Document Type: Research Paper
Hyperspectral data potentially contain more information than multispectral data because of their higher spectral resolution. However, the stochastic data analysis approaches that have been successfully applied to multispectral data are not as effective for hyperspectral data as well. Various investigations indicate that the key problem that causes poor performance in the stochastic approaches to hyperspectral data classification is inaccurate class parameters estimation. It has been found that the conventional approaches can be retained if a preprocessing stage is established before feature extraction procedure in classification of hyperspectral data. For preprocessing stage it has been proposed two steps in this paper including dimensionality reduction and class separability improvement. Sequential Parametric Projection Pursuit was used for dimensionality reduction because of its special characteristics. Projection Pursuit algorithm performs the computation of class parameter estimation at a lower dimensional space, giving better parameter estimation. For class separability improvement a lowpass filter has been used after dimensionality reduction. This paper shows that for different number of features, classification accuracy is improved when the preprocessing stage is applied.