Spectral-Spatial Hyperspectral Image Classification Based on Mathematical Morphology Post-Processing

Hyperspectral remote sensing sensors can provide plenty of valuable information. Fusion of spectral and spatial information plays a key role in the field of HyperSpectral Image (HSI) classification. In this paper, a novel two stages spectral-spatial HSI classification method based on Mathematical Morphology (MM) post-processing is proposed. In first stage, Support Vector Machine (SVM) is adopted to obtain the initial classification results. In second stage, in order to remove salt and pepper noise, MM is used to refine the obtained results of above stage. Experiments are conducted on the Indian Pines dataset. The evaluation results show that the proposed approach achieves better accuracy than several recently proposed post-processing HSI classification methods.