基于樽海鞘群体优化非负矩阵分解的高光谱图像解混算法

Hyperspectral Images Unmixing Based on Nonnegative Matrix Factorization Optimized by Salp Swarm Algorithm

针对鲁棒非负矩阵分解应用于高光谱图像处理时,存在对初始值的敏感性,求解目标函数时易陷入局部最优的缺点,提出基于樽海鞘群体优化鲁棒非负矩阵分解的高光谱图像解混算法.该算法基于鲁棒线性混合模型,在RNMF框架下,采用樽海鞘群体算法取代乘法迭代策略,以增强算法全局搜索能力,在约束空间内随机搜索满足目标函数的全局最优解,可有效地完成非线性高光谱图像解混.仿真数据与真实遥感数据实验结果表明,本文算法在处理高光谱图像时,能够有效地避免RNMF算法易陷入局部最优解的局限性,具有更好的解混性能.

In order to solve the problem that robust nonnegative matrix factorization is sensitive to the initial values and usually trapped in a local optimum when applied to hyperspectral images processing,a new hyperspectral image unmixing algorithm based on robust nonnegative matrix factorization optimized by salp swarm algorithm is proposed.The algorithm is based on robust linear mixed model.Under the framework of robust nonnegative matrix factorization,the global search capability is enhanced by replacing the multiplicative iteration strategy with the salp swarm optimization algorithm,the global optimal solution satisfying the objective function is randomly searched in the constrained space.And then,the hyperspectral image unmixing is effectively fulfilled.The experimental results on synthetic data and real remote sensing data indicate that the algorithm proposed can effectively avoid the limitation of RNMF falling into the local optimal solution,which has better performance of unmixing.