Multi-Scale Hybrid Spectral Network for Feature Learning and Hyperspectral Image Classification

Hyperspectral image (HSI) classification is an important concern in remote sensing, but it complex since few numbers of labelled training samples and the high-dimensional space with many spectral bands. Hence, essential to develop a more efficient neural network architecture improve performance HSI task. Deep learning models are contemporary techniques for pixel-based hyperspectral classification. feature extraction from both spatial channels has led high accuracy. Meanwhile, effectiveness these spatial-spectral methods relies on dimension every patch, there no feasible method determine best take into consideration. It makes better sense retrieve properties through examination at different neighborhood scales dimensions. In this context, paper presents multi-scale hybrid convolutional (MS-HybSN) model that uses three distinct spectral-spatial patches pull out domains. The presented deep framework sizes find possible features. process Hybrid convolution operation (3D-2D) done each selected patch repeated throughout image. To assess model, benchmark datasets openly accessible (Pavia University, Indian Pines, Salinas) new (Ahmedabad-1 Ahmedabad-2) being used experimental studies. Empirically, been demonstrated succeeds over remaining state-of-the-art approaches terms performance.