Abstract

MEDJAHED, Seyyid Ahmed  and  OUALI, Mohammed. A Hybrid Approach for Supervised Spectral Band Selection in Hyperspectral Images Classification. Comp. y Sist. [online]. 2020, vol.24, n.1, pp.213-219.  Epub Sep 27, 2021. ISSN 2007-9737.  https://doi.org/10.13053/cys-24-1-3017.

Recently, hyperspectral imagery has been very active research field in many applications of remote sensing. Unfortunately, the large number of bands reduces the classification accuracy and computational complexity which causes the Hugh phenomenon. In this paper, a new hybrid approach for band selection based is proposed. This approach combines the advantage of filter and wrapper method. The proposed approach is composed of two phases: the first phase consists to reduce the number of bands by merging the highly correlated bands, and, the second phase uses a wrapper approach based on Sin Cosine Algorithm to select the optimal band subset that provides a high classification accuracy. In addition, a new binary version of Sin Cosine Algorithm is proposed to adapt it to the band selection problem. The performance evaluation of the proposed approach is tested on three publicly available benchmark hyperspectral images. The analysis of the results demonstrates the efficiency and performance of the proposed approach.

Keywords : Spectral band selection; hyperspectral image; classification; sin cosine algorithm; optimization.