Speeding Up and Enhancing the Hyperspectral Images Classification

Hyperspectral images (HSI) are adjacent band images generally utilized in many different real-life applications. HSI is rich in spatial and spectral information, providing very accurate information. However, their high dimensionality is the biggest challenge in the classification process time. This study provided the solution to handle the high HSI dimensionality and speed up its classification operation with increased accuracy. The model of this study used the Exponential linear units (ELU) activation function to solve the problems of dying ReLu and the vanishing gradient. Moreover, it used QPCA to enhance dimensional reduction. The proposed model is ELUSNet. It provided the best results and time compared to five other known models with two different HSI datasets.