一种基于空间连续性的高光谱图像分类方法

利用地物在空间上分布连续性这一信息，提出了一种高光谱图像先分块、再分类的分类方法，改进了通常意义下的分类只考虑光谱信息这一缺陷，通过对2000年我们在日本长野南牧村，采用上海技术物理研究所研制的推扫式成像光谱仪(PHI)获得了80个波段的高光谱数据进行实验，验证了我们算法的优越性．

CLASSIFICATION ALGORITHM BASED ON SPATIAL CONTINUITY FOR HYPERSPECTRAL IMAGE

Based on the spatial continuity of ground objects, a new classification algorithm was proposed by two steps, namely blocking images followed by further classification, which improved the disadvantage of the general classification algorithms constrained by only spectral information. The new algorithm proposed in this study was validated by the 80-band hyperspectral image acquired by the push broom hyperspectral imager(PHI) in Minamimaki of Japan, 2000. The PHI sensor was developed by the Shanghai Institute of Technical Physics. It turned out that our algorithm embodied the advantage in classifying hyperspectral image.