光谱异质性参数优化方法及其在QuickBird影像中的应用

提出了一种光谱异质性参数优化方法,与现有方法相比,新方法的改进主要有:①在设置参与分割的波段数据时,通过引入派生波段数据来增强分割判据的数据支撑；②在设置各波段数据层的权值时,综合考虑了各波段数据层包含的信息量和相互间的信息相关度,以抑制波段数据层间的信息冗余对分割过程的不利影响.最后,通过分割对比实验证明:采用新方法设置的光谱异质性参数执行影像分割,“欠分割”现象有了较大幅度地减少,对象间的差异程度和对象内部的均质程度都得到了提高,分割质量有了明显地改善.

An Optimization Method of the Spectral Heterogeneity Parameters and Its Application in QuickBird Imagery

An optimization method of the spectral heterogeneity parameters was proposed in this paper.Compared to the existing method,the advantages of the proposed method were included:(1)data support on the segmentation criterion would be enhanced by adding the derived bands data during setting the data layers which involved in the imagery segmentation;(2)in order to control the adverse effect of information redundancy between the band layers for the segmentation process,not only the information of each band layer but also the relevance between band layers were taken into account during setting the weight of each band layer.Finally,the results of experiments showed that compared to the segmentation result of the existing method,imagery was segmented according to spectral heterogeneity parameters set by the proposed method,less segmentation phenomenon had been greatly reduced.Meanwhile,the difference between objects and the homogeneity of each object were significantly improved,the quality of segmentation was better than before.