高光谱数据的降维及Tabu搜索算法的应用

Dimensionality Reduction in Hyperspectral Classification and the Application of Tabu Search Algorithm

在高光谱影像的分类过程中,如何有效地降低特征空间的维数,又能保证原始数据所包含的丰富地物信息是一项十分重要而繁琐的工作。深入分析了这种降维的必要性,并针对当前常用的降维方法存在的问题,提出了运用Tabu搜索算法获取对分类最为有利的特征波段的思想。考虑到高光谱数据的特点,指出了算法运行中应该注意的若干关键参数设置问题。实验表明,Tabu搜索算法在求解质量和执行效率方面都有着良好的表现,可以用于高光谱数据的降维处理。

In hyperspectral classification, it is very important, also inconvenient, to reduce the number of input dimensionality effectively and meanwhile maintain the information contained in original hyperspectral data. The paper analyzed the necessity of dimensionality reduction in depth. After the presentation of the traditional reduction algorithms, the Tabu search algorithm was proposed for the dimensionality reduction purpose, which could overcome the limitations of the current reduction algorithms to some extent. In consideration of the complexity of the hyperspetral data, the problem of providing several key elements with appropriate values was also referred. The experiment demonstrated the good performance of Tabu search algorithm in hyperspectral imensionality reduction.