一种高光谱图像波段选择的快速混合搜索算法

A fast hybrid search algorithm for band selection in hyperspectral images

由于高光谱图像的高数据维和大数据量,现有的波段选择方法大多不能同时具有良好的效果和较短的计算时间。提出了一种用于高光谱图像波段选择的新方法——快速混合搜索算法。该算法将全局搜索和局部寻优有机的结合起来,能够在较短的时间内获得最佳的波段组合,用于高光谱图像的目标分类识别。快速混合搜索算法克服了传统搜索方法在高光谱图像波段选择中的缺陷,能够在提高所选波段性能的同时节省大量的运算时间。分别利用200波段和126波段的AVIRIS对其数据进行了仿真实验。实验结果表明,快速混合搜索算法在所选波段性能和计算耗时方面都获得了令人满意的效果。

At present, most algorithms for band selection in hyperspectral images can＇t hold both excellent behavior and low computation load because of huge data amount and high dimension. A new algorithm for hand selection, fast hybrid search algorithm, is proposed. The proposed algorithm combines global and local search together, and can select the best hands in a short time. Fast hybrid search algorithm conquers the limitations of traditional band selection algorithms, and its characteristic of global and local search ensures that it can improve the performance of the selected bands and reduce computation load at the same time. Numerical experiments are conducted on AVIRIS data with 200 hands and 126 bands, and the results show that the new algorithm reaches satisfying effects in both performance and computation time.