Fusion of PCA and Segmented-PCA Domain Multiscale 2-D-SSA for Effective Spectral-Spatial Feature Extraction and Data Classification in Hyperspectral Imagery

As hyperspectral imagery (HSI) contains rich spectral and spatial information, a novel principal component analysis (PCA) segmented-PCA (SPCA)-based multiscale 2-D-singular spectrum (2-D-SSA) fusion method is proposed for joint spectral–spatial HSI feature extraction classification. Considering the overall spectra adjacent band correlations of objects, PCA SPCA methods are utilized first dimension reduction, respectively. Then, 2-D-SSA applied onto dimension-reduced images to extract abundant features at different scales, where again dimensionality reduction. The obtained then fused with global derived from form (MSF-PCs). performance extracted MSF-PCs evaluated using support vector machine (SVM) classifier. Experiments on four benchmark data sets have shown that outperforms other state-of-the-art methods, including several deep learning approaches, when only small number training samples available.