Collaborative representation based on enhanced tensor robust PCA for hyperspectral anomaly detection.

Area of Science:

• Remote Sensing
• Signal Processing
• Computer Vision

Background:

• Hyperspectral anomaly detection (HAD) is crucial for identifying targets distinct from their background.
• Existing methods often struggle to fully utilize spectral-spatial information and preserve important signal features.
• Noise and complex background clutter in hyperspectral images (HSIs) pose significant challenges.

Purpose of the Study:

• To propose a novel hyperspectral anomaly detection method, ETRPCA-CRD.
• To enhance the utilization of spectral-spatial information in HSIs.
• To improve the accuracy and robustness of anomaly detection.

Main Methods:

• Integration of enhanced tensor robust principal component analysis (ETRPCA) with collaborative representation detection (CRD).
• Utilizing weighted tensor Schatten-p norm minimization (WTSNM) for noise reduction and information preservation within ETRPCA.
• Employing Fourier transform, generalized soft-thresholding (GST), and T-singular value decomposition (SVD) for efficient ETRPCA problem solving.

Main Results:

• ETRPCA-CRD demonstrates superior detection accuracy compared to state-of-the-art algorithms on multiple datasets.
• The method effectively separates anomalous targets from background data.
• Significant improvement in robustness and preservation of salient signals was observed.

Conclusions:

• The proposed ETRPCA-CRD method offers a powerful and effective solution for hyperspectral anomaly detection.
• The integration of WTSNM within ETRPCA significantly enhances background suppression and anomaly detection.
• The approach successfully leverages spectral-spatial information for improved performance in HSIs.