Local and Global Spectral Features for Hyperspectral Image Classification

Hyperspectral images (HSI) contain powerful spectral characterization capabilities and are widely used especially for classification applications. However, the rich spectrum contained in HSI also increases difficulty of extracting useful information, which makes feature extraction method significant as it enables effective expression utilization spectrum. Traditional methods design features manually, is likely to be limited by complex information within HSI. Recently, data-driven methods, use convolutional neural networks (CNNs), have shown great improvements performance when processing image data owing their automatic learning abilities classification. The CNN extracts based on convolution operation. Nevertheless, local perception operation focus (LSF) weakens description between long-distance ranges, will referred global (GSF) this study. LSF GSF describe from two different perspectives both essential determining Thus, study, a local-global (LGSF) optimization proposed jointly consider To increase relationship spectra possibility obtain with more forms, we first transformed 1D vector into 2D image. Based image, module (LSFEM) (GSFEM) automatically extract LGSF. loss function optimize LGSF improved class separability inspired contrastive learning. We further enhanced introducing spatial relation designed constructed using dilated was evaluated four datasets, results highlighted its comprehensive well effectiveness