Research On Automatic Extraction Of The Body Of Interest In Hyperspectral Image

As one of the products of hyperspectral imaging technology, the essence of hyperspectral image is a three-dimensional data cube which formed by two-dimensional spatial dimension and one dimension spectral dimension. Although it has a very high spectral resolution, it also has a lot of defects such as high dimension and big information quantity accordingly. This situation causes a lot of issues and even triggers "dimension disaster". Therefore, how to extract a interested body which contained by the high-quality spatial information and spectral information from the "information data cube" of the hyperspectral image, has become a imminent demand of research. The article firstly uses a band selection method of innovation in the spectral dimension of hyperspectral data cube, the method introduces the grey incidence decision theory, then optimizes the results of single band selection further. Then carry out image segmentation in the spatial dimension of the results of band selection, this method can determine the region of interest of the spatial dimension, and get the final result——Body of Interest. Thirdly, package the entire algorithm into an automatic extraction algorithm of the Body of Interest and carry out simulation experiment. Last of all, compare the traditional method with the experimental results, the results show that the method has realized the automatic extraction of the Body of Interest, and improved the quality of the dimension reduction in hyperspectral image, then it could provide a powerful data support for the follow-up operation of deeper level.