Spectral recognition of target based on the hyperspectral image data

Abstract

As it contains the spectral and image information of targets, the hyperspectral image measured by the airborne CASI imaging spectrometer can improve the accuracy of target identification effectively. Usually, the spectral matching methods are used in identification by extracting spectral information from hyperspectral image. Two matching algorithms, the spectral angle matching method and the binary encoding method, are investigated under the condition of different spectral resolutions. Firstly the atmospheric correction, the spectral curve smoothing and re-sampling are carried out in ENVI software. Then the different spectral resolutions are constructed by selecting spectral subsets from the hyperspectral image. The matching result indicates that the spectral resolution threshold of the spectral angle matching method is above the threshold of the binary encoding method. The result of this study has important guiding significance to the use of the matching algorithm and the selection of spectral resolution.