[A Hyperspectral Imagery Anomaly Detection Algorithm Based on Gauss-Markov Model].

Area of Science:

• Remote Sensing
• Image Processing
• Computer Vision

Context:

• Hyperspectral anomaly detection is crucial for remote sensing image analysis.
• Traditional RX anomaly detection algorithms struggle with spatial correlations and high dimensionality.
• Hyperspectral data often exhibits Gauss-Markov Random Field (GMRF) properties.

Purpose:

• To propose an improved RX anomaly detection algorithm leveraging 3D GMRF for hyperspectral imagery.
• To enhance detection efficiency and reduce false alarm rates compared to existing methods.
• To address the computational cost and validity issues of traditional algorithms.

Summary:

• A novel anomaly detection algorithm is developed based on a 3D GMRF model, simulating hyperspectral data and estimating GMRF parameters using the Approximated Maximum Likelihood method.
• The algorithm constructs a detection operator using GMRF parameters and calculates anomaly degrees within a moving GMRF detection window.
• Simulations on AVIRIS hyperspectral data demonstrate improved detection efficiency and a reduced false alarm rate.

Impact:

• The proposed algorithm significantly improves operational time by 45.2% over traditional methods, showcasing superior computing efficiency.
• This advancement offers a more effective and computationally efficient solution for hyperspectral anomaly detection in remote sensing.
• The findings contribute to more robust and accurate analysis of hyperspectral imagery.