A local Mahalanobis-distance method based on tensor decomposition for hyperspectral anomaly detection

Recently, some tensor decomposition-based algorithms are proposed and performed well for hyperspectral anomaly detection (AD). This paper proposes a tensor decomposition-based local Mahalanobis-distance (Tensor-LMD) method for hyperspectral AD. First, a three-order tensor is employed to represent hyperspectral data-set and the Tucker decomposition technology is used to decompose such tensor into a core tensor and three factor matrices. Then, the minor PCs are used to eliminate anomaly and noise information along each mode and the more pure background data-set is obtained. Finally, the sliding dual-window strategy is used for both the background data-set and the original hyperspectral data-set, and the local Mahalanobis-distance detector is employed for the final results. The experimental results demonstrate that the proposed Tensor-LMD can achieve a better performance when compared with the comparison algorithms.