[An improved N-FINDR endmember extraction algorithm based on manifold learning and spatial information]

Area of Science:

• Remote Sensing
• Geospatial Analysis
• Signal Processing

Context:

• Hyperspectral imaging generates high-dimensional data, posing challenges for accurate material identification.
• Traditional endmember extraction algorithms often struggle with nonlinear spectral mixing and spatial data integration.
• Existing methods like GSVM, VCA, and SPPNFINDR have limitations in complex hyperspectral scenarios.

Purpose:

• To develop an improved N-FINDR endmember extraction algorithm by combining manifold learning and spatial information.
• To address the challenges of nonlinear mixing and high dimensionality in hyperspectral data.
• To enhance the precision and robustness of endmember extraction.

Summary:

• The proposed algorithm utilizes adaptive local tangent space alignment to reduce data dimensionality while preserving intrinsic structures.
• Spatial preprocessing enhances pixel vectors in homogeneous areas, leveraging spatial continuity for improved accuracy.
• Endmembers are extracted by maximizing simplex volume, effectively handling nonlinear spectral mixtures.

Impact:

• The enhanced algorithm significantly increases the precision of endmember extraction compared to GSVM, VCA, and SPPNFINDR.
• Demonstrated superior performance on both simulated and real hyperspectral datasets.
• Offers a more robust solution for analyzing complex hyperspectral imagery in various applications.