稀疏低秩核非负张量分解高光谱图像解混

Sparse low-rank kernel non-negative tensor decomposition hyperspectral image unmixing

为了解决传统非负张量分解过程中出现大维度矩阵导致计算速度变慢的问题,本文提出了一种基于稀疏低秩约束的核非负张量分解算法。该算法将张量数据映射到核空间内,添加稀疏约束结构,通过交替方向乘子法进行模型求解,并且考虑张量数据的非零元素及其位置,降低了算法计算的复杂度。在模拟和真实数据集上的实验验证了该算法的稳定性和有效性。

In order to solve the problem that large dimensional matrices slow down the calculation speed in the traditional non-negative tensor decomposition process,this paper proposes a kernel non-negative tensor decomposition algorithm based on sparse low-rank constraints.It first maps the tensor data into the kernel space,and then adds a sparse constraint structure,solving the model by alternating direction multiplier,and considering the non-zero elements of the tensor data and their positions in the process,which reduces the complexity of the algorithm calculation.Finally,the stability and effectiveness of the algorithm are verified on simulated and real datasets.