavating the hyperspectral data space information based on the study of the neighborhood mosaic spectral and spatial information used in combination, neighborhood stitching and spectral information integration and the weighted average of the empty Cape Joint strategy, through the experimental comparison with other methods to get the optimal weighted average method empty spectrum joint conclusions.