A novel dynamic classifier selection algorithm using spatial-spectral information for hyperspectral classification
-
摘要: 针对高光谱遥感影像分类面临的小样本、分类器不稳定等问题,在总结现有多分类器动态集成算法的基础上,提出了一种利用空间和光谱信息的多分类器动态集成算法.首先,采用支持向量机等5个基分类器构建多分类器集合;其次,计算各个分类器的分类结果,将大多数分类器分类一致的像元列入样本数据;最后,根据待分类像元的邻域像元的标签分类情况,动态地选择合适的方式进行分类器集成.该算法只在空间邻域信息满足一定条件的情况下,才采用空间和光谱信息结合的方法进行处理,即利用空间信息提高算法的灵活性.采用2幅不同传感器的高光谱遥感影像数据对算法进行实验,并与现有5种多分类器动态集成算法进行对比分析.结果表明,本文提出的多分类器动态集成算法可以保持较高的分类精度,并能有效提升高光谱遥感影像分类的稳定性,对于推动高光谱遥感精细分类研究具有一定的理论和实用价值.Abstract: To further improve the classification accuracy of hyperspectral remotely sensed imagery, this paper proposes a novel dynamic classifier selection algorithm, in which spatial and spectral information is used.The class labels of unlabeled pixels are predicted based on the percentage of their classified neighbors.The experiment is conducted between the proposed DCS-SSI algorithm and five dynamic classifier selection algorithms, and the results show that the proposed DCS-SSI algorithm can improve the robustness of classification performance for hyperspectral image analysis, which would be useful for high level classification of hyperspectral remote sensing images.
-
-
计量
- 文章访问数: 412
- PDF下载数: 24
- 施引文献: 0
下载: