The Automatic Recognition and Classification of Ground Objects in Hyperspectral Images Based on the Projection Pursuit Method

We use sequential projection pursuit method to analyze all kinds of projection index, and put forward the projection index and optimized search algorithm which is fit for hyperspectral image data. We develop sequence in VC++ environment, take dimensionality reduction and feature extraction experiment on two kinds of hyperspectral image data, and also do the accuracy evaluation. We can find from the experiment result that by using sequential projection pursuit method, we remove the redundant information in hyperspectral data to a large extent, and reserve the effective information at the same time.