Answers are grounded in this paper and cite their sources.
In hyperspectral anomaly detection (HAD), the collaborative-representation-based detector (CRD) has recently gained significant attention. Nevertheless, the CRD method's ability is constrained by the issue of background contamination. Meanwhile, the CRD method ignores the abundant spatial information of hyperspectral imagery. To address these two issues, this paper proposes a novel HAD algorithm called spectralspatial enhanced low-rank collaborative representation (SSELRCR). The SSELRCR algorithm first extracts the enhanced low-rank tensor as the updated background for CRD. In this way, the contamination of backgrounds can be suppressed by exploiting multidimensional prominent information embedded in background tensor data. Besides, the SSELRCR algorithm utilizes attribute filtering and differential operations to extract the attribute profiles (APs) of abnormal targets, based on their spatial distribution features. Next, the total variation curvature filter is applied to further smooth the APs of abnormal targets. Finally, the SSELRCR algorithm utilizes an adaptive fusion strategy that relies on the structure tensor to fuse spatial and spectral detection maps without manually setting fusion weights. The performance of the SSELRCR algorithm is assessed using six hyperspectral datasets. Experimental results demonstrate that the SSELRCR method outperforms the other nine methods for HAD.
No takes yet. Share an insight, caveat, or question.
Li et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: