| Hi-index | 0.00 |
With the development of sensor technology, the spectral resolution of remote sensing image is continuously improved. The appearance of the hyperspectral remote sensing is a tremendous leap in the field of remote sensing. The increasing availability of hyperspectral data and image has enriched us with better and finer data and it also enable us a much stronger ability to identify features. However the approaches in the feature identify of hyperspectral images are not as successful as we thought. Too many bands and a large amount of data not only bring difficulties in data storage and transmission, but also bring new challenges in hyperspectral image processing technology, especially the hyperspectral image feature recognize. Band selection aims to recognize the features effectively. We should distinguish the features by utilizing their spectral curve properties. These curves are found to have important information to recognize the different land cover types. So it makes great sense to choose the best combination of many bands and form a new hyperspectral image space. This procedure is usually called features selection. Bhattacharyya-Distance is one of the commonly used methods. It is one kind of statistic distance. It can more reasonably measure the distance between different land types in super multi-dimensional space. The hyperspectral data used in this paper is obtained by the sensors OMIS (Operational Modular Imaging Spectrometer). In this paper, we propose a band selection method based on the Bhattacharyya distance. In the proposed method, we try to find the optimize band combination. We divide land types in the research area into five classes (the five classes are seawater, fishery, building, vegetation and crops); calculate the Bhattacharyya distance between the five class pairs. According the optimal band subset selected by the Bhattacharyya distance, we make a classification and evaluate the classification accuracy. Experimental results show that the proposed band selection method compares favorably with conventional methods.