Unsupervised Hyperspectral Band Selection via Structure-Conserved and Neighborhood-Grouped Evolutionary Algorithm

Hyperspectral images (HSIs) contain hundreds of bands, which provide a wealth of spectral information and enable better characterization of features. However, the excessive dimensions and redundant information also cause a dimensional disaster for subsequent processing. Band selection (BS) is a widely used dimension reduction technique for HSIs. Traditional methods mainly consider the hyperspectral BS problem at the level of data and maintain the information contained in the data, without considering the spatial structures inside HSIs. To fill the gap, in this work, an unsupervised hyperspectral BS method through structure-conserved and neighborhood-grouped evolutionary algorithm (SNEA) is proposed. Different from other EAs for hyperspectral BS, first, two spatial-structure-related optimization objectives are designed, including the locally spatial structure denoted by the pixel’s spatial consistency with its adjacent neighbors and the globally spatial structure denoted by the affinity graph among pixels. With the designed objectives, the hyperspectral BS is formulated as the problem of conserving spatial structures. Moreover, a neighborhood-grouped pairwise learning strategy is proposed to generate high-quality offsprings. In this novel strategy, a neighborhood-grouping operation is developed to divide the band space into several groups. The population can be initialized efficiently, and the offspring solutions can be generated pairwisely under the guidance of grouping. Compared with nine state-of-the-art comparison algorithms, experimental results on three standard hyperspectral datasets demonstrate that the band subset obtained by our proposed SNEA has a better classification performance than the comparison algorithms.