中文摘要：

针对影像分割所需区域内部满足一致性、区域间互不相交的要求,鉴于高光谱影像地物在尺寸、形状、光谱上的异质性,开展高光谱影像分割研究。应用视觉皮层细胞神经元之间存在的局部兴奋全局抑制振荡网络对视觉影像信息进行深入提取和处理,在此基础上,结合小世界神经网络高群集系数、短特征路长的特点,研究了LEGION神经元振荡器所具有的小世界神经网络的同步性能,从而构建了耦合小世界网络的LEGION分割算法。进一步采用直观参数设置,简化高微分方程的计算复杂,减少迭代次数。实验表明：耦合小世界网络的LEGION分割算法,可有效地把高光谱影像中同质地物分割在一起,达到信息提取的目的。

英文摘要：

Consistency and inter-disjoint are requirements of image segmentation. In order to meet these requirements, locally excitatory globally inhibitory oscillator network （LEGION）, a neural oscillator network based on biologically framework, has been extended for HSI segmentation, which is based on band selection and texture analysis. In LEGION, each oscillator has excitatory lateral connections to the oscillators in its local neighborhood, as well as a connection to a global inhibitor. According to the validation of image segmentation, extended LEGION presents promising potential for HSI segmentation. Recently, it is shown that many biological neural networks are typical small-world networks. Thus, small-world model derived from the well-known LEGION model has been proposed and investigated. To form a small-world network, a proper proportion of unidirectional shortcuts （ran- dom long-range connections） have been added to the original LEGION model. With local connections and short- cuts, the neural oscillators can not only communicate with neighbors but also exchange phase information with re- mote partners. It introdHces that excitatory shortcuts can enhance the synchronization within an oscillator group re- presenting the same obje t. The original complex high differential equations have been simplified by using parameter settings. And experiment results indicated that the proposed small-world models could achieve synchronization fas- ter than the original LEGION model and are more likely to bind disconnected image regions belonging together.