高光谱混合像元分解的稀疏优化算法

SPARSE OPTIMAL ALGORITHM FOR HYPERSPECTRAL PIXEL UNMIXING

提出一种参数自适应估计的高光谱混合像元分解算法。为混合像元分解问题建立新的约束优化模型,该模型的目标函数由L2误差项和Lp正则项构成。利用交替优化方法将模型分解为若干子问题,采用邻近算子方法求解这些子问题。在交替迭代的求解过程中,根据每次迭代的结果自适应地选择模型参数。从理论角度分析了算法的收敛性,并通过实验验证了所提算法的有效性。实验结果还表明,与经典的高光谱混合像元分解算法相比,所建立的模型及提出的求解算法可获得更佳的混合像元分解效果。

A hyperspectral pixel unmixing algorithm with adaptive parameter estimation is proposed. A novel constrained optimal model is created for the issue of pixel unmixing, the objective function of the model is composed of L~ error term and Lp regularised term. The alternating optimisation method is used to decompose the optimal model into several sub-problems, and the proximal operator method is employed to solve these sub-problems. During the alternating iterative procedure, the model parameters can be adaptively chosen according to each iterative result. The convergence of the proposed algorithm is theoretically analyzed, and the effectiveness of it is proved by the experiment. Compared with the classic hyperspeetral pixel unmixing algorithms, the experimental results also show that the created model and the proposed solution algorithm can achieve better effect of pixel unmixing.