Feature Extraction for Hyperspectral Image Classification Based on Scale Invariant Feature Transform-Locality Preserving Projection Algorithm

摘要

Feature extraction is an important research topic in the hyperspectral remote sensing area. The new data point in the low-dimensional space is determined by the feature extraction based on LPP algorithm. However, the algorithm seems like a non-supervised linear dimension-reducing method. The category identification information of a sample is not considered when the local graph recently is established. An application of SIFT-LPP algorithm was proposed for the feature extraction of hyperspectral images. First of all, gradient features are obtained from the key points of the image by the extraction of image local invariant features (the SIFT operators). Then, it is mapped into a low-dimensional space by using LPP algorithm. Finally, the classification method of the image is achieved by support vector machines. The classification results illustrate that the total classification accuracy of the proposed method is higher than that of traditional method.