一种针对高光谱遥感影像的波段选择方法

A method of band selection special for the hyperspectral remote sensing image

随着传感器技术的不断发展,高光谱遥感影像已经广泛应用于土地覆盖监测等诸多领域。高光谱遥感影像具有波段数目多、波段间相关性强等特点,因此在图像分类时需要有效的波段选择方法以提高遥感影像的使用效率。文中提出了一种针对高光谱遥感影像的波段选择方法,该方法首先使用信息散度描述波段间的相关性,通过构造信息散度矩阵对子空间进行划分。然后使用波段的信息量和Bhattacharyya距离构建适应度函数,并对粒子群算法中的惯性权值更新方式进行改进。通过对AVIRIS高光谱遥感图像进行实验证明,与现有算法相比文中算法具有更高的分类精度及更快的收敛速度。

With the constant development of sensor technology,the hyperspectral remote sensing image is widely applied in many fields such as the land cover monitoring. The hyperspectral remote sensing image is characterized by the adequate number of bands and strong correlation between bands. Thus it requires the effective selection of bands when handling the image classification,which can increase the operation efficiency of remote sensing image. This paper proposed a method of band selection special for the hyperspectral remote sensing image. Firstly,it made use of the correlation between bands described by the information divergence to divide the sub-space using the matrix of information divergence. Then it constructed the fitness function relying on the information content and Bhattacharyya distance. It also improved the update method of inertia weight in the particle swarm optimization algorithm. Compared with the existing algorithms,the experimental results of AVIRIS hyperspectral remote sensing image showed the higher classification accuracy and faster speed of convergence.