Hyper-spectral remote sensing data can be acquired by imaging the same area with multiple wavelengths, and it normally consists of hundreds of band-images. Hyper-spectral images can not only provide spatial information but also high resolution spectral information, and it has been widely used in environment monitoring, mineral investigation and military reconnaissance. However, because of the corresponding large data volume, it is very difficult to transmit and store Hyper-spectral images. Hyper-spectral image dimensional reduction technique is desired to resolve this problem. Because of the High relation and high redundancy of the hyper-spectral bands, it is very feasible that applying the dimensional reduction method to compress the data volume. This paper proposed a novel band selection-based dimension reduction method which can adaptively select the bands which contain more information and details. The proposed method is based on the principal component analysis (PCA), and then computes the index corresponding to every band. The indexes obtained are then ranked in order of magnitude from large to small. Based on the threshold, system can adaptively and reasonably select the bands. The proposed method can overcome the shortcomings induced by transform-based dimension reduction method and prevent the original spectral information from being lost. The performance of the proposed method has been validated by implementing several experiments. The experimental results show that the proposed algorithm can reduce the dimensions of hyper-spectral image with little information loss by adaptively selecting the band images.

Date Published: 20 November 2014
PDF: 11 pages
Proc. SPIE 9300, International Symposium on Optoelectronic Technology and Application 2014: Infrared Technology and Applications, 930006 (20 November 2014); doi: 10.1117/12.2074382

Published in SPIE Proceedings Vol. 9300:
International Symposium on Optoelectronic Technology and Application 2014: Infrared Technology and Applications
Mircea Guina; Haimei Gong; Zhichuan Niu; Jin Lu, Editor(s)