中文摘要：

高光谱遥感使宽波段遥感中不可探测的物质可以被探测，成为了遥感界的一场新的革命．由于高光谱遥感图像波段多、光谱分辨率高、数据量庞大，给高光谱遥感数据实际应用分析带来极大不便．以特征选择为目的，以协方差矩阵特征值法为评价算法，设计实现了基于遗传算法和差分演化算法的降维过程．通过与传统的序列向前搜索的特征选择进行对比实验，比照搜索结果和算法耗时，验证了演化算法在特征选择的实现过程中具有良好的性能，证明了演化算法在高光谱图像降维中的实用价值．其中差分演化算法搜索结果十分稳定，可以替代完全搜索来寻找最优解．

英文摘要：

Hyperspectral remote sensing makes the wideband remote sensing can be detected. However its high bands, high spectral resolution and large amounts of data bring great inconvenience. This article treats feature selection as purpose, uses covariance matrix estimation as an evaluation method and achieves dimensionality reduction based on two kinds of evolutionary algorithms： GA and DE. The experiments reveal that the differential evolution algorithm performs better than traditional sequence of forward algorithm in both search results and time-consuming, and it finally proves the practical value and significance of evolutionary algorithm used in hyper-spectral image dimensionality reduction. The results of DE are much more stable and can replace full search to get optimal solution.