The election of Spectrum bands in Hyper-spectral image classification

this paper present a framework for the pre-process of hyper-spectral image classification, it seems to be proved an important and well method in this kind of field. We construct a new calculating method in preprocess which use for reference of digital image process, such us morphology method and relations of position. Because of the data of the spectrum bands of hyper-spectral has some degree of redundancy. Some similar bands have the similar information or even same information, so it would waste many counter resource. On the other hands, redundancy means when you matching every model, you will easily addicted to Hughes phenomenon. So bands electing before we use classification model begin to classify directly is very important. It not only can decrease time-cost also improve the accuracy and degree of stability. Our experiment proved our idea practice in real classification get a gorgeous effect, it shows in different complex classification surroundings, our method also perform great.