| Popis: |
In order to improve the Hyperspectral images (HSI) classification accuracy and to preprocess HSI by fully using the spatial and spectral information, a new spatial-spectral dimensionality reduction method, Weighted Spatial and Spectral Principle Component Analysis (WSSPCA), was proposed. This algorithm reconstructed the HSI by using the physical characters of HSI. Then, principle Component Analysis (PCA) was utilized to reduce dimensionality of HSI. The new method not only can lower the influence of singular point in HSI, but also reduce the redundancy between bands, which improves the HSI classification accuracy efficiently. The benchmark tests on Indian Pines and PaviaU demonstrate that the performance of WSSPCA is better than PCA and LPP when 10%, 5% samples in each class. The best values of kappa coefficient obtained by WSSPCA are 89.61% and 95.59% respectively on the HSI datasets, exceeding the baseline 20.5% and 19.38%. |
